Support ZSTD dictionary compression

patch by Yifan Cai; reviewed by Jon Haddad, Stefan Miklosovic for CASSANDRA-17021

CHANGES.txt

@@ -1,4 +1,5 @@
 5.1
+ * Support ZSTD dictionary compression (CASSANDRA-17021)
  * Fix ExceptionsTable when stacktrace has zero elements (CASSANDRA-20992)
  * Replace blocking wait with non-blocking delay in paxos repair (CASSANDRA-20983)
  * Implementation of CEP-55 - Generation of role names (CASSANDRA-20897)

conf/cassandra.yaml

@@ -2870,3 +2870,49 @@ storage_compatibility_mode: NONE
 #         # especially in keyspaces with many tables. The splitter avoids batching tables together if they
 #         # exceed other configuration parameters like bytes_per_assignment or partitions_per_assignment.
 #         max_tables_per_assignment: 64
+
+# Dictionary compression settings for ZSTD dictionary-based compression
+# These settings control the automatic training and caching of compression dictionaries
+# for tables that use ZSTD dictionary compression.
+
+# How often to refresh compression dictionaries across the cluster.
+# During refresh, nodes will check for newer dictionary versions and update their caches.
+# Min unit: s
+compression_dictionary_refresh_interval: 3600s
+
+# Initial delay before starting the first dictionary refresh cycle after node startup.
+# This prevents all nodes from refreshing simultaneously when the cluster starts.
+# Min unit: s
+compression_dictionary_refresh_initial_delay: 10s
+
+# Maximum number of compression dictionaries to cache per table.
+# Each table using dictionary compression can have multiple dictionaries cached
+# (current version plus recently used versions for reading older SSTables).
+compression_dictionary_cache_size: 10
+
+# How long to keep compression dictionaries in the cache before they expire.
+# Expired dictionaries will be removed from memory but can be reloaded if needed.
+# Min unit: s
+compression_dictionary_cache_expire: 24h
+
+# Dictionary training configuration (advanced settings)
+# These settings control how compression dictionaries are trained from sample data.
+
+# Maximum size of a trained compression dictionary.
+# Larger dictionaries may provide better compression but use more memory.
+compression_dictionary_training_max_dictionary_size: 64KiB
+
+# Maximum total size of sample data to collect for dictionary training.
+# More sample data generally produces better dictionaries but takes longer to train.
+# The recommended sample size is 100x the dictionary size.
+compression_dictionary_training_max_total_sample_size: 10MiB
+
+# Enable automatic dictionary training based on sampling of write operations.
+# When enabled, the system will automatically collect samples and train new dictionaries.
+# Manual training via nodetool is always available regardless of this setting.
+compression_dictionary_training_auto_train_enabled: false
+
+# Sampling rate for automatic dictionary training (1-10000).
+# Value of 100 means 1% of writes are sampled. Lower values reduce overhead but may
+# result in less representative sample data for dictionary training.
+compression_dictionary_training_sampling_rate: 0.01

conf/cassandra_latest.yaml

@@ -2621,3 +2621,49 @@ storage_compatibility_mode: NONE
 #         # especially in keyspaces with many tables. The splitter avoids batching tables together if they
 #         # exceed other configuration parameters like bytes_per_assignment or partitions_per_assignment.
 #         max_tables_per_assignment: 64
+
+# Dictionary compression settings for ZSTD dictionary-based compression
+# These settings control the automatic training and caching of compression dictionaries
+# for tables that use ZSTD dictionary compression.
+
+# How often to refresh compression dictionaries across the cluster.
+# During refresh, nodes will check for newer dictionary versions and update their caches.
+# Min unit: s
+compression_dictionary_refresh_interval: 3600s
+
+# Initial delay before starting the first dictionary refresh cycle after node startup.
+# This prevents all nodes from refreshing simultaneously when the cluster starts.
+# Min unit: s
+compression_dictionary_refresh_initial_delay: 10s
+
+# Maximum number of compression dictionaries to cache per table.
+# Each table using dictionary compression can have multiple dictionaries cached
+# (current version plus recently used versions for reading older SSTables).
+compression_dictionary_cache_size: 10
+
+# How long to keep compression dictionaries in the cache before they expire.
+# Expired dictionaries will be removed from memory but can be reloaded if needed.
+# Min unit: s
+compression_dictionary_cache_expire: 24h
+
+# Dictionary training configuration (advanced settings)
+# These settings control how compression dictionaries are trained from sample data.
+
+# Maximum size of a trained compression dictionary.
+# Larger dictionaries may provide better compression but use more memory.
+compression_dictionary_training_max_dictionary_size: 64KiB
+
+# Maximum total size of sample data to collect for dictionary training.
+# More sample data generally produces better dictionaries but takes longer to train.
+# The recommended sample size is 100x the dictionary size.
+compression_dictionary_training_max_total_sample_size: 10MiB
+
+# Enable automatic dictionary training based on sampling of write operations.
+# When enabled, the system will automatically collect samples and train new dictionaries.
+# Manual training via nodetool is always available regardless of this setting.
+compression_dictionary_training_auto_train_enabled: false
+
+# Sampling rate for automatic dictionary training (1-10000).
+# Value of 100 means 1% of writes are sampled. Lower values reduce overhead but may
+# result in less representative sample data for dictionary training.
+compression_dictionary_training_sampling_rate: 0.01

doc/modules/cassandra/pages/managing/operating/compression.adoc

@@ -49,6 +49,8 @@ these areas (A is relatively good, F is relatively bad):
 
 |https://facebook.github.io/zstd/[Zstd] |`ZstdCompressor` | A- | A- | A+ | `>= 4.0`
 
+|https://facebook.github.io/zstd/[Zstd with Dictionary] |`ZstdDictionaryCompressor` | A- | A- | A++ | `>= 6.0`
+
 |http://google.github.io/snappy/[Snappy] |`SnappyCompressor` | A- | A | C | `>= 1.0`
 
 |https://zlib.net[Deflate (zlib)] |`DeflateCompressor` | C | C | A | `>= 1.0`
@@ -60,6 +62,9 @@ cycle spent. This is why it is the default choice in Cassandra.
 
 For storage critical applications (disk footprint), however, `Zstd` may
 be a better choice as it can get significant additional ratio to `LZ4`.
+For workloads with highly repetitive or similar data patterns,
+`ZstdDictionaryCompressor` can achieve even better compression ratios by
+training a compression dictionary on representative data samples.
 
 `Snappy` is kept for backwards compatibility and `LZ4` will typically be
 preferable.
@@ -67,6 +72,102 @@ preferable.
 `Deflate` is kept for backwards compatibility and `Zstd` will typically
 be preferable.
 
+== ZSTD Dictionary Compression
+
+The `ZstdDictionaryCompressor` extends standard ZSTD compression by using
+trained compression dictionaries to achieve superior compression ratios,
+particularly for workloads with repetitive or similar data patterns.
+
+=== How Dictionary Compression Works
+
+Dictionary compression improves upon standard compression by training a
+compression dictionary on representative samples of your data. This
+dictionary captures common patterns, repeated strings, and data structures,
+allowing the compressor to reference these patterns more efficiently than
+discovering them independently in each compression chunk.
+
+=== When to Use Dictionary Compression
+
+Dictionary compression is most effective for:
+
+* *Tables with similar row structures*: JSON documents, XML data, or
+repeated data schemas benefit significantly from dictionary compression.
+* *Storage-critical workloads*: When disk space savings justify the
+additional operational overhead of dictionary training and management.
+* *Large datasets with repetitive patterns*: The more similar your data,
+the better the compression ratio improvement.
+
+Dictionary compression may not be ideal for:
+
+* *Highly random or unique data*: Already-compressed data or cryptographic
+data will see minimal benefit.
+* *Small tables*: The overhead of dictionary management may outweigh the
+storage savings.
+* *Frequently changing schemas*: Schema changes may require retraining
+dictionaries to maintain optimal compression ratios.
+
+=== Dictionary Training
+
+Before dictionary compression can provide optimal results, a compression
+dictionary must be trained on representative data samples. Cassandra
+supports both manual and automatic training approaches.
+
+==== Manual Dictionary Training
+
+Use the `nodetool compressiondictionary train` command to manually train
+a compression dictionary:
+
+[source,bash]
+----
+nodetool compressiondictionary train <keyspace> <table>
+----
+
+The command trains a dictionary by sampling from existing SSTables. If no
+SSTables are available on disk (e.g., all data is in memtables), the command
+will automatically flush the memtable before sampling.
+
+The training process completes synchronously and displays progress information
+including sample count, sample size, and elapsed time. Training typically
+completes within minutes for most workloads.
+
+By default, training will only proceed if enough samples have been collected.
+To force training even with insufficient samples, use the `--force` or `-f` option:
+
+[source,bash]
+----
+nodetool compressiondictionary train --force <keyspace> <table>
+----
+
+This can be useful for testing or when you want to train a dictionary from
+limited data during initial setup.
+
+==== Automatic Dictionary Training
+
+Enable automatic training in `cassandra.yaml`:
+
+[source,yaml]
+----
+compression_dictionary_training_auto_train_enabled: true
+compression_dictionary_training_sampling_rate: 100  # 1% of writes
+----
+
+When enabled, Cassandra automatically samples write operations and
+trains dictionaries in the background based on the configured sampling
+rate (range: 1-10000, where 100 = 1% of writes).
+
+=== Dictionary Storage and Distribution
+
+Compression dictionaries are stored cluster-wide in the
+`system_distributed.compression_dictionaries` table. Each table can
+maintain multiple dictionary versions: the current dictionary for
+compressing new SSTables, plus historical dictionaries needed for
+reading older SSTables.
+
+Dictionaries are identified by `dict_id`, with higher IDs representing
+newer dictionaries. Cassandra automatically refreshes dictionaries
+across the cluster based on configured intervals, and caches them
+locally to minimize lookup overhead.
+
 == Configuring Compression
 
 Compression is configured on a per-table basis as an optional argument
@@ -105,6 +206,17 @@ should be used with caution, as they require more memory. The default of
 `3` is a good choice for competing with `Deflate` ratios and `1` is a
 good choice for competing with `LZ4`.
 
+The `ZstdDictionaryCompressor` supports the same options as
+`ZstdCompressor`:
+
+* `compression_level` (default `3`): Same range and behavior as
+`ZstdCompressor`. Dictionary compression provides improved ratios at
+any compression level compared to standard ZSTD.
+
+NOTE: `ZstdDictionaryCompressor` requires a trained compression
+dictionary to achieve optimal results. See the ZSTD Dictionary
+Compression section above for training instructions.
+
 Users can set compression using the following syntax:
 
 [source,cql]
@@ -121,6 +233,25 @@ ALTER TABLE keyspace.table
    WITH compression = {'class': 'LZ4Compressor', 'chunk_length_in_kb': 64};
 ----
 
+For dictionary compression:
+
+[source,cql]
+----
+CREATE TABLE keyspace.table (id int PRIMARY KEY)
+   WITH compression = {'class': 'ZstdDictionaryCompressor'};
+----
+
+Or with a specific compression level:
+
+[source,cql]
+----
+ALTER TABLE keyspace.table
+   WITH compression = {
+       'class': 'ZstdDictionaryCompressor',
+       'compression_level': '3'
+   };
+----
+
 Once enabled, compression can be disabled with `ALTER TABLE` setting
 `enabled` to `false`:
 
@@ -140,6 +271,63 @@ immediately, the operator can trigger an SSTable rewrite using
 `nodetool scrub` or `nodetool upgradesstables -a`, both of which will
 rebuild the SSTables on disk, re-compressing the data in the process.
 
+== Dictionary Compression Configuration
+
+When using `ZstdDictionaryCompressor`, several additional configuration
+options are available in `cassandra.yaml` to control dictionary
+management, caching, and training behavior.
+
+=== Dictionary Refresh Settings
+
+* `compression_dictionary_refresh_interval` (default: `3600`): How often
+(in seconds) to check for and refresh compression dictionaries
+cluster-wide. Newly trained dictionaries will be picked up by all nodes
+within this interval.
+* `compression_dictionary_refresh_initial_delay` (default: `10`): Initial
+delay (in seconds) before the first dictionary refresh check after node
+startup.
+
+=== Dictionary Caching
+
+* `compression_dictionary_cache_size` (default: `10`): Maximum number of
+compression dictionaries to cache per table. Higher values reduce lookup
+overhead but increase memory usage.
+* `compression_dictionary_cache_expire` (default: `3600`): Dictionary
+cache entry TTL in seconds. Expired entries are evicted and reloaded on
+next access.
+
+=== Training Configuration
+
+* `compression_dictionary_training_max_dictionary_size` (default: `65536`):
+Maximum size of trained dictionaries in bytes. Larger dictionaries can
+capture more patterns but increase memory overhead.
+* `compression_dictionary_training_max_total_sample_size` (default:
+`10485760`): Maximum total size of sample data to collect for training,
+approximately 10MB.
+* `compression_dictionary_training_auto_train_enabled` (default: `false`):
+Enable automatic background dictionary training. When enabled, Cassandra
+samples writes and trains dictionaries automatically.
+* `compression_dictionary_training_sampling_rate` (default: `100`):
+Sampling rate for automatic training, range 1-10000 where 100 = 1% of
+writes. Lower values reduce training overhead but may miss data patterns.
+
+Example configuration:
+
+[source,yaml]
+----
+# Dictionary refresh and caching
+compression_dictionary_refresh_interval: 3600
+compression_dictionary_refresh_initial_delay: 10
+compression_dictionary_cache_size: 10
+compression_dictionary_cache_expire: 3600
+
+# Automatic training
+compression_dictionary_training_auto_train_enabled: false
+compression_dictionary_training_sampling_rate: 100
+compression_dictionary_training_max_dictionary_size: 65536
+compression_dictionary_training_max_total_sample_size: 10485760
+----
+
 == Other options
 
 * `crc_check_chance` (default: `1.0`): determines how likely Cassandra
@@ -186,6 +374,39 @@ correctness of data on disk, compressed tables allow the user to set
 probabilistically validate chunks on read to verify bits on disk are not
 corrupt.
 
+=== Dictionary Compression Operational Considerations
+
+When using `ZstdDictionaryCompressor`, additional operational factors
+apply:
+
+* *Dictionary Storage*: Compression dictionaries are stored in the
+`system_distributed.compression_dictionaries` table and replicated
+cluster-wide. Each table maintains current and historical dictionary
+versions.
+* *Dictionary Cache Memory*: Dictionaries are cached locally on each node
+according to `compression_dictionary_cache_size`. Memory overhead is
+typically minimal (default 64KB per dictionary × cache size).
+* *Dictionary Training Overhead*: Manual training via
+`nodetool compressiondictionary train` samples SSTable chunk data and
+performs CPU-intensive dictionary training. Consider running training
+during off-peak hours.
+* *Automatic Training Impact*: When
+`compression_dictionary_training_auto_train_enabled` is true, write
+operations are sampled based on `compression_dictionary_training_sampling_rate`.
+This adds minimal overhead but should be monitored in write-intensive
+workloads.
+* *Dictionary Refresh*: The dictionary refresh process
+(`compression_dictionary_refresh_interval`) checks for new dictionaries
+cluster-wide. The default 1-hour interval balances freshness with
+overhead.
+* *SSTable Compatibility*: Each SSTable is compressed with a specific
+dictionary version. Historical dictionaries must be retained to read
+older SSTables until they are compacted with new dictionaries.
+* *Schema Changes*: Significant schema changes or data pattern shifts may
+require retraining dictionaries to maintain optimal compression ratios.
+Monitor the `SSTable Compression Ratio` via `nodetool tablestats` to
+detect degradation.
+
 == Advanced Use
 
 Advanced users can provide their own compression class by implementing

pylib/cqlshlib/cqlhandling.py

@@ -44,6 +44,7 @@ class CqlParsingRuleSet(pylexotron.ParsingRuleSet):
         'SnappyCompressor',
         'LZ4Compressor',
         'ZstdCompressor',
+        'ZstdDictionaryCompressor'
     )
 
     available_compaction_classes = (

src/java/org/apache/cassandra/config/Config.java

@@ -514,6 +514,17 @@ public class Config
     public volatile DurationSpec.IntSecondsBound counter_cache_save_period = new DurationSpec.IntSecondsBound("7200s");
     public volatile int counter_cache_keys_to_save = Integer.MAX_VALUE;
 
+    public volatile DurationSpec.IntSecondsBound compression_dictionary_refresh_interval = new DurationSpec.IntSecondsBound("3600s"); // 1 hour - TODO: re-assess whether daily (86400s) is more appropriate
+    public volatile DurationSpec.IntSecondsBound compression_dictionary_refresh_initial_delay = new DurationSpec.IntSecondsBound("10s"); // 10 seconds default
+    public volatile int compression_dictionary_cache_size = 10; // max dictionaries per table
+    public volatile DurationSpec.IntSecondsBound compression_dictionary_cache_expire = new DurationSpec.IntSecondsBound("24h");
+
+    // Dictionary training settings
+    public volatile DataStorageSpec.IntKibibytesBound compression_dictionary_training_max_dictionary_size = new DataStorageSpec.IntKibibytesBound("64KiB");
+    public volatile DataStorageSpec.IntKibibytesBound compression_dictionary_training_max_total_sample_size = new DataStorageSpec.IntKibibytesBound("10MiB");
+    public volatile boolean compression_dictionary_training_auto_train_enabled = false;
+    public volatile float compression_dictionary_training_sampling_rate = 0.01f; // samples 1%
+
     public DataStorageSpec.LongMebibytesBound paxos_cache_size = null;
 
     public DataStorageSpec.LongMebibytesBound consensus_migration_cache_size = null;

src/java/org/apache/cassandra/config/DatabaseDescriptor.java

@@ -4361,6 +4361,47 @@ public class DatabaseDescriptor
         conf.counter_cache_keys_to_save = counterCacheKeysToSave;
     }
 
+    public static int getCompressionDictionaryRefreshIntervalSeconds()
+    {
+        return conf.compression_dictionary_refresh_interval.toSeconds();
+    }
+
+    public static int getCompressionDictionaryRefreshInitialDelaySeconds()
+    {
+        return conf.compression_dictionary_refresh_initial_delay.toSeconds();
+    }
+
+    public static int getCompressionDictionaryCacheSize()
+    {
+        return conf.compression_dictionary_cache_size;
+    }
+
+    public static int getCompressionDictionaryCacheExpireSeconds()
+    {
+        return conf.compression_dictionary_cache_expire.toSeconds();
+    }
+
+    public static int getCompressionDictionaryTrainingMaxDictionarySize()
+    {
+        return conf.compression_dictionary_training_max_dictionary_size.toBytes();
+    }
+
+    public static int getCompressionDictionaryTrainingMaxTotalSampleSize()
+    {
+        return conf.compression_dictionary_training_max_total_sample_size.toBytes();
+    }
+
+    public static boolean getCompressionDictionaryTrainingAutoTrainEnabled()
+    {
+        return conf.compression_dictionary_training_auto_train_enabled;
+    }
+
+
+    public static float getCompressionDictionaryTrainingSamplingRate()
+    {
+        return conf.compression_dictionary_training_sampling_rate;
+    }
+
     public static int getStreamingKeepAlivePeriod()
     {
         return conf.streaming_keep_alive_period.toSeconds();

src/java/org/apache/cassandra/db/ColumnFamilyStore.java

@@ -83,6 +83,7 @@ import org.apache.cassandra.db.compaction.CompactionInfo;
 import org.apache.cassandra.db.compaction.CompactionManager;
 import org.apache.cassandra.db.compaction.CompactionStrategyManager;
 import org.apache.cassandra.db.compaction.OperationType;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.db.filter.ClusteringIndexFilter;
 import org.apache.cassandra.db.filter.DataLimits;
 import org.apache.cassandra.db.lifecycle.ILifecycleTransaction;
@@ -320,6 +321,7 @@ public class ColumnFamilyStore implements ColumnFamilyStoreMBean, Memtable.Owner
     public final TopPartitionTracker topPartitions;
 
     private final SSTableImporter sstableImporter;
+    private final CompressionDictionaryManager compressionDictionaryManager;
 
     private volatile boolean compactionSpaceCheck = true;
 
@@ -390,6 +392,7 @@ public class ColumnFamilyStore implements ColumnFamilyStoreMBean, Memtable.Owner
                 cfs.crcCheckChance = new DefaultValue<>(tableMetadata.params.crcCheckChance);
 
         compactionStrategyManager.maybeReloadParamsFromSchema(tableMetadata.params.compaction);
+        compressionDictionaryManager.maybeReloadFromSchema(tableMetadata.params.compression);
 
         indexManager.reload(tableMetadata);
 
@@ -576,6 +579,7 @@ public class ColumnFamilyStore implements ColumnFamilyStoreMBean, Memtable.Owner
         streamManager = new CassandraStreamManager(this);
         repairManager = new CassandraTableRepairManager(this);
         sstableImporter = new SSTableImporter(this);
+        compressionDictionaryManager = new CompressionDictionaryManager(this, registerBookeeping);
 
         if (DatabaseDescriptor.isClientOrToolInitialized() || SchemaConstants.isSystemKeyspace(getKeyspaceName()))
             topPartitions = null;
@@ -733,6 +737,8 @@ public class ColumnFamilyStore implements ColumnFamilyStoreMBean, Memtable.Owner
         invalidateCaches();
         if (topPartitions != null)
             topPartitions.close();
+
+        compressionDictionaryManager.close();
     }
 
     /**
@@ -3420,6 +3426,12 @@ public class ColumnFamilyStore implements ColumnFamilyStoreMBean, Memtable.Owner
         return metric;
     }
 
+    @Override
+    public CompressionDictionaryManager compressionDictionaryManager()
+    {
+        return compressionDictionaryManager;
+    }
+
     public TableId getTableId()
     {
         return metadata().id;

src/java/org/apache/cassandra/db/compaction/CompactionManager.java

@@ -1796,6 +1796,7 @@ public class CompactionManager implements CompactionManagerMBean, ICompactionMan
                          .setSerializationHeader(sstable.header)
                          .addDefaultComponents(cfs.indexManager.listIndexGroups())
                          .setSecondaryIndexGroups(cfs.indexManager.listIndexGroups())
+                         .setCompressionDictionaryManager(cfs.compressionDictionaryManager())
                          .build(txn, cfs);
     }
 
@@ -1836,6 +1837,7 @@ public class CompactionManager implements CompactionManagerMBean, ICompactionMan
                          .setSerializationHeader(SerializationHeader.make(cfs.metadata(), sstables))
                          .addDefaultComponents(cfs.indexManager.listIndexGroups())
                          .setSecondaryIndexGroups(cfs.indexManager.listIndexGroups())
+                         .setCompressionDictionaryManager(cfs.compressionDictionaryManager())
                          .build(txn, cfs);
     }
 

src/java/org/apache/cassandra/db/compaction/Upgrader.java

@@ -85,6 +85,7 @@ public class Upgrader
                          .setSerializationHeader(SerializationHeader.make(cfs.metadata(), Sets.newHashSet(sstable)))
                          .addDefaultComponents(cfs.indexManager.listIndexGroups())
                          .setSecondaryIndexGroups(cfs.indexManager.listIndexGroups())
+                         .setCompressionDictionaryManager(cfs.compressionDictionaryManager())
                          .build(transaction, cfs);
     }
 

src/java/org/apache/cassandra/db/compaction/unified/ShardedMultiWriter.java

@@ -118,6 +118,7 @@ public class ShardedMultiWriter implements SSTableMultiWriter
                          .setSerializationHeader(header)
                          .addDefaultComponents(indexGroups)
                          .setSecondaryIndexGroups(indexGroups)
+                         .setCompressionDictionaryManager(cfs.compressionDictionaryManager())
                          .build(txn, cfs);
     }
 

src/java/org/apache/cassandra/db/compaction/writers/CompactionAwareWriter.java

@@ -329,6 +329,7 @@ public abstract class CompactionAwareWriter extends Transactional.AbstractTransa
                          .setRepairedAt(minRepairedAt)
                          .setPendingRepair(pendingRepair)
                          .setSecondaryIndexGroups(cfs.indexManager.listIndexGroups())
-                         .addDefaultComponents(cfs.indexManager.listIndexGroups());
+                         .addDefaultComponents(cfs.indexManager.listIndexGroups())
+                         .setCompressionDictionaryManager(cfs.compressionDictionaryManager());
     }
 }

src/java/org/apache/cassandra/db/compression/CompressionDictionary.java

@@ -0,0 +1,284 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.io.DataInput;
+import java.io.DataOutput;
+import java.io.EOFException;
+import java.io.IOException;
+import java.util.Objects;
+import javax.annotation.Nullable;
+
+import com.google.common.base.Preconditions;
+import com.google.common.hash.Hasher;
+import com.google.common.hash.Hashing;
+
+import org.apache.cassandra.cql3.UntypedResultSet;
+import org.apache.cassandra.io.compress.ICompressor;
+import org.apache.cassandra.io.compress.ZstdDictionaryCompressor;
+
+public interface CompressionDictionary extends AutoCloseable
+{
+    /**
+     * Get the dictionary id
+     *
+     * @return dictionary id
+     */
+    DictId dictId();
+
+    /**
+     * Get the raw bytes of the compression dictionary
+     *
+     * @return raw compression dictionary
+     */
+    byte[] rawDictionary();
+
+    /**
+     * Get the kind of the compression algorithm
+     *
+     * @return compression algorithm kind
+     */
+    default Kind kind()
+    {
+        return dictId().kind;
+    }
+
+    /**
+     * Write compression dictionary to file
+     *
+     * @param out file output stream
+     * @throws IOException on any I/O exception when writing to the file
+     */
+    default void serialize(DataOutput out) throws IOException
+    {
+        DictId dictId = dictId();
+        int ordinal = dictId.kind.ordinal();
+        out.writeByte(ordinal);
+        out.writeLong(dictId.id);
+        byte[] dict = rawDictionary();
+        out.writeInt(dict.length);
+        out.write(dict);
+        int checksum = calculateChecksum((byte) ordinal, dictId.id, dict);
+        out.writeInt(checksum);
+    }
+
+    /**
+     * A factory method to create concrete CompressionDictionary from the file content
+     *
+     * @param input   file input stream
+     * @param manager compression dictionary manager that caches the dictionaries
+     * @return compression dictionary; otherwise, null if there is no dictionary
+     * @throws IOException on any I/O exception when reading from the file
+     */
+    @Nullable
+    static CompressionDictionary deserialize(DataInput input, @Nullable CompressionDictionaryManager manager) throws IOException
+    {
+        int kindOrdinal;
+        try
+        {
+            kindOrdinal = input.readByte();
+        }
+        catch (EOFException eof)
+        {
+            // no dictionary
+            return null;
+        }
+
+        if (kindOrdinal < 0 || kindOrdinal >= Kind.values().length)
+        {
+            throw new IOException("Invalid compression dictionary kind: " + kindOrdinal);
+        }
+        Kind kind = Kind.values()[kindOrdinal];
+        long id = input.readLong();
+        DictId dictId = new DictId(kind, id);
+
+        if (manager != null)
+        {
+            CompressionDictionary dictionary = manager.get(dictId);
+            if (dictionary != null)
+            {
+                return dictionary;
+            }
+        }
+
+        int length = input.readInt();
+        byte[] dict = new byte[length];
+        input.readFully(dict);
+        int checksum = input.readInt();
+        int calculatedChecksum = calculateChecksum((byte) kindOrdinal, id, dict);
+        if (checksum != calculatedChecksum)
+            throw new IOException("Compression dictionary checksum does not match. " +
+                                  "Expected: " + checksum + "; actual: " + calculatedChecksum);
+
+        CompressionDictionary dictionary = kind.createDictionary(dictId, dict);
+
+        // update the dictionary manager if it exists
+        if (manager != null)
+        {
+            manager.add(dictionary);
+        }
+
+        return dictionary;
+    }
+
+    static CompressionDictionary createFromRow(UntypedResultSet.Row row)
+    {
+        String kindStr = row.getString("kind");
+        long dictId = row.getLong("dict_id");
+        byte[] dict = row.getByteArray("dict");
+        int storedLength = row.getInt("dict_length");
+        int storedChecksum = row.getInt("dict_checksum");
+
+        try
+        {
+            Kind kind = CompressionDictionary.Kind.valueOf(kindStr);
+
+            // Validate length
+            if (dict.length != storedLength)
+            {
+                throw new IllegalStateException(String.format("Dictionary length mismatch for %s dict id %d. Expected: %d, actual: %d",
+                                                               kindStr, dictId, storedLength, dict.length));
+            }
+
+            // Validate checksum
+            int calculatedChecksum = calculateChecksum((byte) kind.ordinal(), dictId, dict);
+            if (calculatedChecksum != storedChecksum)
+            {
+                throw new IllegalStateException(String.format("Dictionary checksum mismatch for %s dict id %d. Expected: %d, actual: %d",
+                                                               kindStr, dictId, storedChecksum, calculatedChecksum));
+            }
+
+            return kind.createDictionary(new DictId(kind, dictId), dict);
+        }
+        catch (IllegalArgumentException ex)
+        {
+            throw new IllegalStateException(kindStr + " compression dictionary is not created for dict id " + dictId);
+        }
+    }
+
+    @SuppressWarnings("UnstableApiUsage")
+    static int calculateChecksum(byte kindOrdinal, long dictId, byte[] dict)
+    {
+        Hasher hasher = Hashing.crc32c().newHasher();
+        hasher.putByte(kindOrdinal);
+        hasher.putLong(dictId);
+        hasher.putBytes(dict);
+        return hasher.hash().asInt();
+    }
+
+    // Defines the compression dictionary kind, as well as serves as a factory for creating components of dictionary compression
+    enum Kind
+    {
+        // Order matters: the enum ordinal is serialized
+        ZSTD
+        {
+            public CompressionDictionary createDictionary(DictId dictId, byte[] dict)
+            {
+                return new ZstdCompressionDictionary(dictId, dict);
+            }
+
+            @Override
+            public ICompressor createCompressor(CompressionDictionary dictionary)
+            {
+                Preconditions.checkArgument(dictionary instanceof ZstdCompressionDictionary,
+                                            "Expected dictionary to be ZstdCompressionDictionary; actual: %s",
+                                            dictionary.getClass().getSimpleName());
+                return ZstdDictionaryCompressor.create((ZstdCompressionDictionary) dictionary);
+            }
+
+            @Override
+            public ICompressionDictionaryTrainer createTrainer(String keyspaceName,
+                                                               String tableName,
+                                                               CompressionDictionaryTrainingConfig config,
+                                                               ICompressor compressor)
+            {
+                Preconditions.checkArgument(compressor instanceof ZstdDictionaryCompressor,
+                                            "Expected compressor to be ZstdDictionaryCompressor; actual: %s",
+                                            compressor.getClass().getSimpleName());
+                return new ZstdDictionaryTrainer(keyspaceName, tableName, config, ((ZstdDictionaryCompressor) compressor).compressionLevel());
+            }
+        };
+
+        /**
+         * Creates a compression dictionary instance for this kind
+         *
+         * @param dictId the dictionary identifier
+         * @param dict the raw dictionary bytes
+         * @return a compression dictionary instance
+         */
+        public abstract CompressionDictionary createDictionary(CompressionDictionary.DictId dictId, byte[] dict);
+
+        /**
+         * Creates a dictionary compressor for this kind
+         *
+         * @param dictionary the compression dictionary to use for compression
+         * @return a dictionary compressor instance
+         */
+        public abstract ICompressor createCompressor(CompressionDictionary dictionary);
+
+        /**
+         * Creates a dictionary trainer for this kind
+         *
+         * @param keyspaceName the keyspace name
+         * @param tableName the table name
+         * @param config the training configuration
+         * @param compressor the compressor to use for training
+         * @return a dictionary trainer instance
+         */
+        public abstract ICompressionDictionaryTrainer createTrainer(String keyspaceName,
+                                                                    String tableName,
+                                                                    CompressionDictionaryTrainingConfig config,
+                                                                    ICompressor compressor);
+    }
+
+    final class DictId
+    {
+        public final Kind kind;
+        public final long id; // A value of negative or 0 means no dictionary
+
+        public DictId(Kind kind, long id)
+        {
+            this.kind = kind;
+            this.id = id;
+        }
+
+        @Override
+        public boolean equals(Object o)
+        {
+            if (!(o instanceof DictId)) return false;
+            DictId dictId = (DictId) o;
+            return id == dictId.id && kind == dictId.kind;
+        }
+
+        @Override
+        public int hashCode()
+        {
+            return Objects.hash(kind, id);
+        }
+
+        @Override
+        public String toString()
+        {
+            return "DictId{" +
+                   "kind=" + kind +
+                   ", id=" + id +
+                   '}';
+        }
+    }
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryCache.java

@@ -0,0 +1,124 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.time.Duration;
+import java.util.concurrent.atomic.AtomicReference;
+import javax.annotation.Nullable;
+
+import com.google.common.annotations.VisibleForTesting;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import com.github.benmanes.caffeine.cache.Cache;
+import com.github.benmanes.caffeine.cache.Caffeine;
+import com.github.benmanes.caffeine.cache.RemovalCause;
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+
+/**
+ * Manages caching and current dictionary state for compression dictionaries.
+ * <p>
+ * This class handles:
+ * - Local caching of compression dictionaries with automatic cleanup
+ * - Managing the current active dictionary for write operations
+ * - Thread-safe access to cached dictionaries
+ */
+public class CompressionDictionaryCache implements ICompressionDictionaryCache
+{
+    private static final Logger logger = LoggerFactory.getLogger(CompressionDictionaryCache.class);
+
+    private final Cache<DictId, CompressionDictionary> cache;
+    private final AtomicReference<DictId> currentDictId = new AtomicReference<>();
+
+    public CompressionDictionaryCache()
+    {
+        this(DatabaseDescriptor.getCompressionDictionaryCacheSize(), DatabaseDescriptor.getCompressionDictionaryCacheExpireSeconds());
+    }
+
+    @VisibleForTesting
+    CompressionDictionaryCache(int maximumSize, int expireAfterSeconds)
+    {
+        this.cache = Caffeine.newBuilder()
+                             .maximumSize(maximumSize)
+                             .expireAfterAccess(Duration.ofSeconds(expireAfterSeconds))
+                             .removalListener((DictId dictId,
+                                               CompressionDictionary dictionary,
+                                               RemovalCause cause) -> {
+                                 // Close dictionary when evicted from cache to free native resources
+                                 // SelfRefCounted ensures dictionary won't be actually closed if still referenced by compressors
+                                 if (dictionary != null)
+                                 {
+                                     try
+                                     {
+                                         dictionary.close();
+                                     }
+                                     catch (Exception e)
+                                     {
+                                         logger.warn("Failed to close compression dictionary {}", dictId, e);
+                                     }
+                                 }
+                             })
+                             .build();
+    }
+
+    @Nullable
+    @Override
+    public CompressionDictionary getCurrent()
+    {
+        DictId dictId = currentDictId.get();
+        return dictId == null ? null : get(dictId);
+    }
+
+    @Nullable
+    @Override
+    public CompressionDictionary get(DictId dictId)
+    {
+        return cache.getIfPresent(dictId);
+    }
+
+    @Override
+    public void add(@Nullable CompressionDictionary compressionDictionary)
+    {
+        if (compressionDictionary == null)
+            return;
+
+        // Only update cache if not already in the cache
+        DictId newDictId = compressionDictionary.dictId();
+        cache.get(newDictId, id -> compressionDictionary);
+
+        // Update current dictionary if we don't have one or the new one has a higher ID (newer)
+        DictId currentId = currentDictId.get();
+        while ((currentId == null || newDictId.id > currentId.id)
+               && !currentDictId.compareAndSet(currentId, newDictId))
+        {
+            currentId = currentDictId.get();
+        }
+    }
+
+    @Override
+    public synchronized void close()
+    {
+        currentDictId.set(null);
+        // Invalidate cache will trigger removalListener to close all cached dictionaries, including the currentDictionary
+        cache.invalidateAll();
+        // Force synchronous cleanup to ensure removal listener executes immediately
+        cache.cleanUp();
+    }
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryEventHandler.java

@@ -0,0 +1,121 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import org.apache.cassandra.concurrent.ScheduledExecutors;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.locator.InetAddressAndPort;
+import org.apache.cassandra.net.Message;
+import org.apache.cassandra.net.MessagingService;
+import org.apache.cassandra.net.Verb;
+import org.apache.cassandra.schema.SystemDistributedKeyspace;
+import org.apache.cassandra.tcm.ClusterMetadata;
+import org.apache.cassandra.utils.FBUtilities;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.util.Collection;
+
+/**
+ * Handles compression dictionary events including training completion and cluster notifications.
+ * <p>
+ * This class handles:
+ * - Broadcasting dictionary updates to cluster nodes
+ * - Retrieving new dictionaries when notified by other nodes
+ * - Managing dictionary cache updates
+ */
+public class CompressionDictionaryEventHandler implements ICompressionDictionaryEventHandler
+{
+    private static final Logger logger = LoggerFactory.getLogger(CompressionDictionaryEventHandler.class);
+
+    private final ColumnFamilyStore cfs;
+    private final String keyspaceName;
+    private final String tableName;
+    private final ICompressionDictionaryCache cache;
+
+    public CompressionDictionaryEventHandler(ColumnFamilyStore cfs, ICompressionDictionaryCache cache)
+    {
+        this.cfs = cfs;
+        this.keyspaceName = cfs.keyspace.getName();
+        this.tableName = cfs.getTableName();
+        this.cache = cache;
+    }
+
+    @Override
+    public void onNewDictionaryTrained(CompressionDictionary.DictId dictionaryId)
+    {
+        logger.info("Notifying cluster about dictionary update for {}.{} with {}",
+                    keyspaceName, tableName, dictionaryId);
+
+        CompressionDictionaryUpdateMessage message = new CompressionDictionaryUpdateMessage(cfs.metadata().id, dictionaryId);
+        Collection<InetAddressAndPort> allNodes = ClusterMetadata.current().directory.allJoinedEndpoints();
+        // Broadcast notification using the fire-and-forget fashion
+        for (InetAddressAndPort node : allNodes)
+        {
+            if (node.equals(FBUtilities.getBroadcastAddressAndPort())) // skip ourself
+                continue;
+            sendNotification(node, message);
+        }
+    }
+
+    @Override
+    public void onNewDictionaryAvailable(CompressionDictionary.DictId dictionaryId)
+    {
+        // Best effort to retrieve the dictionary; otherwise, the periodic task should retrieve the dictionary later
+        ScheduledExecutors.nonPeriodicTasks.submit(() -> {
+            try
+            {
+                if (!cfs.metadata().params.compression.isDictionaryCompressionEnabled())
+                {
+                    return;
+                }
+
+                CompressionDictionary dictionary = SystemDistributedKeyspace.retrieveCompressionDictionary(keyspaceName, tableName, dictionaryId);
+                cache.add(dictionary);
+            }
+            catch (Exception e)
+            {
+                logger.warn("Failed to retrieve compression dictionary for {}.{}. {}",
+                            keyspaceName, tableName, dictionaryId, e);
+            }
+        });
+    }
+
+    // Best effort to notify the peer regarding the new dictionary being available to pull.
+    // If the request fails, each peer has periodic task scheduled to pull.
+    private void sendNotification(InetAddressAndPort target, CompressionDictionaryUpdateMessage message)
+    {
+        logger.debug("Sending dictionary update notification for {} to {}", message.dictionaryId, target);
+
+        Message<CompressionDictionaryUpdateMessage> msg = Message.out(Verb.DICTIONARY_UPDATE_REQ, message);
+        MessagingService.instance()
+                        .sendWithResponse(target, msg)
+                        .addListener(future -> {
+                            if (future.isSuccess())
+                            {
+                                logger.debug("Successfully sent dictionary update notification to {}", target);
+                            }
+                            else
+                            {
+                                logger.warn("Failed to send dictionary update notification to {}",
+                                            target, future.cause());
+                            }
+                        });
+    }
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryManager.java

@@ -0,0 +1,355 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.nio.ByteBuffer;
+import java.util.Set;
+import javax.annotation.Nullable;
+import javax.management.openmbean.CompositeData;
+
+import com.google.common.annotations.VisibleForTesting;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+import org.apache.cassandra.schema.CompressionParams;
+import org.apache.cassandra.schema.SystemDistributedKeyspace;
+import org.apache.cassandra.utils.MBeanWrapper;
+import org.apache.cassandra.utils.MBeanWrapper.OnException;
+
+public class CompressionDictionaryManager implements CompressionDictionaryManagerMBean,
+                                                     ICompressionDictionaryCache,
+                                                     ICompressionDictionaryEventHandler,
+                                                     AutoCloseable
+{
+    private static final Logger logger = LoggerFactory.getLogger(CompressionDictionaryManager.class);
+
+    private final String keyspaceName;
+    private final String tableName;
+    private final ColumnFamilyStore columnFamilyStore;
+    private volatile boolean mbeanRegistered;
+    private volatile boolean isEnabled;
+
+    // Components
+    private final ICompressionDictionaryEventHandler eventHandler;
+    private final ICompressionDictionaryCache cache;
+    private final ICompressionDictionaryScheduler scheduler;
+    private ICompressionDictionaryTrainer trainer = null;
+
+    public CompressionDictionaryManager(ColumnFamilyStore columnFamilyStore, boolean registerBookkeeping)
+    {
+        this.keyspaceName = columnFamilyStore.keyspace.getName();
+        this.tableName = columnFamilyStore.getTableName();
+        this.columnFamilyStore = columnFamilyStore;
+
+        this.isEnabled = columnFamilyStore.metadata().params.compression.isDictionaryCompressionEnabled();
+        this.cache = new CompressionDictionaryCache();
+        this.eventHandler = new CompressionDictionaryEventHandler(columnFamilyStore, cache);
+        this.scheduler = new CompressionDictionaryScheduler(keyspaceName, tableName, cache, isEnabled);
+        if (isEnabled)
+        {
+            // Initialize components
+            this.trainer = ICompressionDictionaryTrainer.create(keyspaceName, tableName,
+                                                                columnFamilyStore.metadata().params.compression,
+                                                                createTrainingConfig());
+            trainer.setDictionaryTrainedListener(this::handleNewDictionary);
+
+            scheduler.scheduleRefreshTask();
+
+            trainer.start(false);
+        }
+
+        if (registerBookkeeping && isEnabled)
+        {
+            registerMbean();
+        }
+    }
+
+    static String mbeanName(String keyspaceName, String tableName)
+    {
+        return MBEAN_NAME + ",keyspace=" + keyspaceName + ",table=" + tableName;
+    }
+
+    public boolean isEnabled()
+    {
+        return isEnabled;
+    }
+
+    /**
+     * Reloads dictionary management configuration when compression parameters change.
+     * This method enables or disables dictionary compression based on the new parameters,
+     * and properly manages the lifecycle of training and refresh tasks.
+     *
+     * @param newParams the new compression parameters to apply
+     */
+    public synchronized void maybeReloadFromSchema(CompressionParams newParams)
+    {
+        this.isEnabled = newParams.isDictionaryCompressionEnabled();
+        scheduler.setEnabled(isEnabled);
+        if (isEnabled)
+        {
+            registerMbean();
+            // Check if we need a new trainer due to compression parameter changes
+            boolean needsNewTrainer = shouldCreateNewTrainer(newParams);
+
+            if (needsNewTrainer)
+            {
+                // Close existing trainer and create a new one
+                if (trainer != null)
+                {
+                    try
+                    {
+                        trainer.close();
+                    }
+                    catch (Exception e)
+                    {
+                        logger.warn("Failed to close existing trainer for {}.{}", keyspaceName, tableName, e);
+                    }
+                }
+
+                trainer = ICompressionDictionaryTrainer.create(keyspaceName, tableName, newParams, createTrainingConfig());
+                trainer.setDictionaryTrainedListener(this::handleNewDictionary);
+            }
+
+            scheduler.scheduleRefreshTask();
+
+            // Start trainer if it exists
+            if (trainer != null)
+            {
+                trainer.start(false);
+            }
+            return;
+        }
+
+        // Clean up when dictionary compression is disabled
+        try
+        {
+            close();
+        }
+        catch (Exception e)
+        {
+            logger.warn("Failed to close CompressionDictionaryManager on disabling " +
+                        "dictionary-based compression for table {}.{}", keyspaceName, tableName);
+        }
+    }
+
+    /**
+     * Adds a sample to the dictionary trainer for learning compression patterns.
+     * Samples are randomly selected to avoid bias and improve dictionary quality.
+     *
+     * @param sample the sample data to potentially add for training
+     */
+    public void addSample(ByteBuffer sample)
+    {
+        ICompressionDictionaryTrainer dictionaryTrainer = trainer;
+        if (dictionaryTrainer != null && dictionaryTrainer.shouldSample())
+        {
+            dictionaryTrainer.addSample(sample);
+        }
+    }
+    
+    @Nullable
+    @Override
+    public CompressionDictionary getCurrent()
+    {
+        return cache.getCurrent();
+    }
+
+    @Override
+    public CompressionDictionary get(CompressionDictionary.DictId dictId)
+    {
+        return cache.get(dictId);
+    }
+
+    @Override
+    public void add(@Nullable CompressionDictionary compressionDictionary)
+    {
+        cache.add(compressionDictionary);
+    }
+
+    @Override
+    public void onNewDictionaryTrained(CompressionDictionary.DictId dictionaryId)
+    {
+        eventHandler.onNewDictionaryTrained(dictionaryId);
+    }
+
+    @Override
+    public void onNewDictionaryAvailable(CompressionDictionary.DictId dictionaryId)
+    {
+        eventHandler.onNewDictionaryAvailable(dictionaryId);
+    }
+
+    @Override
+    public synchronized void train(boolean force)
+    {
+        // Validate table supports dictionary compression
+        if (!isEnabled)
+        {
+            throw new UnsupportedOperationException("Table " + keyspaceName + '.' + tableName + " does not support dictionary compression");
+        }
+
+        if (trainer == null)
+        {
+            throw new IllegalStateException("Dictionary trainer is not available for table " + keyspaceName + '.' + tableName);
+        }
+
+        // SSTable-based training: sample from existing SSTables
+        Set<SSTableReader> sstables = columnFamilyStore.getLiveSSTables();
+        if (sstables.isEmpty())
+        {
+            logger.info("No SSTables available for training in table {}.{}, flushing memtable first",
+                       keyspaceName, tableName);
+            columnFamilyStore.forceBlockingFlush(ColumnFamilyStore.FlushReason.USER_FORCED);
+            sstables = columnFamilyStore.getLiveSSTables();
+
+            if (sstables.isEmpty())
+            {
+                throw new IllegalStateException("No SSTables available for training in table " + keyspaceName + '.' + tableName + " after flush");
+            }
+        }
+
+        logger.info("Starting SSTable-based training for {}.{} with {} SSTables",
+                   keyspaceName, tableName, sstables.size());
+
+        trainer.start(true);
+        scheduler.scheduleSSTableBasedTraining(trainer, sstables, createTrainingConfig(), force);
+    }
+
+    @Override
+    public CompositeData getTrainingState()
+    {
+        ICompressionDictionaryTrainer dictionaryTrainer = trainer;
+        if (dictionaryTrainer == null)
+        {
+            return TrainingState.notStarted().toCompositeData();
+        }
+        return dictionaryTrainer.getTrainingState().toCompositeData();
+    }
+
+    /**
+     * Close all the resources. The method can be called multiple times.
+     */
+    @Override
+    public synchronized void close()
+    {
+        unregisterMbean();
+        if (trainer != null)
+        {
+            closeQuitely(trainer, "CompressionDictionaryTrainer");
+            trainer = null;
+        }
+        closeQuitely(cache, "CompressionDictionaryCache");
+        closeQuitely(scheduler, "CompressionDictionaryScheduler");
+    }
+
+    private void handleNewDictionary(CompressionDictionary dictionary)
+    {
+        // sequence meatters; persist the new dictionary before broadcasting to others.
+        storeDictionary(dictionary);
+        onNewDictionaryTrained(dictionary.dictId());
+    }
+
+    private CompressionDictionaryTrainingConfig createTrainingConfig()
+    {
+        CompressionParams compressionParams = columnFamilyStore.metadata().params.compression;
+        return CompressionDictionaryTrainingConfig
+               .builder()
+               .maxDictionarySize(DatabaseDescriptor.getCompressionDictionaryTrainingMaxDictionarySize())
+               .maxTotalSampleSize(DatabaseDescriptor.getCompressionDictionaryTrainingMaxTotalSampleSize())
+               .samplingRate(DatabaseDescriptor.getCompressionDictionaryTrainingSamplingRate())
+               .chunkSize(compressionParams.chunkLength())
+               .build();
+    }
+
+    private void storeDictionary(CompressionDictionary dictionary)
+    {
+        if (!isEnabled)
+        {
+            return;
+        }
+
+        SystemDistributedKeyspace.storeCompressionDictionary(keyspaceName, tableName, dictionary);
+        cache.add(dictionary);
+    }
+
+    /**
+     * Determines if a new trainer should be created based on compression parameter changes.
+     * A new trainer is needed when no existing trainer exists or when the existing trainer
+     * is not compatible with the new compression parameters.
+     *
+     * The method is (and should be) only invoked inside {@link #maybeReloadFromSchema(CompressionParams)},
+     * which is guarded by synchronized.
+     *
+     * @param newParams the new compression parameters
+     * @return true if a new trainer should be created
+     */
+    private boolean shouldCreateNewTrainer(CompressionParams newParams)
+    {
+        if (trainer == null)
+        {
+            return true;
+        }
+
+        return !trainer.isCompatibleWith(newParams);
+    }
+
+    private void registerMbean()
+    {
+        if (!mbeanRegistered)
+        {
+            MBeanWrapper.instance.registerMBean(this, mbeanName(keyspaceName, tableName));
+            mbeanRegistered = true;
+        }
+    }
+
+    private void unregisterMbean()
+    {
+        if (mbeanRegistered)
+        {
+            MBeanWrapper.instance.unregisterMBean(mbeanName(keyspaceName, tableName), OnException.IGNORE);
+            mbeanRegistered = false;
+        }
+    }
+
+    private void closeQuitely(AutoCloseable closeable, String objectName)
+    {
+        try
+        {
+            closeable.close();
+        }
+        catch (Exception exception)
+        {
+            logger.warn("Failed closing {}", objectName, exception);
+        }
+    }
+
+    @VisibleForTesting
+    boolean isReady()
+    {
+        return trainer != null && trainer.isReady();
+    }
+
+    @VisibleForTesting
+    ICompressionDictionaryTrainer trainer()
+    {
+        return trainer;
+    }
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryManagerMBean.java

@@ -0,0 +1,47 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import javax.management.openmbean.CompositeData;
+
+public interface CompressionDictionaryManagerMBean
+{
+    String MBEAN_NAME = "org.apache.cassandra.db.compression:type=CompressionDictionaryManager";
+
+    /**
+     * Starts training from existing SSTables for this table.
+     * Samples chunks from all live SSTables and trains a compression dictionary.
+     * If no SSTables are available, automatically flushes the memtable first.
+     * This operation runs synchronously and blocks until training completes.
+     *
+     * @param force force the dictionary training even if there are not enough samples;
+     *              otherwise, dictionary training won't start if the trainer is not ready
+     * @throws UnsupportedOperationException if table doesn't support dictionary compression
+     * @throws IllegalStateException if no SSTables available after flush
+     */
+    void train(boolean force);
+
+    /**
+     * Gets the current training state for this table.
+     * Returns a snapshot of {@link TrainingState} as JMX CompositeData.
+     *
+     * @return CompositeData representing {@link TrainingState}
+     */
+    CompositeData getTrainingState();
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryScheduler.java

@@ -0,0 +1,260 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.ArrayList;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+import java.util.concurrent.ScheduledFuture;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicBoolean;
+
+import com.google.common.annotations.VisibleForTesting;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.cassandra.concurrent.ScheduledExecutors;
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+import org.apache.cassandra.schema.SystemDistributedKeyspace;
+import org.apache.cassandra.utils.concurrent.Ref;
+
+/**
+ * Manages scheduled tasks for compression dictionary operations.
+ * <p>
+ * This class handles:
+ * - Periodic refresh of dictionaries from system tables
+ * - Manual training task scheduling and monitoring
+ * - Cleanup of scheduled tasks
+ */
+public class CompressionDictionaryScheduler implements ICompressionDictionaryScheduler
+{
+    private static final Logger logger = LoggerFactory.getLogger(CompressionDictionaryScheduler.class);
+
+    private final String keyspaceName;
+    private final String tableName;
+    private final ICompressionDictionaryCache cache;
+    private final AtomicBoolean manualTrainingInProgress = new AtomicBoolean(false);
+
+    private volatile ScheduledFuture<?> scheduledRefreshTask;
+    private volatile boolean isEnabled;
+
+    public CompressionDictionaryScheduler(String keyspaceName,
+                                          String tableName,
+                                          ICompressionDictionaryCache cache,
+                                          boolean isEnabled)
+    {
+        this.keyspaceName = keyspaceName;
+        this.tableName = tableName;
+        this.cache = cache;
+        this.isEnabled = isEnabled;
+    }
+
+    /**
+     * Schedules the periodic dictionary refresh task if not already scheduled.
+     */
+    public void scheduleRefreshTask()
+    {
+        if (scheduledRefreshTask != null)
+            return;
+
+        this.scheduledRefreshTask = ScheduledExecutors.scheduledTasks.scheduleWithFixedDelay(
+        this::refreshDictionaryFromSystemTable,
+        DatabaseDescriptor.getCompressionDictionaryRefreshInitialDelaySeconds(),
+        DatabaseDescriptor.getCompressionDictionaryRefreshIntervalSeconds(),
+        TimeUnit.SECONDS
+        );
+    }
+
+    @Override
+    public void scheduleSSTableBasedTraining(ICompressionDictionaryTrainer trainer,
+                                             Set<SSTableReader> sstables,
+                                             CompressionDictionaryTrainingConfig config,
+                                             boolean force)
+    {
+        if (!manualTrainingInProgress.compareAndSet(false, true))
+        {
+            throw new IllegalStateException("Training already in progress for table " + keyspaceName + '.' + tableName);
+        }
+
+        logger.info("Starting SSTable-based dictionary training for {}.{} from {} SSTables",
+                    keyspaceName, tableName, sstables.size());
+
+        // Run the SSTableSamplingTask asynchronously
+        SSTableSamplingTask task = new SSTableSamplingTask(sstables, trainer, config, force);
+        ScheduledExecutors.nonPeriodicTasks.submit(task);
+    }
+
+    /**
+     * Cancels the in-progress manual training task.
+     */
+    private void cancelManualTraining()
+    {
+        manualTrainingInProgress.compareAndSet(true, false);
+    }
+
+    /**
+     * Sets the enabled state of the scheduler. When disabled, refresh tasks will not execute.
+     *
+     * @param enabled whether the scheduler should be enabled
+     */
+    @Override
+    public void setEnabled(boolean enabled)
+    {
+        this.isEnabled = enabled;
+    }
+
+    /**
+     * Refreshes dictionary from system table and updates the cache.
+     * This method is called periodically by the scheduled refresh task.
+     */
+    private void refreshDictionaryFromSystemTable()
+    {
+        try
+        {
+            if (!isEnabled)
+            {
+                return;
+            }
+
+            CompressionDictionary dictionary = SystemDistributedKeyspace.retrieveLatestCompressionDictionary(keyspaceName, tableName);
+            cache.add(dictionary);
+        }
+        catch (Exception e)
+        {
+            logger.warn("Failed to refresh compression dictionary for {}.{}",
+                        keyspaceName, tableName, e);
+        }
+    }
+
+    @Override
+    public void close()
+    {
+        if (scheduledRefreshTask != null)
+        {
+            scheduledRefreshTask.cancel(false);
+            scheduledRefreshTask = null;
+        }
+
+        cancelManualTraining();
+    }
+
+    /**
+     * Task that samples chunks from existing SSTables and triggers training.
+     * Acquires references to SSTables to prevent them from being deleted during sampling.
+     */
+    private class SSTableSamplingTask implements Runnable
+    {
+        private final Set<SSTableReader> sstables;
+        private final ICompressionDictionaryTrainer trainer;
+        private final CompressionDictionaryTrainingConfig config;
+        private final List<Ref<SSTableReader>> sstableRefs;
+        private final boolean force;
+
+        private SSTableSamplingTask(Set<SSTableReader> sstables,
+                                    ICompressionDictionaryTrainer trainer,
+                                    CompressionDictionaryTrainingConfig config,
+                                    boolean force)
+        {
+            this.trainer = trainer;
+            this.config = config;
+            this.force = force;
+
+            // Acquire references to all SSTables to prevent deletion during sampling
+            this.sstableRefs = new ArrayList<>();
+            Set<SSTableReader> referencedSSTables = new HashSet<>();
+
+            for (SSTableReader sstable : sstables)
+            {
+                Ref<SSTableReader> ref = sstable.tryRef();
+                if (ref != null)
+                {
+                    sstableRefs.add(ref);
+                    referencedSSTables.add(sstable);
+                }
+                else
+                {
+                    logger.debug("Couldn't acquire reference to SSTable {}. It may have been removed.",
+                                 sstable.descriptor);
+                }
+            }
+
+            this.sstables = referencedSSTables;
+        }
+
+        @Override
+        public void run()
+        {
+            try
+            {
+                if (sstables.isEmpty())
+                {
+                    logger.warn("No SSTables available for sampling in {}.{}", keyspaceName, tableName);
+                    cancelManualTraining();
+                    return;
+                }
+
+                logger.info("Sampling chunks from {} SSTables for {}.{}",
+                            sstables.size(), keyspaceName, tableName);
+
+                // Sample chunks from SSTables and add to trainer
+                SSTableChunkSampler.sampleFromSSTables(sstables, trainer, config);
+
+                logger.info("Completed sampling for {}.{}, now training dictionary",
+                            keyspaceName, tableName);
+
+                // Use the force parameter from the task
+                trainer.trainDictionaryAsync(force)
+                       .addCallback((dictionary, throwable) -> {
+                           cancelManualTraining();
+                           if (throwable != null)
+                           {
+                               logger.error("SSTable-based dictionary training failed for {}.{}: {}",
+                                            keyspaceName, tableName, throwable.getMessage());
+                           }
+                           else
+                           {
+                               logger.info("SSTable-based dictionary training completed for {}.{}",
+                                           keyspaceName, tableName);
+                           }
+                       });
+            }
+            catch (Exception e)
+            {
+                logger.error("Failed to sample from SSTables for {}.{}", keyspaceName, tableName, e);
+                cancelManualTraining();
+            }
+            finally
+            {
+                // Release all SSTable references
+                for (Ref<SSTableReader> ref : sstableRefs)
+                {
+                    ref.release();
+                }
+            }
+        }
+    }
+
+    @VisibleForTesting
+    boolean isManualTrainingRunning()
+    {
+        return manualTrainingInProgress.get();
+    }
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryTrainingConfig.java

@@ -0,0 +1,88 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import com.google.common.base.Preconditions;
+
+/**
+ * Configuration for dictionary training parameters.
+ */
+public class CompressionDictionaryTrainingConfig
+{
+    public final int maxDictionarySize;
+    public final int maxTotalSampleSize;
+    public final int acceptableTotalSampleSize;
+    public final int samplingRate;
+    public final int chunkSize;
+
+    private CompressionDictionaryTrainingConfig(Builder builder)
+    {
+        this.maxDictionarySize = builder.maxDictionarySize;
+        this.maxTotalSampleSize = builder.maxTotalSampleSize;
+        this.acceptableTotalSampleSize = builder.maxTotalSampleSize / 10 * 8;
+        this.samplingRate = builder.samplingRate;
+        this.chunkSize = builder.chunkSize;
+    }
+
+    public static Builder builder()
+    {
+        return new Builder();
+    }
+
+    public static class Builder
+    {
+        private int maxDictionarySize = 65536; // 64KB default
+        private int maxTotalSampleSize = 10 * 1024 * 1024; // 10MB total
+        private int samplingRate = 100; // Sampling 1%
+        private int chunkSize = 64 * 1024; // 64KB default
+
+        public Builder maxDictionarySize(int size)
+        {
+            this.maxDictionarySize = size;
+            return this;
+        }
+
+        public Builder maxTotalSampleSize(int size)
+        {
+            this.maxTotalSampleSize = size;
+            return this;
+        }
+
+        public Builder samplingRate(float samplingRate)
+        {
+            this.samplingRate = Math.round(1 / samplingRate);
+            return this;
+        }
+
+        public Builder chunkSize(int chunkSize)
+        {
+            this.chunkSize = chunkSize;
+            return this;
+        }
+
+        public CompressionDictionaryTrainingConfig build()
+        {
+            Preconditions.checkArgument(maxDictionarySize > 0, "maxDictionarySize must be positive");
+            Preconditions.checkArgument(maxTotalSampleSize > 0, "maxTotalSampleSize must be positive");
+            Preconditions.checkArgument(samplingRate > 0, "samplingRate must be positive");
+            Preconditions.checkArgument(chunkSize > 0, "chunkSize must be positive");
+            return new CompressionDictionaryTrainingConfig(this);
+        }
+    }
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryUpdateMessage.java

@@ -0,0 +1,70 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.io.IOException;
+
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import org.apache.cassandra.io.IVersionedSerializer;
+import org.apache.cassandra.io.util.DataInputPlus;
+import org.apache.cassandra.io.util.DataOutputPlus;
+import org.apache.cassandra.schema.TableId;
+
+public class CompressionDictionaryUpdateMessage
+{
+    public static final IVersionedSerializer<CompressionDictionaryUpdateMessage> serializer = new DictionaryUpdateMessageSerializer();
+
+    public final TableId tableId;
+    public final DictId dictionaryId;
+
+    public CompressionDictionaryUpdateMessage(TableId tableId, DictId dictionaryId)
+    {
+        this.tableId = tableId;
+        this.dictionaryId = dictionaryId;
+    }
+
+    public static class DictionaryUpdateMessageSerializer implements IVersionedSerializer<CompressionDictionaryUpdateMessage>
+    {
+        @Override
+        public void serialize(CompressionDictionaryUpdateMessage message, DataOutputPlus out, int version) throws IOException
+        {
+            TableId.serializer.serialize(message.tableId, out, version);
+            out.writeByte(message.dictionaryId.kind.ordinal());
+            out.writeLong(message.dictionaryId.id);
+        }
+
+        @Override
+        public CompressionDictionaryUpdateMessage deserialize(DataInputPlus in, int version) throws IOException
+        {
+            TableId tableId = TableId.serializer.deserialize(in, version);
+            int kindOrdinal = in.readByte();
+            long dictionaryId = in.readLong();
+            DictId dictId = new DictId(CompressionDictionary.Kind.values()[kindOrdinal], dictionaryId);
+            return new CompressionDictionaryUpdateMessage(tableId, dictId);
+        }
+
+        @Override
+        public long serializedSize(CompressionDictionaryUpdateMessage message, int version)
+        {
+            return TableId.serializer.serializedSize(message.tableId, version) +
+                   1 + // byte for kind ordinal
+                   8; // long for dictionaryId
+        }
+    }
+}

src/java/org/apache/cassandra/db/compression/CompressionDictionaryUpdateVerbHandler.java

@@ -0,0 +1,61 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.net.IVerbHandler;
+import org.apache.cassandra.net.Message;
+import org.apache.cassandra.schema.Schema;
+
+public class CompressionDictionaryUpdateVerbHandler implements IVerbHandler<CompressionDictionaryUpdateMessage>
+{
+    private static final Logger logger = LoggerFactory.getLogger(CompressionDictionaryUpdateVerbHandler.class);
+    public static final CompressionDictionaryUpdateVerbHandler instance = new CompressionDictionaryUpdateVerbHandler();
+
+    private CompressionDictionaryUpdateVerbHandler() {}
+
+    @Override
+    public void doVerb(Message<CompressionDictionaryUpdateMessage> message)
+    {
+        CompressionDictionaryUpdateMessage payload = message.payload;
+
+        try
+        {
+            ColumnFamilyStore cfs = Schema.instance.getColumnFamilyStoreInstance(payload.tableId);
+            if (cfs == null)
+            {
+                logger.warn("Received dictionary update for unknown table with tableId {}", payload.tableId);
+                return;
+            }
+
+            logger.debug("Received dictionary update notification for {}.{} with dictionaryId {}",
+                         cfs.keyspace, cfs.name, payload.dictionaryId);
+            CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+            manager.onNewDictionaryAvailable(payload.dictionaryId);
+        }
+        catch (Exception e)
+        {
+            logger.error("Failed to process dictionary update notification for tableId {}",
+                         payload.tableId, e);
+        }
+    }
+}

src/java/org/apache/cassandra/db/compression/ICompressionDictionaryCache.java

@@ -0,0 +1,56 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import javax.annotation.Nullable;
+
+/**
+ * Interface for managing compression dictionary caching and current dictionary state.
+ * <p>
+ * Implementations handle:
+ * - Local caching of compression dictionaries with automatic cleanup
+ * - Managing the current active dictionary for write operations
+ * - Thread-safe access to cached dictionaries
+ */
+public interface ICompressionDictionaryCache extends AutoCloseable
+{
+    /**
+     * Gets the current active compression dictionary.
+     *
+     * @return the current compression dictionary, or null if no dictionary is available
+     */
+    @Nullable
+    CompressionDictionary getCurrent();
+
+    /**
+     * Retrieves a specific compression dictionary by its identifier.
+     *
+     * @param dictId the dictionary identifier to look up
+     * @return the compression dictionary with the given identifier, or null if not found in cache
+     */
+    @Nullable
+    CompressionDictionary get(CompressionDictionary.DictId dictId);
+
+    /**
+     * Stores a compression dictionary in the local cache and updates the current dictionary if the new one is newer.
+     *
+     * @param compressionDictionary the compression dictionary to cache, may be null
+     */
+    void add(@Nullable CompressionDictionary compressionDictionary);
+}

src/java/org/apache/cassandra/db/compression/ICompressionDictionaryEventHandler.java

@@ -0,0 +1,35 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+public interface ICompressionDictionaryEventHandler
+{
+    /**
+     * Invoked when a new dictionary is trained
+     * @param dictionaryId dictionary id
+     */
+    void onNewDictionaryTrained(CompressionDictionary.DictId dictionaryId);
+
+    /**
+     * Invoked when {@link CompressionDictionaryUpdateMessage} is received indicating
+     * a dictionary is trained and local node should retrieve the specified dictionary
+     * @param dictionaryId dictionary id
+     */
+    void onNewDictionaryAvailable(CompressionDictionary.DictId dictionaryId);
+}

src/java/org/apache/cassandra/db/compression/ICompressionDictionaryScheduler.java

@@ -0,0 +1,60 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.Set;
+
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+
+/**
+ * Interface for managing scheduled tasks for compression dictionary operations.
+ * <p>
+ * Implementations handle:
+ * - Periodic refresh of dictionaries from system tables
+ * - Manual training task scheduling and monitoring
+ * - Cleanup of scheduled tasks
+ */
+public interface ICompressionDictionaryScheduler extends AutoCloseable
+{
+    /**
+     * Schedules the periodic dictionary refresh task if not already scheduled.
+     */
+    void scheduleRefreshTask();
+
+    /**
+     * Schedules SSTable-based training that samples from existing SSTables.
+     *
+     * @param trainer the trainer to use
+     * @param sstables the set of SSTables to sample from
+     * @param config the training configuration
+     * @param force force the dictionary training even if there are not enough samples
+     * @throws IllegalStateException if training is already in progress
+     */
+    void scheduleSSTableBasedTraining(ICompressionDictionaryTrainer trainer,
+                                      Set<SSTableReader> sstables,
+                                      CompressionDictionaryTrainingConfig config,
+                                      boolean force);
+
+    /**
+     * Sets the enabled state of the scheduler. When disabled, refresh tasks will not execute.
+     *
+     * @param enabled whether the scheduler should be enabled
+     */
+    void setEnabled(boolean enabled);
+}

src/java/org/apache/cassandra/db/compression/ICompressionDictionaryTrainer.java

@@ -0,0 +1,163 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.nio.ByteBuffer;
+import java.util.function.Consumer;
+
+import org.apache.cassandra.concurrent.ScheduledExecutors;
+import org.apache.cassandra.utils.concurrent.Future;
+import org.apache.cassandra.io.compress.ICompressor;
+import org.apache.cassandra.io.compress.IDictionaryCompressor;
+import org.apache.cassandra.schema.CompressionParams;
+
+/**
+ * Interface for training compression dictionaries from sample data.
+ * <p>
+ * Implementations handle:
+ * - Sample collection and management
+ * - Dictionary training lifecycle
+ * - Asynchronous training execution
+ * - Training status tracking
+ */
+public interface ICompressionDictionaryTrainer extends AutoCloseable
+{
+    /**
+     * Starts the trainer for collecting samples.
+     *
+     * @param manualTraining true if this is manual training, false for automatic
+     * @return true if the trainer is started; otherwise false. The trainer is started
+     *         in any of those conditions: 1. trainer closed; 2. not requested for
+     *         either manual or auto training; 3. failed to start
+     */
+    boolean start(boolean manualTraining);
+
+    /**
+     * @return true if the trainer is ready to take a new sample; otherwise, false
+     */
+    boolean shouldSample();
+
+    /**
+     * Adds a sample to the training dataset.
+     *
+     * @param sample the sample data to add for training
+     */
+    void addSample(ByteBuffer sample);
+
+    /**
+     * Trains and produces a compression dictionary from collected samples synchronously.
+     *
+     * @param force force the dictionary training even if there are not enough samples;
+     *              otherwise, dictionary training won't start if the trainer is not ready
+     * @return the trained compression dictionary
+     */
+    CompressionDictionary trainDictionary(boolean force);
+
+    /**
+     * Trains and produces a compression dictionary from collected samples asynchronously.
+     *
+     * @param force force the dictionary training even if there are not enough samples
+     * @return Future that completes when training is done
+     */
+    default Future<CompressionDictionary> trainDictionaryAsync(boolean force)
+    {
+        return ScheduledExecutors.nonPeriodicTasks.submit(() -> trainDictionary(force));
+    }
+
+    /**
+     * @return true if enough samples have been collected for training
+     */
+    boolean isReady();
+
+    /**
+     * Clears all collected samples and resets trainer state.
+     */
+    void reset();
+
+    /**
+     * Gets the current training state including status, progress, and failure details.
+     *
+     * @return the current training state as an atomic snapshot
+     */
+    TrainingState getTrainingState();
+
+    /**
+     * @return the compression algorithm kind this trainer supports
+     */
+    CompressionDictionary.Kind kind();
+
+    /**
+     * Determines if this trainer is compatible with the given compression parameters.
+     * This method allows the trainer to decide whether it can continue operating
+     * with new compression parameters or if a new trainer instance is needed.
+     *
+     * @param newParams the new compression parameters to check compatibility against
+     * @return true if this trainer is compatible with the new parameters, false otherwise
+     */
+    boolean isCompatibleWith(CompressionParams newParams);
+
+    /**
+     * Sets the listener for dictionary training events.
+     *
+     * @param listener the listener to be notified when dictionaries are trained, null to remove listener
+     */
+    void setDictionaryTrainedListener(Consumer<CompressionDictionary> listener);
+
+    /**
+     * Updates the sampling rate for this trainer.
+     *
+     * @param newSamplingRate the new sampling rate. For exmaple, 1 = sample every time (100%),
+     *                        2 = expect sample 1/2 of data (50%), n = expect sample 1/n of data
+     */
+    void updateSamplingRate(int newSamplingRate);
+
+    /**
+     * Factory method to create appropriate trainer based on compression parameters.
+     *
+     * @param keyspaceName the keyspace name for logging
+     * @param tableName the table name for logging
+     * @param params the compression parameters
+     * @param config the training configuration
+     * @return a dictionary trainer for the specified compression algorithm
+     * @throws IllegalArgumentException if no dictionary trainer is available for the compression algorithm
+     */
+    static ICompressionDictionaryTrainer create(String keyspaceName,
+                                                String tableName,
+                                                CompressionParams params,
+                                                CompressionDictionaryTrainingConfig config)
+    {
+        ICompressor compressor = params.getSstableCompressor();
+        if (!(compressor instanceof IDictionaryCompressor))
+        {
+            throw new IllegalArgumentException("Compressor does not support dictionary training: " + params.getSstableCompressor());
+        }
+
+        IDictionaryCompressor dictionaryCompressor = (IDictionaryCompressor) compressor;
+        return dictionaryCompressor.acceptableDictionaryKind().createTrainer(keyspaceName, tableName, config, compressor);
+    }
+
+    enum TrainingStatus
+    {
+        NOT_STARTED,
+        SAMPLING,
+        TRAINING,
+        COMPLETED,
+        FAILED;
+    }
+}

src/java/org/apache/cassandra/db/compression/SSTableChunkSampler.java

@@ -0,0 +1,443 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.util.ArrayList;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+import java.util.concurrent.ThreadLocalRandom;
+
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.cassandra.db.compression.ICompressionDictionaryTrainer.TrainingStatus;
+import org.apache.cassandra.io.compress.CompressionMetadata;
+import org.apache.cassandra.io.compress.ICompressor;
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+import org.apache.cassandra.io.util.ChannelProxy;
+import org.apache.cassandra.utils.ChecksumType;
+
+/**
+ * Samples uncompressed chunks from existing SSTables for dictionary training.
+ * Uses random sampling to locate the chunk offsets to avoid sequential scanning while ensuring representative samples.
+ * Supports both compressed and uncompressed SSTables.
+ */
+public class SSTableChunkSampler
+{
+    private static final Logger logger = LoggerFactory.getLogger(SSTableChunkSampler.class);
+
+    /**
+     * Information about an SSTable and its chunks for sampling.
+     */
+    static class SSTableChunkInfo
+    {
+        final SSTableReader sstable;
+        final CompressionMetadata metadata; // null for uncompressed
+        final long chunkCount;
+        final long dataLength;
+        final int chunkSize;
+        final boolean isCompressed;
+
+        SSTableChunkInfo(SSTableReader sstable, CompressionDictionaryTrainingConfig config)
+        {
+            this.sstable = sstable;
+            this.isCompressed = sstable.compression;
+
+            if (isCompressed)
+            {
+                this.metadata = sstable.getCompressionMetadata();
+                this.dataLength = metadata.dataLength;
+                this.chunkSize = metadata.chunkLength();
+                // Use the logical chunk count from metadata (each offset is 8 bytes)
+                this.chunkCount = metadata.chunkOffsetsSize >> 3;
+            }
+            else
+            {
+                this.metadata = null;
+                this.dataLength = sstable.uncompressedLength();
+                this.chunkSize = config.chunkSize;
+                // Calculate number of chunks for uncompressed: dataLength divided by chunkSize, rounded up
+                this.chunkCount = (dataLength + chunkSize - 1) / chunkSize;
+            }
+        }
+    }
+
+    /**
+     * Samples chunks from existing SSTables and adds them to the trainer.
+     * Uses two-level sampling to avoid memory issues with large datasets:
+     * 1. Select SSTables (potentially all, weighted by size)
+     * 2. For each SSTable, randomly select specific chunks to sample
+     *
+     * @param sstables the set of SSTables to sample from
+     * @param trainer  the trainer to add samples to
+     * @param config   the training configuration with sample size limits
+     */
+    public static void sampleFromSSTables(Set<SSTableReader> sstables,
+                                          ICompressionDictionaryTrainer trainer,
+                                          CompressionDictionaryTrainingConfig config) throws IOException
+    {
+        if (sstables.isEmpty())
+        {
+            throw new IllegalArgumentException("No SSTables provided for sampling");
+        }
+
+        TrainingStatus status = trainer.getTrainingState().status;
+        if (status != TrainingStatus.SAMPLING)
+        {
+            throw new IllegalStateException("Trainer is not ready to accept samples. Current status: " + status);
+        }
+
+        // Build metadata for all SSTables
+        List<SSTableChunkInfo> sstableInfos = buildSSTableInfos(sstables, config);
+        long totalChunks = sstableInfos.stream().mapToLong(info -> info.chunkCount).sum();
+
+        // Calculate how many chunks to sample
+        long targetChunkCount = calculateTargetChunkCount(sstableInfos, totalChunks, config);
+
+        logger.debug("Target chunk count for sampling: {} (max sample size: {} bytes)",
+                     targetChunkCount, config.maxTotalSampleSize);
+
+        // Sample chunks from each SSTable
+        SamplingStats stats = sampleChunksFromSSTables(sstableInfos, totalChunks, targetChunkCount, trainer, config);
+
+        logger.info("Completed sampling: {} chunks, total size: {} bytes", stats.sampleCount, stats.totalSampleSize);
+    }
+
+    /**
+     * Builds SSTableChunkInfo objects for all SSTables and logs statistics.
+     */
+    static List<SSTableChunkInfo> buildSSTableInfos(Set<SSTableReader> sstables,
+                                                    CompressionDictionaryTrainingConfig config)
+    {
+        List<SSTableChunkInfo> sstableInfos = new ArrayList<>();
+        long totalChunks = 0;
+        int compressedCount = 0;
+        int uncompressedCount = 0;
+
+        for (SSTableReader sstable : sstables)
+        {
+            SSTableChunkInfo info = new SSTableChunkInfo(sstable, config);
+            sstableInfos.add(info);
+            totalChunks += info.chunkCount;
+
+            if (info.isCompressed)
+                compressedCount++;
+            else
+                uncompressedCount++;
+        }
+
+        logger.info("Sampling from {} SSTables ({} compressed, {} uncompressed) with {} total chunks",
+                    sstableInfos.size(), compressedCount, uncompressedCount, totalChunks);
+
+        return sstableInfos;
+    }
+
+    /**
+     * Calculates the target number of chunks to sample based on available data and constraints.
+     */
+    static long calculateTargetChunkCount(List<SSTableChunkInfo> sstableInfos,
+                                          long totalChunks,
+                                          CompressionDictionaryTrainingConfig config)
+    {
+        long totalDataSize = sstableInfos.stream().mapToLong(info -> info.dataLength).sum();
+        int averageChunkSize = totalDataSize > 0 ? (int) (totalDataSize / totalChunks) : config.chunkSize;
+        return config.maxTotalSampleSize / averageChunkSize;
+    }
+
+    /**
+     * Result of sampling operation containing statistics.
+     */
+    static class SamplingStats
+    {
+        final long sampleCount;
+        final long totalSampleSize;
+
+        SamplingStats(long sampleCount, long totalSampleSize)
+        {
+            this.sampleCount = sampleCount;
+            this.totalSampleSize = totalSampleSize;
+        }
+    }
+
+    /**
+     * Samples chunks from all SSTables proportionally to their size.
+     * Each SSTable contributes samples in proportion to its chunk count relative to the total.
+     * Stops early if either the target chunk count or max total sample size limit is reached.
+     * <p>
+     * For example,
+     * <pre>
+     * Given:
+     *   - SSTable A: 40 chunks, chunkSize=64KB (40% of total)
+     *   - SSTable B: 60 chunks, chunkSize=64KB (60% of total)
+     *   - Target chunk count: 100
+     *   - Max total sample size: 5MB
+     *
+     * Result:
+     *   - Sample 32 chunks from A (5MiB / 64KiB * 0.4 = 32 chunks)
+     *   - Sample 48 chunks from B (5MiB / 64KiB * 0.6 = 48 chunks)
+     *   - Total sampled: 80 chunks (stopped due to size limit, not target)
+     * </pre>
+     */
+    static SamplingStats sampleChunksFromSSTables(List<SSTableChunkInfo> sstableInfos,
+                                                  long totalChunks,
+                                                  long targetChunkCount,
+                                                  ICompressionDictionaryTrainer trainer,
+                                                  CompressionDictionaryTrainingConfig config) throws IOException
+    {
+        long totalSampleSize = 0;
+        long sampleCount = 0;
+
+        for (SSTableChunkInfo info : sstableInfos)
+        {
+            if (sampleCount >= targetChunkCount || totalSampleSize >= config.maxTotalSampleSize)
+            {
+                break;
+            }
+
+            // Calculate how many chunks to sample from this SSTable (proportional to its size)
+            long remainingTarget = Math.min(targetChunkCount - sampleCount, (config.maxTotalSampleSize - totalSampleSize) / info.chunkSize);
+            long chunksFromThisSSTable = Math.min((targetChunkCount * info.chunkCount) / totalChunks, remainingTarget);
+
+            if (chunksFromThisSSTable <= 0)
+            {
+                continue;
+            }
+
+            // Sample chunks from this SSTable
+            SamplingStats sstableStats = sampleChunksFromSSTable(info, chunksFromThisSSTable, trainer, config);
+            totalSampleSize += sstableStats.totalSampleSize;
+            sampleCount += sstableStats.sampleCount;
+
+            if (sampleCount % 100 == 0)
+            {
+                logger.debug("Sampled {} chunks, total size: {} bytes", sampleCount, totalSampleSize);
+            }
+        }
+
+        return new SamplingStats(sampleCount, totalSampleSize);
+    }
+
+    /**
+     * Samples a specified number of chunks from a single SSTable.
+     */
+    static SamplingStats sampleChunksFromSSTable(SSTableChunkInfo info,
+                                                 long chunksToSample,
+                                                 ICompressionDictionaryTrainer trainer,
+                                                 CompressionDictionaryTrainingConfig config) throws IOException
+    {
+        long totalSampleSize = 0;
+        long sampleCount = 0;
+
+        // Generate random chunk indices for this SSTable (without building full list)
+        Set<Long> selectedIndices = selectRandomChunkIndices(info.chunkCount, chunksToSample);
+
+        // Sample the selected chunks
+        for (long chunkIndex : selectedIndices)
+        {
+            if (totalSampleSize >= config.maxTotalSampleSize)
+            {
+                logger.debug("Reached max total sample size limit");
+                break;
+            }
+
+            long position = chunkIndex * info.chunkSize;
+            ByteBuffer chunk = readChunk(info, position);
+
+            // Check if adding this sample would exceed the max total sample size
+            if (totalSampleSize + chunk.remaining() > config.maxTotalSampleSize)
+            {
+                logger.debug("Next chunk would exceed max total sample size limit");
+                break;
+            }
+
+            trainer.addSample(chunk);
+            totalSampleSize += chunk.remaining();
+            sampleCount++;
+        }
+
+        return new SamplingStats(sampleCount, totalSampleSize);
+    }
+
+    /**
+     * Selects random chunk indices.
+     *
+     * @param totalChunks the total number of chunks available
+     * @param count       the number of chunks to select
+     * @return set of randomly selected chunk indices
+     */
+    static Set<Long> selectRandomChunkIndices(long totalChunks, long count)
+    {
+        // If we need to sample more than half, it's more efficient to select what to exclude
+        if (count > totalChunks / 2)
+        {
+            long excludeCount = totalChunks - count;
+            Set<Long> toExclude = floydRandomSampling(totalChunks, excludeCount);
+            Set<Long> selected = new HashSet<>();
+            // Add all indices except those in toExclude
+            for (long i = 0; i < totalChunks; i++)
+            {
+                if (!toExclude.contains(i))
+                {
+                    selected.add(i);
+                }
+            }
+            return selected;
+        }
+        else
+        {
+            return floydRandomSampling(totalChunks, count);
+        }
+    }
+
+    /**
+     * Floyd's algorithm for random sampling without replacement.
+     * Efficiently selects a random subset by iterating only through the sample size, not the total population.
+     * Guarantees no duplication.
+     *
+     * @param total   the total number of items available
+     * @param samples the number of items to select
+     * @return set of randomly selected indices
+     * @see <a href="https://fermatslibrary.com/s/a-sample-of-brilliance">Floyd's Sampling Algorithm</a>
+     */
+    static Set<Long> floydRandomSampling(long total, long samples)
+    {
+        Set<Long> set = new HashSet<>();
+        long requested = Math.min(total, samples);
+        for (long i = total - requested; i < total; i++)
+        {
+            long randomIndex = ThreadLocalRandom.current().nextLong(i + 1);
+            if (!set.add(randomIndex))
+            {
+                set.add(i);
+            }
+        }
+        return set;
+    }
+
+    /**
+     * Reads a chunk from an SSTable at the given position.
+     * Handles both compressed and uncompressed SSTables.
+     *
+     * @param sstableInfo the SSTable info
+     * @param position    the position to read from
+     * @return the chunk data (uncompressed if source was compressed)
+     * @throws IOException if reading or decompression fails
+     */
+    static ByteBuffer readChunk(SSTableChunkInfo sstableInfo, long position) throws IOException
+    {
+        if (sstableInfo.isCompressed)
+        {
+            return readAndDecompressChunk(sstableInfo, position);
+        }
+        else
+        {
+            return readUncompressedChunk(sstableInfo, position);
+        }
+    }
+
+    /**
+     * Reads and decompresses a single chunk from a compressed SSTable.
+     *
+     * @param sstableInfo the SSTable info
+     * @param position    the uncompressed position (will be mapped to chunk)
+     * @return the uncompressed chunk data
+     * @throws IOException if reading or decompression fails
+     */
+    static ByteBuffer readAndDecompressChunk(SSTableChunkInfo sstableInfo, long position) throws IOException
+    {
+        CompressionMetadata metadata = sstableInfo.metadata;
+        CompressionMetadata.Chunk chunk = metadata.chunkFor(position);
+
+        // Read the compressed chunk from disk
+        ChannelProxy channel = sstableInfo.sstable.getDataChannel();
+
+        // Allocate buffer for compressed data + checksum
+        int compressedLength = chunk.length;
+        ByteBuffer compressed = ByteBuffer.allocateDirect(compressedLength + Integer.BYTES);
+
+        int read = channel.read(compressed, chunk.offset);
+        if (read != compressedLength + Integer.BYTES)
+        {
+            throw new IOException(String.format("Expected to read %d bytes but got %d",
+                                                compressedLength + Integer.BYTES, read));
+        }
+
+        compressed.flip();
+        compressed.limit(compressedLength);
+
+        // Verify checksum
+        int expectedChecksum = (int) ChecksumType.CRC32.of(compressed);
+        compressed.limit(compressedLength + Integer.BYTES);
+        int actualChecksum = compressed.getInt(compressedLength);
+
+        if (expectedChecksum != actualChecksum)
+        {
+            throw new IOException(String.format("Checksum mismatch for chunk at position %d in SSTable %s (expected: %d, actual: %d)",
+                                                position, sstableInfo.sstable, expectedChecksum, actualChecksum));
+        }
+
+        // Reset for decompression
+        compressed.position(0).limit(compressedLength);
+
+        // Decompress the chunk
+        ICompressor compressor = metadata.compressor();
+        ByteBuffer uncompressed = ByteBuffer.allocateDirect(metadata.chunkLength());
+
+        compressor.uncompress(compressed, uncompressed);
+        uncompressed.flip();
+        return uncompressed;
+    }
+
+    /**
+     * Reads a chunk directly from an uncompressed SSTable.
+     *
+     * @param sstableInfo the SSTable info
+     * @param position    the position to read from
+     * @return the chunk data
+     * @throws IOException if reading fails
+     */
+    static ByteBuffer readUncompressedChunk(SSTableChunkInfo sstableInfo, long position) throws IOException
+    {
+        ChannelProxy channel = sstableInfo.sstable.getDataChannel();
+
+        // Calculate how much to read (might be less than chunkSize at end of file)
+        long remainingData = sstableInfo.dataLength - position;
+        int readSize = (int) Math.min(sstableInfo.chunkSize, remainingData);
+
+        if (readSize <= 0)
+        {
+            throw new IOException(String.format("Invalid read size %d at position %d (dataLength: %d) for SSTable %s",
+                                                readSize, position, sstableInfo.dataLength, sstableInfo.sstable));
+        }
+
+        ByteBuffer buffer = ByteBuffer.allocateDirect(readSize);
+        int read = channel.read(buffer, position);
+
+        if (read != readSize)
+        {
+            throw new IOException(String.format("Expected to read %d bytes but got %d", readSize, read));
+        }
+
+        buffer.flip();
+        return buffer;
+    }
+}

src/java/org/apache/cassandra/db/compression/TrainingState.java

@@ -0,0 +1,218 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.HashMap;
+import java.util.Map;
+import javax.management.openmbean.CompositeData;
+import javax.management.openmbean.CompositeDataSupport;
+import javax.management.openmbean.CompositeType;
+import javax.management.openmbean.OpenDataException;
+import javax.management.openmbean.OpenType;
+import javax.management.openmbean.SimpleType;
+
+import org.apache.cassandra.db.compression.ICompressionDictionaryTrainer.TrainingStatus;
+
+/**
+ * Represents the current state of compression dictionary training.
+ * This class encapsulates training status, progress information, and failure details in a single snapshot.
+ */
+public class TrainingState
+{
+    public final TrainingStatus status;
+    public final String failureMessage;  // null unless status is FAILED
+    public final long sampleCount;
+    public final long totalSampleSize;
+
+    // JMX CompositeData support
+    private static final String[] ITEM_NAMES = new String[]{ "status",
+                                                             "failure_message",
+                                                             "sample_count",
+                                                             "total_sample_size" };
+
+    private static final String[] ITEM_DESC = new String[]{ "current training status",
+                                                            "failure message if training failed, null otherwise",
+                                                            "number of samples collected",
+                                                            "total size of samples collected in bytes" };
+
+    private static final OpenType<?>[] ITEM_TYPES;
+
+    public static final CompositeType COMPOSITE_TYPE;
+
+    static
+    {
+        try
+        {
+            ITEM_TYPES = new OpenType[]{ SimpleType.STRING,
+                                         SimpleType.STRING,
+                                         SimpleType.LONG,
+                                         SimpleType.LONG };
+
+            COMPOSITE_TYPE = new CompositeType(TrainingState.class.getName(),
+                                               "TrainingState",
+                                               ITEM_NAMES,
+                                               ITEM_DESC,
+                                               ITEM_TYPES);
+        }
+        catch (OpenDataException e)
+        {
+            throw new RuntimeException(e);
+        }
+    }
+
+    private TrainingState(TrainingStatus status, String failureMessage, long sampleCount, long totalSampleSize)
+    {
+        this.status = status;
+        this.failureMessage = failureMessage;
+        this.sampleCount = sampleCount;
+        this.totalSampleSize = totalSampleSize;
+    }
+
+    // Factory methods for clarity and type safety
+    public static TrainingState notStarted()
+    {
+        return new TrainingState(TrainingStatus.NOT_STARTED, null, 0, 0);
+    }
+
+    public static TrainingState sampling(long samples, long totalSize)
+    {
+        return new TrainingState(TrainingStatus.SAMPLING, null, samples, totalSize);
+    }
+
+    public static TrainingState training(long samples, long totalSize)
+    {
+        return new TrainingState(TrainingStatus.TRAINING, null, samples, totalSize);
+    }
+
+    public static TrainingState completed(long samples, long totalSize)
+    {
+        return new TrainingState(TrainingStatus.COMPLETED, null, samples, totalSize);
+    }
+
+    public static TrainingState failed(String message, long samples, long totalSize)
+    {
+        return new TrainingState(TrainingStatus.FAILED, message, samples, totalSize);
+    }
+
+    public boolean isFailed()
+    {
+        return status == TrainingStatus.FAILED;
+    }
+
+    public boolean isCompleted()
+    {
+        return status == TrainingStatus.COMPLETED;
+    }
+
+    public TrainingStatus getStatus()
+    {
+        return status;
+    }
+
+    public long getSampleCount()
+    {
+        return sampleCount;
+    }
+
+    public long getTotalSampleSize()
+    {
+        return totalSampleSize;
+    }
+
+    // returns null unless status is FAILED
+    public String getFailureMessage()
+    {
+        return isFailed() ? failureMessage : null;
+    }
+
+    // JMX CompositeData conversion methods
+
+    /**
+     * Converts this TrainingState to JMX CompositeData format.
+     *
+     * @return CompositeData representation of this training state
+     */
+    public CompositeData toCompositeData()
+    {
+        Map<String, Object> valueMap = new HashMap<>();
+        valueMap.put(ITEM_NAMES[0], status.toString());
+        valueMap.put(ITEM_NAMES[1], getFailureMessage());
+        valueMap.put(ITEM_NAMES[2], sampleCount);
+        valueMap.put(ITEM_NAMES[3], totalSampleSize);
+
+        try
+        {
+            return new CompositeDataSupport(COMPOSITE_TYPE, valueMap);
+        }
+        catch (final OpenDataException e)
+        {
+            throw new RuntimeException(e);
+        }
+    }
+
+    /**
+     * Converts JMX CompositeData back to TrainingState.
+     *
+     * @param data the CompositeData to convert
+     * @return TrainingState reconstructed from the CompositeData
+     */
+    public static TrainingState fromCompositeData(final CompositeData data)
+    {
+        assert data.getCompositeType().equals(COMPOSITE_TYPE);
+
+        final Object[] values = data.getAll(ITEM_NAMES);
+
+        TrainingStatus status = TrainingStatus.valueOf((String) values[0]);
+        String failureMessage = (String) values[1];
+        long sampleCount = (Long) values[2];
+        long totalSampleSize = (Long) values[3];
+
+        // Reconstruct TrainingState based on status
+        switch (status)
+        {
+            case NOT_STARTED:
+                return TrainingState.notStarted();
+            case SAMPLING:
+                return TrainingState.sampling(sampleCount, totalSampleSize);
+            case TRAINING:
+                return TrainingState.training(sampleCount, totalSampleSize);
+            case COMPLETED:
+                return TrainingState.completed(sampleCount, totalSampleSize);
+            case FAILED:
+                return TrainingState.failed(failureMessage, sampleCount, totalSampleSize);
+            default:
+                throw new IllegalStateException("Unknown training status: " + status);
+        }
+    }
+
+    @Override
+    public String toString()
+    {
+        StringBuilder sb = new StringBuilder("TrainingState{status=");
+        sb.append(status);
+        sb.append(", samples=").append(sampleCount);
+        sb.append(", totalSize=").append(totalSampleSize);
+        if (isFailed() && failureMessage != null)
+        {
+            sb.append(", failure='").append(failureMessage).append('\'');
+        }
+        sb.append('}');
+        return sb.toString();
+    }
+}

src/java/org/apache/cassandra/db/compression/ZstdCompressionDictionary.java

@@ -0,0 +1,217 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.Objects;
+import java.util.concurrent.ConcurrentHashMap;
+import java.util.concurrent.atomic.AtomicBoolean;
+
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import com.github.luben.zstd.ZstdDictCompress;
+import com.github.luben.zstd.ZstdDictDecompress;
+import org.apache.cassandra.io.compress.ZstdCompressorBase;
+import org.apache.cassandra.utils.concurrent.Ref;
+import org.apache.cassandra.utils.concurrent.RefCounted;
+import org.apache.cassandra.utils.concurrent.SelfRefCounted;
+
+public class ZstdCompressionDictionary implements CompressionDictionary, SelfRefCounted<ZstdCompressionDictionary>
+{
+    private static final Logger logger = LoggerFactory.getLogger(ZstdCompressionDictionary.class);
+
+    private final DictId dictId;
+    private final byte[] rawDictionary;
+    // One ZstdDictDecompress and multiple ZstdDictCompress (per level) can be derived from the same raw dictionary content
+    private final ConcurrentHashMap<Integer, ZstdDictCompress> zstdDictCompressPerLevel = new ConcurrentHashMap<>();
+    private volatile ZstdDictDecompress dictDecompress;
+    private final AtomicBoolean closed = new AtomicBoolean(false);
+    private final Ref<ZstdCompressionDictionary> selfRef;
+
+    public ZstdCompressionDictionary(DictId dictId, byte[] rawDictionary)
+    {
+        this.dictId = dictId;
+        this.rawDictionary = rawDictionary;
+        this.selfRef = new Ref<>(this, new Tidy(zstdDictCompressPerLevel, dictDecompress));
+    }
+
+    @Override
+    public DictId dictId()
+    {
+        return dictId;
+    }
+
+    @Override
+    public Kind kind()
+    {
+        return Kind.ZSTD;
+    }
+
+    @Override
+    public byte[] rawDictionary()
+    {
+        return rawDictionary;
+    }
+
+    @Override
+    public boolean equals(Object o)
+    {
+        if (!(o instanceof ZstdCompressionDictionary)) return false;
+        ZstdCompressionDictionary that = (ZstdCompressionDictionary) o;
+        return Objects.equals(dictId, that.dictId);
+    }
+
+    @Override
+    public int hashCode()
+    {
+        return dictId.hashCode();
+    }
+
+    /**
+     * Get a pre-processed compression tables that is optimized for compression.
+     * It is derived/computed from dictionary bytes.
+     * The internal data structure is different from the tables for decompression.
+     *
+     * @param compressionLevel compression level to create the compression table
+     * @return ZstdDictCompress
+     */
+    public ZstdDictCompress dictionaryForCompression(int compressionLevel)
+    {
+        if (closed.get())
+            throw new IllegalStateException("Dictionary has been closed. " + dictId);
+
+        ZstdCompressorBase.validateCompressionLevel(compressionLevel);
+
+        return zstdDictCompressPerLevel.computeIfAbsent(compressionLevel, level -> {
+            if (closed.get())
+                throw new IllegalStateException("Dictionary has been closed");
+            return new ZstdDictCompress(rawDictionary, level);
+        });
+    }
+
+    /**
+     * Get a pre-processed decompression tables that is optimized for decompression.
+     * It is derived/computed from dictionary bytes.
+     * The internal data structure is different from the tables for compression.
+     *
+     * @return ZstdDictDecompress
+     */
+    public ZstdDictDecompress dictionaryForDecompression()
+    {
+        if (closed.get())
+            throw new IllegalStateException("Dictionary has been closed");
+
+        ZstdDictDecompress result = dictDecompress;
+        if (result != null)
+            return result;
+
+        synchronized (this)
+        {
+            if (closed.get())
+                throw new IllegalStateException("Dictionary has been closed");
+
+            result = dictDecompress;
+            if (result == null)
+            {
+                result = new ZstdDictDecompress(rawDictionary);
+                dictDecompress = result;
+            }
+            return result;
+        }
+    }
+
+    @Override
+    public Ref<ZstdCompressionDictionary> tryRef()
+    {
+        return selfRef.tryRef();
+    }
+
+    @Override
+    public Ref<ZstdCompressionDictionary> selfRef()
+    {
+        return selfRef;
+    }
+
+    @Override
+    public Ref<ZstdCompressionDictionary> ref()
+    {
+        return selfRef.ref();
+    }
+
+    @Override
+    public void close()
+    {
+        if (closed.compareAndSet(false, true))
+        {
+            selfRef.release();
+        }
+    }
+
+    private static class Tidy implements RefCounted.Tidy
+    {
+        private final ConcurrentHashMap<Integer, ZstdDictCompress> zstdDictCompressPerLevel;
+        private volatile ZstdDictDecompress dictDecompress;
+
+        Tidy(ConcurrentHashMap<Integer, ZstdDictCompress> zstdDictCompressPerLevel, ZstdDictDecompress dictDecompress)
+        {
+            this.zstdDictCompressPerLevel = zstdDictCompressPerLevel;
+            this.dictDecompress = dictDecompress;
+        }
+
+        @Override
+        public void tidy()
+        {
+            // Close all compression dictionaries
+            for (ZstdDictCompress compressDict : zstdDictCompressPerLevel.values())
+            {
+                try
+                {
+                    compressDict.close();
+                }
+                catch (Exception e)
+                {
+                    // Log but don't fail - continue closing other resources
+                    logger.warn("Failed to close ZstdDictCompress", e);
+                }
+            }
+            zstdDictCompressPerLevel.clear();
+
+            // Close decompression dictionary
+            ZstdDictDecompress decompressDict = dictDecompress;
+            if (decompressDict != null)
+            {
+                try
+                {
+                    decompressDict.close();
+                }
+                catch (Exception e)
+                {
+                    logger.warn("Failed to close ZstdDictDecompress", e);
+                }
+                dictDecompress = null;
+            }
+        }
+
+        @Override
+        public String name()
+        {
+            return ZstdCompressionDictionary.class.getSimpleName();
+        }
+    }
+}

src/java/org/apache/cassandra/db/compression/ZstdDictionaryTrainer.java

@@ -0,0 +1,470 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.nio.ByteBuffer;
+import java.util.concurrent.ThreadLocalRandom;
+import java.util.concurrent.atomic.AtomicLong;
+import java.util.function.Consumer;
+
+import com.google.common.annotations.VisibleForTesting;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import com.github.luben.zstd.Zstd;
+import com.github.luben.zstd.ZstdDictTrainer;
+import org.apache.cassandra.concurrent.ScheduledExecutors;
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import org.apache.cassandra.io.compress.IDictionaryCompressor;
+import org.apache.cassandra.io.compress.ZstdDictionaryCompressor;
+import org.apache.cassandra.io.util.FileUtils;
+import org.apache.cassandra.schema.CompressionParams;
+import org.apache.cassandra.utils.Clock;
+import org.apache.cassandra.utils.concurrent.AsyncFuture;
+import org.apache.cassandra.utils.concurrent.Future;
+
+/**
+ * Zstd implementation of dictionary trainer with lifecycle management.
+ */
+public class ZstdDictionaryTrainer implements ICompressionDictionaryTrainer
+{
+    private static final Logger logger = LoggerFactory.getLogger(ZstdDictionaryTrainer.class);
+
+    private final String keyspaceName;
+    private final String tableName;
+    private final CompressionDictionaryTrainingConfig config;
+    private final AtomicLong totalSampleSize;
+    private final AtomicLong sampleCount;
+    private final int compressionLevel; // optimal if using the same level for training as when compressing.
+
+    // Sampling rate can be updated during training
+    private volatile int samplingRate;
+
+    // Minimum number of samples required by ZSTD library
+    private static final int MIN_SAMPLES_REQUIRED = 11;
+
+    private volatile Consumer<CompressionDictionary> dictionaryTrainedListener;
+    // TODO: manage the samples in this class for auto-train (follow-up). The ZstdDictTrainer cannot be re-used for multiple training runs.
+    private ZstdDictTrainer zstdTrainer;
+    private volatile boolean closed = false;
+    private volatile TrainingStatus currentTrainingStatus;
+    private volatile String failureMessage;
+
+    public ZstdDictionaryTrainer(String keyspaceName, String tableName,
+                                 CompressionDictionaryTrainingConfig config,
+                                 int compressionLevel)
+    {
+        this.keyspaceName = keyspaceName;
+        this.tableName = tableName;
+        this.config = config;
+        this.totalSampleSize = new AtomicLong(0);
+        this.sampleCount = new AtomicLong(0);
+        this.compressionLevel = compressionLevel;
+        this.samplingRate = config.samplingRate;
+        this.currentTrainingStatus = TrainingStatus.NOT_STARTED;
+    }
+
+    @Override
+    public boolean shouldSample()
+    {
+        return zstdTrainer != null && ThreadLocalRandom.current().nextInt(samplingRate) == 0;
+    }
+
+    @Override
+    public void addSample(ByteBuffer sample)
+    {
+        if (closed || sample == null || !sample.hasRemaining() || zstdTrainer == null)
+            return;
+
+        byte[] sampleBytes = new byte[sample.remaining()];
+        sample.duplicate().get(sampleBytes);
+
+        if (zstdTrainer.addSample(sampleBytes))
+        {
+            // Update the totalSampleSize and sampleCount if the sample is added
+            totalSampleSize.addAndGet(sampleBytes.length);
+            sampleCount.incrementAndGet();
+        }
+    }
+
+    @Override
+    public CompressionDictionary trainDictionary(boolean force)
+    {
+        boolean isReady = isReady();
+        if (!force && !isReady)
+        {
+            failureMessage = buildNotReadyMessage();
+            currentTrainingStatus = TrainingStatus.FAILED;
+            throw new IllegalStateException(failureMessage);
+        }
+
+        long currentSampleCount = sampleCount.get();
+        if (currentSampleCount < MIN_SAMPLES_REQUIRED) // minimum samples should be required even if force training
+        {
+            failureMessage = String.format("Insufficient samples for training: %d (minimum required: %d)",
+                                           currentSampleCount, MIN_SAMPLES_REQUIRED);
+            currentTrainingStatus = TrainingStatus.FAILED;
+            throw new IllegalStateException(failureMessage);
+        }
+
+        currentTrainingStatus = TrainingStatus.TRAINING;
+        failureMessage = null; // Clear any previous failure message
+        try
+        {
+            logger.debug("Training with sample count: {}, sample size: {}, isReady: {}",
+                         currentSampleCount, totalSampleSize.get(), isReady);
+            byte[] dictBytes = zstdTrainer.trainSamples();
+            long zstdDictId = Zstd.getDictIdFromDict(dictBytes);
+            DictId dictId = new DictId(Kind.ZSTD, makeDictionaryId(Clock.Global.currentTimeMillis(), zstdDictId));
+            currentTrainingStatus = TrainingStatus.COMPLETED;
+            logger.debug("New dictionary is trained with {}", dictId);
+            CompressionDictionary dictionary = new ZstdCompressionDictionary(dictId, dictBytes);
+            notifyDictionaryTrainedListener(dictionary);
+            return dictionary;
+        }
+        catch (Exception e)
+        {
+            failureMessage = "Failed to train Zstd dictionary: " + e.getMessage();
+            currentTrainingStatus = TrainingStatus.FAILED;
+            throw new RuntimeException(failureMessage, e);
+        }
+    }
+
+    @Override
+    public Future<CompressionDictionary> trainDictionaryAsync(boolean force)
+    {
+        DictionaryTrainingTask task = new DictionaryTrainingTask(force);
+        ScheduledExecutors.nonPeriodicTasks.execute(task);
+        return task;
+    }
+
+    /**
+     * Async task for training dictionary that handles failures without routing through JVMStabilityInspector.
+     * Follows the pattern used by repair tasks (ValidationTask, SyncTask, etc.):
+     * - Extends AsyncFuture and implements Runnable
+     * - Never throws exceptions from run()
+     * - Calls trySuccess()/tryFailure() to complete the future
+     *
+     * This ensures validation failures are handled cleanly without ERROR logging to JVMStabilityInspector,
+     * while unexpected errors are still properly logged.
+     */
+    private class DictionaryTrainingTask extends AsyncFuture<CompressionDictionary> implements Runnable
+    {
+        private final boolean force;
+
+        DictionaryTrainingTask(boolean force)
+        {
+            this.force = force;
+        }
+
+        @Override
+        public void run()
+        {
+            try
+            {
+                CompressionDictionary dict = trainDictionary(force);
+                trySuccess(dict);
+            }
+            catch (IllegalStateException e)
+            {
+                logger.debug("Dictionary training validation failed for {}.{}: {}",
+                             keyspaceName, tableName, e.getMessage());
+                tryFailure(e);
+            }
+            catch (Throwable t)
+            {
+                // Unexpected failures - log at error level for visibility
+                logger.error("Unexpected error during dictionary training for {}.{}",
+                             keyspaceName, tableName, t);
+                tryFailure(t);
+            }
+        }
+    }
+
+    /**
+     * Builds a detailed message explaining why the trainer is not ready.
+     *
+     * @return error message with details about what conditions are not met
+     */
+    private String buildNotReadyMessage()
+    {
+        StringBuilder message = new StringBuilder("Trainer is not ready");
+
+        if (closed)
+        {
+            message.append(": trainer is closed");
+            return message.toString();
+        }
+
+        if (currentTrainingStatus == TrainingStatus.TRAINING)
+        {
+            message.append(": training is already in progress");
+            return message.toString();
+        }
+
+        if (zstdTrainer == null)
+        {
+            message.append(": trainer not initialized (call start() first)");
+            return message.toString();
+        }
+
+        long currentSampleCount = sampleCount.get();
+        long currentTotalSampleSize = totalSampleSize.get();
+
+        // Check both sample count and total sample size
+        boolean hasEnoughSamples = currentSampleCount >= MIN_SAMPLES_REQUIRED;
+        boolean hasEnoughSampleSize = currentTotalSampleSize >= config.acceptableTotalSampleSize;
+
+        if (!hasEnoughSamples && !hasEnoughSampleSize)
+        {
+            message.append(String.format(": insufficient samples collected (have %d/%d samples, %s/%s). Use --force to train anyway.",
+                                         currentSampleCount, MIN_SAMPLES_REQUIRED,
+                                         FileUtils.stringifyFileSize(currentTotalSampleSize, true),
+                                         FileUtils.stringifyFileSize(config.acceptableTotalSampleSize, true)));
+        }
+        else if (!hasEnoughSamples)
+        {
+            message.append(String.format(": insufficient sample count (have %d/%d samples). Use --force to train anyway.",
+                                         currentSampleCount, MIN_SAMPLES_REQUIRED));
+        }
+        else if (!hasEnoughSampleSize)
+        {
+            message.append(String.format(": insufficient sample size (have %s/%s). Use --force to train anyway.",
+                                         FileUtils.stringifyFileSize(currentTotalSampleSize, true),
+                                         FileUtils.stringifyFileSize(config.acceptableTotalSampleSize, true)));
+        }
+
+        return message.toString();
+    }
+
+    @Override
+    public boolean isReady()
+    {
+        return currentTrainingStatus != TrainingStatus.TRAINING
+               && !closed
+               && zstdTrainer != null
+               && totalSampleSize.get() >= config.acceptableTotalSampleSize
+               && sampleCount.get() >= MIN_SAMPLES_REQUIRED;
+    }
+
+    @Override
+    public TrainingState getTrainingState()
+    {
+        long currentSampleCount = sampleCount.get();
+        long currentTotalSampleSize = totalSampleSize.get();
+
+        switch (currentTrainingStatus)
+        {
+            case NOT_STARTED:
+                return TrainingState.notStarted();
+            case SAMPLING:
+                return TrainingState.sampling(currentSampleCount, currentTotalSampleSize);
+            case TRAINING:
+                return TrainingState.training(currentSampleCount, currentTotalSampleSize);
+            case COMPLETED:
+                return TrainingState.completed(currentSampleCount, currentTotalSampleSize);
+            case FAILED:
+                return TrainingState.failed(failureMessage, currentSampleCount, currentTotalSampleSize);
+            default:
+                throw new IllegalStateException("Unknown training status: " + currentTrainingStatus);
+        }
+    }
+
+    @Override
+    public boolean start(boolean manualTraining)
+    {
+        if (closed || !(manualTraining || shouldAutoStartTraining()))
+            return false;
+
+        try
+        {
+            // reset on starting; a new zstdTrainer instance is created during reset
+            reset();
+            logger.info("Started dictionary training for {}.{}", keyspaceName, tableName);
+            currentTrainingStatus = TrainingStatus.SAMPLING;
+            failureMessage = null; // Clear any previous failure message
+            return true;
+        }
+        catch (Exception e)
+        {
+            logger.warn("Failed to create ZstdDictTrainer for {}.{}", keyspaceName, tableName, e);
+            failureMessage = "Failed to create ZstdDictTrainer: " + e.getMessage();
+            currentTrainingStatus = TrainingStatus.FAILED;
+        }
+        return false;
+    }
+
+    /**
+     * Determines if training should auto-start based on configuration.
+     */
+    private boolean shouldAutoStartTraining()
+    {
+        return DatabaseDescriptor.getCompressionDictionaryTrainingAutoTrainEnabled();
+    }
+
+    @Override
+    public void reset()
+    {
+        if (closed)
+        {
+            return;
+        }
+
+        currentTrainingStatus = TrainingStatus.NOT_STARTED;
+        synchronized (this)
+        {
+            totalSampleSize.set(0);
+            sampleCount.set(0);
+            zstdTrainer = new ZstdDictTrainer(config.maxTotalSampleSize, config.maxDictionarySize, compressionLevel);
+        }
+    }
+
+    @Override
+    public Kind kind()
+    {
+        return Kind.ZSTD;
+    }
+
+    @Override
+    public void setDictionaryTrainedListener(Consumer<CompressionDictionary> listener)
+    {
+        this.dictionaryTrainedListener = listener;
+    }
+
+    @Override
+    public void updateSamplingRate(int newSamplingRate)
+    {
+        if (newSamplingRate <= 0)
+        {
+            throw new IllegalArgumentException("Sampling rate must be positive, got: " + newSamplingRate);
+        }
+        this.samplingRate = newSamplingRate;
+        logger.debug("Updated sampling rate to {} for {}.{}", newSamplingRate, keyspaceName, tableName);
+    }
+
+    /**
+     * Notifies the registered listener that a dictionary has been trained.
+     *
+     * @param dictionary the newly trained dictionary
+     */
+    private void notifyDictionaryTrainedListener(CompressionDictionary dictionary)
+    {
+        Consumer<CompressionDictionary> listener = this.dictionaryTrainedListener;
+        if (listener != null)
+        {
+            try
+            {
+                listener.accept(dictionary);
+            }
+            catch (Exception e)
+            {
+                logger.warn("Error notifying dictionary trained listener for {}.{}", keyspaceName, tableName, e);
+            }
+        }
+    }
+
+    @Override
+    public boolean isCompatibleWith(CompressionParams newParams)
+    {
+        if (!newParams.isDictionaryCompressionEnabled())
+        {
+            return false;
+        }
+
+        IDictionaryCompressor newCompressor = (IDictionaryCompressor) newParams.getSstableCompressor();
+
+        // Check if the compressor type is compatible with this trainer
+        if (newCompressor.acceptableDictionaryKind() != Kind.ZSTD)
+        {
+            return false;
+        }
+
+        ZstdDictionaryCompressor zstdDictionaryCompressor = (ZstdDictionaryCompressor) newCompressor;
+        // For Zstd compressors, check if compression level matches
+        return this.compressionLevel == zstdDictionaryCompressor.compressionLevel();
+    }
+
+    @Override
+    public void close()
+    {
+        if (closed)
+            return;
+
+        closed = true;
+        currentTrainingStatus = TrainingStatus.NOT_STARTED;
+
+        synchronized (this)
+        {
+            // Permanent shutdown: clear all state and prevent restart
+            totalSampleSize.set(0);
+            sampleCount.set(0);
+            zstdTrainer = null;
+        }
+
+        logger.info("Permanently closed dictionary trainer for {}.{}", keyspaceName, tableName);
+    }
+
+    /**
+     * Custom epoch for dictionary ID timestamps: October 20, 2025 00:00:00 UTC
+     * Calculated as: Instant.parse("2025-10-20T00:00:00Z").toEpochMilli()
+     * This allows using signed 32-bit seconds for ±68 years (~1957 to ~2093)
+     */
+    public static final long CUSTOM_EPOCH_MILLIS = 1760889600000L;
+
+    /**
+     * Creates a monotonically increasing dictionary ID by combining timestamp and Zstd dictionary ID.
+     * This is a public API to support external dictionary imports.
+     * <p>
+     * The resulting dictionary ID has the following structure:
+     * - Upper 32 bits (first 4 bytes): timestamp in seconds since custom epoch (signed int)
+     * - Lower 32 bits (last 4 bytes): Zstd dictionary ID (unsigned int, passed as long)
+     * <p>
+     * Custom epoch: October 20, 2025 00:00:00 UTC ({@link #CUSTOM_EPOCH_MILLIS})
+     * <p>
+     * Using signed 32-bit seconds allows representing ±2,147,483,648 seconds (±68.1 years),
+     * giving a valid range from ~1957 to ~2093, which is sufficient for the software's lifespan.
+     * <p>
+     * This ensures dictionary IDs are monotonically increasing over time, helping to identify
+     * the latest dictionary.
+     *
+     * @param currentTimeMillis the current time in milliseconds since Unix epoch
+     * @param dictId            Zstd dictionary ID (unsigned 32-bit value represented as long)
+     * @return combined dictionary ID that is monotonically increasing over time
+     */
+    public static long makeDictionaryId(long currentTimeMillis, long dictId)
+    {
+        // timestamp in seconds since custom epoch
+        long timestampSeconds = (currentTimeMillis - CUSTOM_EPOCH_MILLIS) / 1000;
+        // Shift timestamp to upper 32 bits
+        long combined = timestampSeconds << 32;
+
+        // Add the unsigned int (already as long) to lower 32 bits
+        combined |= (dictId & 0xFFFFFFFFL);
+
+        return combined;
+    }
+
+    @VisibleForTesting
+    Object trainer()
+    {
+        return zstdTrainer;
+    }
+}

src/java/org/apache/cassandra/io/compress/CompressedSequentialWriter.java

@@ -24,7 +24,10 @@ import java.nio.ByteBuffer;
 import java.nio.channels.Channels;
 import java.util.Optional;
 import java.util.zip.CRC32;
+import javax.annotation.Nullable;
 
+import org.apache.cassandra.db.compression.CompressionDictionary;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.io.FSReadError;
 import org.apache.cassandra.io.FSWriteError;
 import org.apache.cassandra.io.sstable.CorruptSSTableException;
@@ -61,11 +64,24 @@ public class CompressedSequentialWriter extends SequentialWriter
     private long uncompressedSize = 0, compressedSize = 0;
 
     private final MetadataCollector sstableMetadataCollector;
+    private final CompressionDictionaryManager compressionDictionaryManager;
 
     private final ByteBuffer crcCheckBuffer = ByteBuffer.allocate(4);
     private final Optional<File> digestFile;
 
     private final int maxCompressedLength;
+    private final boolean isDictionaryEnabled;
+
+    public CompressedSequentialWriter(File file,
+                                      File offsetsFile,
+                                      File digestFile,
+                                      SequentialWriterOption option,
+                                      CompressionParams parameters,
+                                      MetadataCollector sstableMetadataCollector)
+    {
+        this(file, offsetsFile, digestFile, option, parameters, sstableMetadataCollector, null);
+    }
+
 
     /**
      * Create CompressedSequentialWriter without digest file.
@@ -74,15 +90,17 @@ public class CompressedSequentialWriter extends SequentialWriter
      * @param offsetsFile File to write compression metadata
      * @param digestFile File to write digest
      * @param option Write option (buffer size and type will be set the same as compression params)
-     * @param parameters Compression mparameters
+     * @param parameters Compression parameters
      * @param sstableMetadataCollector Metadata collector
+     * @param compressionDictionaryManager manages compression dictionary; null if absent
      */
     public CompressedSequentialWriter(File file,
                                       File offsetsFile,
                                       File digestFile,
                                       SequentialWriterOption option,
                                       CompressionParams parameters,
-                                      MetadataCollector sstableMetadataCollector)
+                                      MetadataCollector sstableMetadataCollector,
+                                      @Nullable CompressionDictionaryManager compressionDictionaryManager)
     {
         super(file, SequentialWriterOption.newBuilder()
                             .bufferSize(option.bufferSize())
@@ -91,7 +109,7 @@ public class CompressedSequentialWriter extends SequentialWriter
                             .bufferType(parameters.getSstableCompressor().preferredBufferType())
                             .finishOnClose(option.finishOnClose())
                             .build());
-        this.compressor = parameters.getSstableCompressor();
+        ICompressor compressor = parameters.getSstableCompressor();
         this.digestFile = Optional.ofNullable(digestFile);
 
         // buffer for compression should be the same size as buffer itself
@@ -99,8 +117,28 @@ public class CompressedSequentialWriter extends SequentialWriter
 
         maxCompressedLength = parameters.maxCompressedLength();
 
+        // Note that we cannot rely on the compressor type to tell whether dictionary compression is enabled.
+        // Because the `CompressionParams` for this method is updated at the callsite, `DataComponent.buildWriter`.
+        // See CASSANDRA-15379 for details regarding the optimization.
+        // Meanwhile, as long as dictionary-based compression is enabled, we want to collect samples.
+        this.isDictionaryEnabled = compressionDictionaryManager != null && compressionDictionaryManager.isEnabled();
+
+        CompressionDictionary compressionDictionary = compressionDictionaryManager == null ? null : compressionDictionaryManager.getCurrent();
+        if (compressionDictionary != null && compressor instanceof IDictionaryCompressor)
+        {
+            compressor = ((IDictionaryCompressor) compressor).getOrCopyWithDictionary(compressionDictionary);
+        }
+        else
+        {
+            // It is likely on the sstable flushing path and LZ4 compressor or something else is picked.
+            // In this case, we disable the compression dictionary, i.e. do not attach the dictionary
+            // bytes to the CompressionInfo component.
+            compressionDictionary = null;
+        }
+        this.compressor = compressor;
+        this.compressionDictionaryManager = compressionDictionaryManager;
         /* Index File (-CompressionInfo.db component) and it's header */
-        metadataWriter = CompressionMetadata.Writer.open(parameters, offsetsFile);
+        metadataWriter = CompressionMetadata.Writer.open(parameters, offsetsFile, compressionDictionary);
 
         this.sstableMetadataCollector = sstableMetadataCollector;
         crcMetadata = new ChecksumWriter(new DataOutputStream(Channels.newOutputStream(channel)));
@@ -145,6 +183,13 @@ public class CompressedSequentialWriter extends SequentialWriter
         {
             // compressing data with buffer re-use
             buffer.flip();
+            
+            // Collect sample for dictionary training before compression
+            if (isDictionaryEnabled)
+            {
+                compressionDictionaryManager.addSample(buffer.duplicate());
+            }
+            
             compressed.clear();
             compressor.compress(buffer, compressed);
         }
@@ -440,4 +485,4 @@ public class CompressedSequentialWriter extends SequentialWriter
             this.nextChunkIndex = nextChunkIndex;
         }
     }
-}
+}

src/java/org/apache/cassandra/io/compress/CompressionMetadata.java

@@ -27,11 +27,14 @@ import java.util.HashMap;
 import java.util.Map;
 import java.util.SortedSet;
 import java.util.TreeSet;
+import javax.annotation.Nullable;
 
 import com.google.common.annotations.VisibleForTesting;
 import com.google.common.primitives.Longs;
 
 import org.apache.cassandra.db.TypeSizes;
+import org.apache.cassandra.db.compression.CompressionDictionary;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.exceptions.ConfigurationException;
 import org.apache.cassandra.io.FSReadError;
 import org.apache.cassandra.io.FSWriteError;
@@ -62,16 +65,31 @@ public class CompressionMetadata extends WrappedSharedCloseable
     public final long dataLength;
     public final long compressedFileLength;
     private final Memory chunkOffsets;
-    private final long chunkOffsetsSize;
+    public final long chunkOffsetsSize;
     public final File chunksIndexFile;
     public final CompressionParams parameters;
+    @Nullable // null when no dictionary
+    private final CompressionDictionary compressionDictionary;
+    private volatile ICompressor resolvedCompressor;
 
     @VisibleForTesting
-    public static CompressionMetadata open(File chunksIndexFile, long compressedLength, boolean hasMaxCompressedSize)
+    public static CompressionMetadata open(File chunksIndexFile,
+                                           long compressedLength,
+                                           boolean hasMaxCompressedSize)
+    {
+        return open(chunksIndexFile, compressedLength, hasMaxCompressedSize, null);
+    }
+
+    @VisibleForTesting
+    public static CompressionMetadata open(File chunksIndexFile,
+                                           long compressedLength,
+                                           boolean hasMaxCompressedSize,
+                                           @Nullable CompressionDictionaryManager compressionDictionaryManager)
     {
         CompressionParams parameters;
         long dataLength;
         Memory chunkOffsets;
+        CompressionDictionary compressionDictionary;
 
         try (FileInputStreamPlus stream = chunksIndexFile.newInputStream())
         {
@@ -99,6 +117,7 @@ public class CompressionMetadata extends WrappedSharedCloseable
 
             dataLength = stream.readLong();
             chunkOffsets = readChunkOffsets(stream);
+            compressionDictionary = CompressionDictionary.deserialize(stream, compressionDictionaryManager);
         }
         catch (FileNotFoundException | NoSuchFileException e)
         {
@@ -109,7 +128,9 @@ public class CompressionMetadata extends WrappedSharedCloseable
             throw new CorruptSSTableException(e, chunksIndexFile);
         }
 
-        return new CompressionMetadata(chunksIndexFile, parameters, chunkOffsets, chunkOffsets.size(), dataLength, compressedLength);
+        return new CompressionMetadata(chunksIndexFile, parameters,
+                                       chunkOffsets, chunkOffsets.size(), dataLength,
+                                       compressedLength, compressionDictionary);
     }
 
     // do not call this constructor directly, unless used in testing
@@ -119,7 +140,8 @@ public class CompressionMetadata extends WrappedSharedCloseable
                                Memory chunkOffsets,
                                long chunkOffsetsSize,
                                long dataLength,
-                               long compressedFileLength)
+                               long compressedFileLength,
+                               CompressionDictionary compressionDictionary)
     {
         super(chunkOffsets);
         this.chunksIndexFile = chunksIndexFile;
@@ -128,6 +150,7 @@ public class CompressionMetadata extends WrappedSharedCloseable
         this.compressedFileLength = compressedFileLength;
         this.chunkOffsets = chunkOffsets;
         this.chunkOffsetsSize = chunkOffsetsSize;
+        this.compressionDictionary = compressionDictionary;
     }
 
     private CompressionMetadata(CompressionMetadata copy)
@@ -139,11 +162,46 @@ public class CompressionMetadata extends WrappedSharedCloseable
         this.compressedFileLength = copy.compressedFileLength;
         this.chunkOffsets = copy.chunkOffsets;
         this.chunkOffsetsSize = copy.chunkOffsetsSize;
+        this.compressionDictionary = copy.compressionDictionary;
     }
 
     public ICompressor compressor()
     {
-        return parameters.getSstableCompressor();
+        // classic double-checked locking to call resolveCompressor method just once per CompressionMetadata object
+        ICompressor result = resolvedCompressor;
+        if (result != null)
+            return result;
+
+        synchronized (this)
+        {
+            result = resolvedCompressor;
+            if (result == null)
+            {
+                result = resolveCompressor(parameters.getSstableCompressor(), compressionDictionary);
+                resolvedCompressor = result;
+            }
+            return result;
+        }
+    }
+
+    static ICompressor resolveCompressor(ICompressor compressor, CompressionDictionary dictionary)
+    {
+        if (dictionary == null)
+            return compressor;
+
+        // When the attached dictionary can be consumed by the current dictionary compressor
+        if (compressor instanceof IDictionaryCompressor)
+        {
+            IDictionaryCompressor dictionaryCompressor = (IDictionaryCompressor) compressor;
+            if (dictionaryCompressor.canConsumeDictionary(dictionary))
+                return dictionaryCompressor.getOrCopyWithDictionary(dictionary);
+        }
+
+        // When the current compressor is not compatible with the dictionary. It could happen in the read path when:
+        // 1. The current compressor is not a dictionary compressor, but there is dictionary attached
+        // 2. The current dictionary compressor is a different type, e.g. table schema is changed
+        // In those cases, we should get the compatible dictionary compressor based on the dictionary
+        return dictionary.kind().createCompressor(dictionary);
     }
 
     public int chunkLength()
@@ -349,16 +407,21 @@ public class CompressionMetadata extends WrappedSharedCloseable
 
         // provided by user when setDescriptor
         private long dataLength, chunkCount;
+        @Nullable
+        private CompressionDictionary compressionDictionary;
 
-        private Writer(CompressionParams parameters, File file)
+        private Writer(CompressionParams parameters, File file, CompressionDictionary compressionDictionary)
         {
             this.parameters = parameters;
             this.file = file;
+            this.compressionDictionary = compressionDictionary;
         }
 
-        public static Writer open(CompressionParams parameters, File file)
+        public static Writer open(CompressionParams parameters,
+                                  File file,
+                                  CompressionDictionary compressionDictionary)
         {
-            return new Writer(parameters, file);
+            return new Writer(parameters, file, compressionDictionary);
         }
 
         public void addOffset(long offset)
@@ -397,6 +460,21 @@ public class CompressionMetadata extends WrappedSharedCloseable
             }
         }
 
+        private void writeCompressionDictionary(DataOutput out)
+        {
+            if (compressionDictionary == null)
+                return;
+
+            try
+            {
+                compressionDictionary.serialize(out);
+            }
+            catch (IOException e)
+            {
+                throw new FSWriteError(e, file);
+            }
+        }
+
         // we've written everything; wire up some final metadata state
         public Writer finalizeLength(long dataLength, int chunkCount)
         {
@@ -426,6 +504,7 @@ public class CompressionMetadata extends WrappedSharedCloseable
                 for (int i = 0; i < count; i++)
                     out.writeLong(offsets.getLong(i * 8L));
 
+                writeCompressionDictionary(out);
                 out.flush();
                 out.sync();
             }
@@ -453,7 +532,9 @@ public class CompressionMetadata extends WrappedSharedCloseable
             if (tCount < this.count)
                 compressedLength = tOffsets.getLong(tCount * 8L);
 
-            return new CompressionMetadata(file, parameters, tOffsets, tCount * 8L, dataLength, compressedLength);
+            return new CompressionMetadata(file, parameters,
+                                           tOffsets, tCount * 8L, dataLength,
+                                           compressedLength, compressionDictionary);
         }
 
         /**

src/java/org/apache/cassandra/io/compress/ICompressor.java

@@ -37,6 +37,11 @@ public interface ICompressor
         FAST_COMPRESSION
     }
 
+    /**
+     * Get the maximum compressed size in the worst case scenario
+     * @param chunkLength input data (chunk) size
+     * @return compressed size upper bound in the worse case
+     */
     public int initialCompressedBufferLength(int chunkLength);
 
     public int uncompress(byte[] input, int inputOffset, int inputLength, byte[] output, int outputOffset) throws IOException;

src/java/org/apache/cassandra/io/compress/IDictionaryCompressor.java

@@ -0,0 +1,71 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.io.compress;
+
+import org.apache.cassandra.db.compression.CompressionDictionary;
+
+/**
+ * Interface for compressors that support dictionary-based compression.
+ * <br>
+ * Dictionary compressors can use pre-trained compression dictionaries to achieve
+ * better compression ratios, especially for small data chunks that are similar
+ * to the training data used to create the dictionary.
+ * 
+ * @param <T> the specific type of compression dictionary this compressor supports
+ */
+public interface IDictionaryCompressor<T extends CompressionDictionary>
+{
+    /**
+     * Returns a compressor instance configured with the specified compression dictionary.
+     * <br>
+     * This method may return the same instance if it already uses the given dictionary,
+     * or create a new instance configured with the dictionary. The implementation should
+     * be efficient and avoid unnecessary object creation when possible.
+     * 
+     * @param compressionDictionary the dictionary to use for compression/decompression
+     * @return a compressor instance that will use the specified dictionary
+     */
+    ICompressor getOrCopyWithDictionary(T compressionDictionary);
+
+    /**
+     * Returns the kind of compression dictionary that this compressor can accept.
+     * <br>
+     * This is used to validate dictionary compatibility before attempting to use
+     * a dictionary with this compressor. Only dictionaries of the returned kind
+     * should be passed to {@link #getOrCopyWithDictionary(CompressionDictionary)}.
+     * 
+     * @return the compression dictionary kind supported by this compressor
+     */
+    CompressionDictionary.Kind acceptableDictionaryKind();
+
+    /**
+     * Checks whether this compressor can use the given compression dictionary.
+     * <br>
+     * The default implementation compares the dictionary's kind with the kind
+     * returned by {@link #acceptableDictionaryKind()}. Compressor implementations
+     * may override this method to provide more sophisticated compatibility checks.
+     * 
+     * @param dictionary the compression dictionary to check for compatibility
+     * @return true if this compressor can use the dictionary, false otherwise
+     */
+    default boolean canConsumeDictionary(CompressionDictionary dictionary)
+    {
+        return dictionary.kind() == acceptableDictionaryKind();
+    }
+}

src/java/org/apache/cassandra/io/compress/ZstdCompressor.java

@@ -18,72 +18,31 @@
 
 package org.apache.cassandra.io.compress;
 
-import java.io.IOException;
-import java.nio.ByteBuffer;
 import java.util.Collections;
-import java.util.HashSet;
 import java.util.Map;
-import java.util.Set;
 import java.util.concurrent.ConcurrentHashMap;
 
-import com.google.common.annotations.VisibleForTesting;
-import com.google.common.collect.ImmutableSet;
-import org.slf4j.Logger;
-import org.slf4j.LoggerFactory;
-
-import com.github.luben.zstd.Zstd;
-
 /**
  * ZSTD Compressor
  */
-public class ZstdCompressor implements ICompressor
+public class ZstdCompressor extends ZstdCompressorBase implements ICompressor
 {
-    private static final Logger logger = LoggerFactory.getLogger(ZstdCompressor.class);
-
-    // These might change with the version of Zstd we're using
-    public static final int FAST_COMPRESSION_LEVEL = Zstd.minCompressionLevel();
-    public static final int BEST_COMPRESSION_LEVEL = Zstd.maxCompressionLevel();
-
-    // Compressor Defaults
-    public static final int DEFAULT_COMPRESSION_LEVEL = 3;
-    private static final boolean ENABLE_CHECKSUM_FLAG = true;
-
-    @VisibleForTesting
-    public static final String COMPRESSION_LEVEL_OPTION_NAME = "compression_level";
-
     private static final ConcurrentHashMap<Integer, ZstdCompressor> instances = new ConcurrentHashMap<>();
 
-    private final int compressionLevel;
-    private final Set<Uses> recommendedUses;
-
     /**
      * Create a Zstd compressor with the given options
+     * Invoked by {@link org.apache.cassandra.schema.CompressionParams#createCompressor} via reflection
      *
-     * @param options
-     * @return
+     * @param options compression options
+     * @return ZstdCompressor
      */
     public static ZstdCompressor create(Map<String, String> options)
     {
         int level = getOrDefaultCompressionLevel(options);
-
-        if (!isValid(level))
-            throw new IllegalArgumentException(String.format("%s=%d is invalid", COMPRESSION_LEVEL_OPTION_NAME, level));
-
+        validateCompressionLevel(level);
         return getOrCreate(level);
     }
 
-    /**
-     * Private constructor
-     *
-     * @param compressionLevel
-     */
-    private ZstdCompressor(int compressionLevel)
-    {
-        this.compressionLevel = compressionLevel;
-        this.recommendedUses = ImmutableSet.of(Uses.GENERAL);
-        logger.trace("Creating Zstd Compressor with compression level={}", compressionLevel);
-    }
-
     /**
      * Get a cached instance or return a new one
      *
@@ -92,157 +51,16 @@ public class ZstdCompressor implements ICompressor
      */
     public static ZstdCompressor getOrCreate(int level)
     {
-        return instances.computeIfAbsent(level, l -> new ZstdCompressor(level));
+        return instances.computeIfAbsent(level, ZstdCompressor::new);
     }
 
     /**
-     * Get initial compressed buffer length
+     * Private constructor
      *
-     * @param chunkLength
-     * @return
+     * @param compressionLevel
      */
-    @Override
-    public int initialCompressedBufferLength(int chunkLength)
+    private ZstdCompressor(int compressionLevel)
     {
-        return (int) Zstd.compressBound(chunkLength);
-    }
-
-    /**
-     * Decompress data using arrays
-     *
-     * @param input
-     * @param inputOffset
-     * @param inputLength
-     * @param output
-     * @param outputOffset
-     * @return
-     * @throws IOException
-     */
-    @Override
-    public int uncompress(byte[] input, int inputOffset, int inputLength, byte[] output, int outputOffset)
-    throws IOException
-    {
-        long dsz = Zstd.decompressByteArray(output, outputOffset, output.length - outputOffset,
-                                            input, inputOffset, inputLength);
-
-        if (Zstd.isError(dsz))
-            throw new IOException(String.format("Decompression failed due to %s", Zstd.getErrorName(dsz)));
-
-        return (int) dsz;
-    }
-
-    /**
-     * Decompress data via ByteBuffers
-     *
-     * @param input
-     * @param output
-     * @throws IOException
-     */
-    @Override
-    public void uncompress(ByteBuffer input, ByteBuffer output) throws IOException
-    {
-        try
-        {
-            Zstd.decompress(output, input);
-        } catch (Exception e)
-        {
-            throw new IOException("Decompression failed", e);
-        }
-    }
-
-    /**
-     * Compress using ByteBuffers
-     *
-     * @param input
-     * @param output
-     * @throws IOException
-     */
-    @Override
-    public void compress(ByteBuffer input, ByteBuffer output) throws IOException
-    {
-        try
-        {
-            Zstd.compress(output, input, compressionLevel, ENABLE_CHECKSUM_FLAG);
-        } catch (Exception e)
-        {
-            throw new IOException("Compression failed", e);
-        }
-    }
-
-    /**
-     * Check if the given compression level is valid. This can be a negative value as well.
-     *
-     * @param level
-     * @return
-     */
-    private static boolean isValid(int level)
-    {
-        return (level >= FAST_COMPRESSION_LEVEL && level <= BEST_COMPRESSION_LEVEL);
-    }
-
-    /**
-     * Parse the compression options
-     *
-     * @param options
-     * @return
-     */
-    private static int getOrDefaultCompressionLevel(Map<String, String> options)
-    {
-        if (options == null)
-            return DEFAULT_COMPRESSION_LEVEL;
-
-        String val = options.get(COMPRESSION_LEVEL_OPTION_NAME);
-
-        if (val == null)
-            return DEFAULT_COMPRESSION_LEVEL;
-
-        return Integer.valueOf(val);
-    }
-
-    /**
-     * Return the preferred BufferType
-     *
-     * @return
-     */
-    @Override
-    public BufferType preferredBufferType()
-    {
-        return BufferType.OFF_HEAP;
-    }
-
-    /**
-     * Check whether the given BufferType is supported
-     *
-     * @param bufferType
-     * @return
-     */
-    @Override
-    public boolean supports(BufferType bufferType)
-    {
-        return bufferType == BufferType.OFF_HEAP;
-    }
-
-    /**
-     * Lists the supported options by this compressor
-     *
-     * @return
-     */
-    @Override
-    public Set<String> supportedOptions()
-    {
-        return new HashSet<>(Collections.singletonList(COMPRESSION_LEVEL_OPTION_NAME));
-    }
-
-
-    @VisibleForTesting
-    public int getCompressionLevel()
-    {
-        return compressionLevel;
-    }
-
-    @Override
-    public Set<Uses> recommendedUses()
-    {
-        return recommendedUses;
+        super(compressionLevel, Collections.singleton(COMPRESSION_LEVEL_OPTION_NAME));
     }
 }

src/java/org/apache/cassandra/io/compress/ZstdCompressorBase.java

@@ -0,0 +1,197 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.io.compress;
+
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.util.Collections;
+import java.util.Map;
+import java.util.Set;
+
+import com.google.common.annotations.VisibleForTesting;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import com.github.luben.zstd.Zstd;
+
+public abstract class ZstdCompressorBase implements ICompressor
+{
+    // These might change with the version of Zstd we're using
+    public static final int FAST_COMPRESSION_LEVEL = Zstd.minCompressionLevel();
+    public static final int BEST_COMPRESSION_LEVEL = Zstd.maxCompressionLevel();
+
+    // Compressor Defaults
+    public static final int DEFAULT_COMPRESSION_LEVEL = 3;
+    public static final boolean ENABLE_CHECKSUM_FLAG = true;
+
+    // Compressor option names
+    public static final String COMPRESSION_LEVEL_OPTION_NAME = "compression_level";
+
+    protected final Logger logger = LoggerFactory.getLogger(getClass());
+
+    private final int compressionLevel;
+    private final Set<ICompressor.Uses> recommendedUses;
+    private final Set<String> supportedOptions;
+
+    protected ZstdCompressorBase(int compressionLevel, Set<String> supportedOptions)
+    {
+        this.compressionLevel = compressionLevel;
+        this.supportedOptions = Collections.unmodifiableSet(supportedOptions);
+        this.recommendedUses = Set.of(ICompressor.Uses.GENERAL);
+        logger.trace("Creating Zstd Compressor with compression level={}", compressionLevel);
+    }
+
+    @Override
+    public int initialCompressedBufferLength(int chunkLength)
+    {
+        return (int) Zstd.compressBound(chunkLength);
+    }
+
+    @Override
+    public BufferType preferredBufferType()
+    {
+        return BufferType.OFF_HEAP;
+    }
+
+    @Override
+    public boolean supports(BufferType bufferType)
+    {
+        return bufferType == BufferType.OFF_HEAP;
+    }
+
+    @Override
+    public Set<Uses> recommendedUses()
+    {
+        return recommendedUses;
+    }
+
+    @VisibleForTesting
+    public int compressionLevel()
+    {
+        return compressionLevel;
+    }
+
+    @Override
+    public Set<String> supportedOptions()
+    {
+        return supportedOptions;
+    }
+
+    /**
+     * Decompress data using arrays
+     *
+     * @param input
+     * @param inputOffset
+     * @param inputLength
+     * @param output
+     * @param outputOffset
+     * @return
+     * @throws IOException
+     */
+    @Override
+    public int uncompress(byte[] input, int inputOffset, int inputLength, byte[] output, int outputOffset)
+    throws IOException
+    {
+        long dsz;
+        try
+        {
+            dsz = Zstd.decompressByteArray(output, outputOffset, output.length - outputOffset,
+                                           input, inputOffset, inputLength);
+        }
+        catch (Exception e)
+        {
+            throw new IOException("Decompression failed", e);
+        }
+
+        if (Zstd.isError(dsz))
+            throw new IOException("Decompression failed due to " + Zstd.getErrorName(dsz));
+
+        return (int) dsz;
+    }
+
+    /**
+     * Decompress data via ByteBuffers
+     *
+     * @param input
+     * @param output
+     * @throws IOException
+     */
+    @Override
+    public void uncompress(ByteBuffer input, ByteBuffer output) throws IOException
+    {
+        try
+        {
+            Zstd.decompress(output, input);
+        } catch (Exception e)
+        {
+            throw new IOException("Decompression failed", e);
+        }
+    }
+
+    /**
+     * Compress using ByteBuffers
+     *
+     * @param input
+     * @param output
+     * @throws IOException
+     */
+    @Override
+    public void compress(ByteBuffer input, ByteBuffer output) throws IOException
+    {
+        try
+        {
+            Zstd.compress(output, input, compressionLevel(), ENABLE_CHECKSUM_FLAG);
+        } catch (Exception e)
+        {
+            throw new IOException("Compression failed", e);
+        }
+    }
+
+    /**
+     * Check if the given compression level is valid. This can be a negative value as well.
+     *
+     * @param level compression level
+     */
+    public static void validateCompressionLevel(int level)
+    {
+        if (level < FAST_COMPRESSION_LEVEL || level > BEST_COMPRESSION_LEVEL)
+        {
+            throw new IllegalArgumentException(String.format("%s=%d is invalid", COMPRESSION_LEVEL_OPTION_NAME, level));
+        }
+    }
+
+    /**
+     * Get the supplied compression level; otherwise, use the default
+     *
+     * @param options compression options
+     * @return compression level
+     */
+    public static int getOrDefaultCompressionLevel(Map<String, String> options)
+    {
+        if (options == null)
+            return DEFAULT_COMPRESSION_LEVEL;
+
+        String val = options.get(COMPRESSION_LEVEL_OPTION_NAME);
+
+        if (val == null)
+            return DEFAULT_COMPRESSION_LEVEL;
+
+        return Integer.parseInt(val);
+    }
+}

src/java/org/apache/cassandra/io/compress/ZstdDictionaryCompressor.java

@@ -0,0 +1,216 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.io.compress;
+
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.time.Duration;
+import java.util.Map;
+import java.util.Set;
+import java.util.concurrent.ConcurrentHashMap;
+
+import com.github.benmanes.caffeine.cache.Cache;
+import com.github.benmanes.caffeine.cache.Caffeine;
+import com.github.benmanes.caffeine.cache.RemovalCause;
+import com.github.luben.zstd.Zstd;
+
+import com.google.common.annotations.VisibleForTesting;
+
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compression.ZstdCompressionDictionary;
+import org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import org.apache.cassandra.utils.concurrent.Ref;
+
+import javax.annotation.Nullable;
+
+public class ZstdDictionaryCompressor extends ZstdCompressorBase implements ICompressor, IDictionaryCompressor<ZstdCompressionDictionary>
+{
+    private static final ConcurrentHashMap<Integer, ZstdDictionaryCompressor> instancesPerLevel = new ConcurrentHashMap<>();
+    private static final Cache<ZstdCompressionDictionary, ZstdDictionaryCompressor> instancePerDict =
+    Caffeine.newBuilder()
+            .maximumSize(DatabaseDescriptor.getCompressionDictionaryCacheSize())
+            .expireAfterAccess(Duration.ofSeconds(DatabaseDescriptor.getCompressionDictionaryCacheExpireSeconds()))
+            .removalListener((ZstdCompressionDictionary dictionary,
+                              ZstdDictionaryCompressor compressor,
+                              RemovalCause cause) -> {
+                // Release dictionary reference when compressor is evicted from cache
+                if (compressor != null && compressor.dictionaryRef != null)
+                {
+                    compressor.dictionaryRef.release();
+                }
+            })
+            .build();
+
+    // dictionary and its ref are null, when they are absent.
+    // In this case, the compressor falls back to be the same as ZstdCompressor
+    @Nullable
+    private final ZstdCompressionDictionary dictionary;
+    @Nullable
+    private final Ref<ZstdCompressionDictionary> dictionaryRef;
+
+    /**
+     * Create a ZstdDictionaryCompressor with the given options
+     * Invoked by {@link org.apache.cassandra.schema.CompressionParams#createCompressor} via reflection
+     *
+     * @param options compression options
+     * @return ZstdDictionaryCompressor
+     */
+    public static ZstdDictionaryCompressor create(Map<String, String> options)
+    {
+        int level = getOrDefaultCompressionLevel(options);
+        validateCompressionLevel(level);
+        return getOrCreate(level, null);
+    }
+
+    // Constructor used to create the compressor for reading the sstable; the compression level is not relevant
+    public static ZstdDictionaryCompressor create(ZstdCompressionDictionary dictionary)
+    {
+        return getOrCreate(DEFAULT_COMPRESSION_LEVEL, dictionary);
+    }
+
+    private static ZstdDictionaryCompressor getOrCreate(int level, ZstdCompressionDictionary dictionary)
+    {
+        if (dictionary == null)
+        {
+            return instancesPerLevel.computeIfAbsent(level, ZstdDictionaryCompressor::new);
+        }
+
+        return instancePerDict.get(dictionary, dict -> {
+            // Get a reference to the dictionary when creating new compressor
+            Ref<ZstdCompressionDictionary> ref = dict != null ? dict.tryRef() : null;
+            if (ref == null && dict != null)
+            {
+                // Dictionary is being closed, cannot create compressor
+                throw new IllegalStateException("Dictionary is being closed");
+            }
+            return new ZstdDictionaryCompressor(level, dictionary, ref);
+        });
+    }
+
+    private ZstdDictionaryCompressor(int level)
+    {
+        this(level, null, null);
+    }
+
+    private ZstdDictionaryCompressor(int level, ZstdCompressionDictionary dictionary, Ref<ZstdCompressionDictionary> dictionaryRef)
+    {
+        super(level, Set.of(COMPRESSION_LEVEL_OPTION_NAME));
+        this.dictionary = dictionary;
+        this.dictionaryRef = dictionaryRef;
+    }
+
+    @Override
+    public ZstdDictionaryCompressor getOrCopyWithDictionary(ZstdCompressionDictionary compressionDictionary)
+    {
+        return getOrCreate(compressionLevel(), compressionDictionary);
+    }
+
+    @Override
+    public Kind acceptableDictionaryKind()
+    {
+        return Kind.ZSTD;
+    }
+
+    @Override
+    public int uncompress(byte[] input, int inputOffset, int inputLength, byte[] output, int outputOffset) throws IOException
+    {
+        // fallback to non-dict zstd compressor
+        if (dictionary == null)
+        {
+            return super.uncompress(input, inputOffset, inputLength, output, outputOffset);
+        }
+
+        int dsz;
+        try
+        {
+            dsz = (int) Zstd.decompressFastDict(output, outputOffset,
+                                                input, inputOffset, inputLength,
+                                                dictionary.dictionaryForDecompression());
+        }
+        catch (Exception e)
+        {
+            throw new IOException("Decompression failed", e);
+        }
+
+        if (Zstd.isError(dsz))
+            throw new IOException("Decompression failed due to " + Zstd.getErrorName(dsz));
+
+        return dsz;
+    }
+
+    @Override
+    public void uncompress(ByteBuffer input, ByteBuffer output) throws IOException
+    {
+        if (dictionary == null)
+        {
+            super.uncompress(input, output);
+            return;
+        }
+
+        try
+        {
+            // Zstd compressors expect only direct bytebuffer. See ZstdCompressorBase.preferredBufferType and supports
+            int decompressedSize = (int) Zstd.decompressDirectByteBufferFastDict(output, output.position(), output.limit() - output.position(),
+                                                                                 input, input.position(), input.limit() - input.position(),
+                                                                                 dictionary.dictionaryForDecompression());
+            output.position(output.position() + decompressedSize);
+            input.position(input.limit());
+        }
+        catch (Exception e)
+        {
+            throw new IOException("Decompression failed", e);
+        }
+    }
+
+    @Override
+    public void compress(ByteBuffer input, ByteBuffer output) throws IOException
+    {
+        if (dictionary == null)
+        {
+            super.compress(input, output);
+            return;
+        }
+
+        try
+        {
+            // Zstd compressors expect only direct bytebuffer. See ZstdCompressorBase.preferredBufferType and supports
+            int compressedSize = (int) Zstd.compressDirectByteBufferFastDict(output, output.position(), output.limit() - output.position(),
+                                                                             input, input.position(), input.limit() - input.position(),
+                                                                             dictionary.dictionaryForCompression(compressionLevel()));
+            output.position(output.position() + compressedSize);
+            input.position(input.limit());
+        }
+        catch (Exception e)
+        {
+            throw new IOException("Compression failed", e);
+        }
+    }
+
+    @VisibleForTesting
+    ZstdCompressionDictionary dictionary()
+    {
+        return dictionary;
+    }
+
+    @VisibleForTesting
+    public static void invalidateCache()
+    {
+        instancePerDict.invalidateAll();
+    }
+}

src/java/org/apache/cassandra/io/sstable/SSTable.java

@@ -43,6 +43,7 @@ import org.apache.cassandra.db.lifecycle.Tracker;
 import org.apache.cassandra.dht.AbstractBounds;
 import org.apache.cassandra.dht.IPartitioner;
 import org.apache.cassandra.dht.Token;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.io.sstable.format.SSTableFormat;
 import org.apache.cassandra.io.sstable.format.SSTableFormat.Components;
 import org.apache.cassandra.io.sstable.format.TOCComponent;
@@ -369,6 +370,8 @@ public abstract class SSTable
         OpOrder.Barrier newReadOrderingBarrier();
 
         TableMetrics getMetrics();
+
+        CompressionDictionaryManager compressionDictionaryManager();
     }
 
     /**

src/java/org/apache/cassandra/io/sstable/SimpleSSTableMultiWriter.java

@@ -23,6 +23,7 @@ import java.util.Collections;
 import org.apache.cassandra.db.SerializationHeader;
 import org.apache.cassandra.db.commitlog.CommitLogPosition;
 import org.apache.cassandra.db.commitlog.IntervalSet;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.db.lifecycle.ILifecycleTransaction;
 import org.apache.cassandra.db.rows.UnfilteredRowIterator;
 import org.apache.cassandra.index.Index;
@@ -122,17 +123,21 @@ public class SimpleSSTableMultiWriter implements SSTableMultiWriter
         MetadataCollector metadataCollector = new MetadataCollector(metadata.get().comparator)
                                               .commitLogIntervals(commitLogPositions != null ? commitLogPositions : IntervalSet.empty())
                                               .sstableLevel(sstableLevel);
-        SSTableWriter writer = descriptor.getFormat().getWriterFactory().builder(descriptor)
-                                            .setKeyCount(keyCount)
-                                            .setRepairedAt(repairedAt)
-                                            .setPendingRepair(pendingRepair)
-                                            .setTransientSSTable(isTransient)
-                                            .setTableMetadataRef(metadata)
-                                            .setMetadataCollector(metadataCollector)
-                                            .setSerializationHeader(header)
-                                            .addDefaultComponents(indexGroups)
-                                            .setSecondaryIndexGroups(indexGroups)
-                                            .build(txn, owner);
+        CompressionDictionaryManager compressionDictionaryManager = owner == null ? null : owner.compressionDictionaryManager();
+        SSTableWriter writer = descriptor.getFormat()
+                                         .getWriterFactory()
+                                         .builder(descriptor)
+                                         .setKeyCount(keyCount)
+                                         .setRepairedAt(repairedAt)
+                                         .setPendingRepair(pendingRepair)
+                                         .setTransientSSTable(isTransient)
+                                         .setTableMetadataRef(metadata)
+                                         .setMetadataCollector(metadataCollector)
+                                         .setSerializationHeader(header)
+                                         .addDefaultComponents(indexGroups)
+                                         .setSecondaryIndexGroups(indexGroups)
+                                         .setCompressionDictionaryManager(compressionDictionaryManager)
+                                         .build(txn, owner);
         return new SimpleSSTableMultiWriter(writer, txn);
     }
 }

src/java/org/apache/cassandra/io/sstable/format/CompressionInfoComponent.java

@@ -22,6 +22,9 @@ import java.io.IOException;
 import java.nio.file.NoSuchFileException;
 import java.util.Set;
 
+import javax.annotation.Nullable;
+
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.io.FSReadError;
 import org.apache.cassandra.io.compress.CompressionMetadata;
 import org.apache.cassandra.io.sstable.Component;
@@ -32,27 +35,31 @@ import org.apache.cassandra.io.util.File;
 
 public class CompressionInfoComponent
 {
-    public static CompressionMetadata maybeLoad(Descriptor descriptor, Set<Component> components)
+    public static CompressionMetadata maybeLoad(Descriptor descriptor, Set<Component> components,
+                                                @Nullable CompressionDictionaryManager compressionDictionaryManager)
     {
         if (components.contains(Components.COMPRESSION_INFO))
-            return load(descriptor);
+            return load(descriptor, compressionDictionaryManager);
 
         return null;
     }
 
-    public static CompressionMetadata loadIfExists(Descriptor descriptor)
+    public static CompressionMetadata loadIfExists(Descriptor descriptor,
+                                                   @Nullable CompressionDictionaryManager compressionDictionaryManager)
     {
         if (descriptor.fileFor(Components.COMPRESSION_INFO).exists())
-            return load(descriptor);
+            return load(descriptor, compressionDictionaryManager);
 
         return null;
     }
 
-    public static CompressionMetadata load(Descriptor descriptor)
+    public static CompressionMetadata load(Descriptor descriptor,
+                                           @Nullable CompressionDictionaryManager compressionDictionaryManager)
     {
         return CompressionMetadata.open(descriptor.fileFor(Components.COMPRESSION_INFO),
                                         descriptor.fileFor(Components.DATA).length(),
-                                        descriptor.version.hasMaxCompressedLength());
+                                        descriptor.version.hasMaxCompressedLength(),
+                                        compressionDictionaryManager);
     }
 
     /**

src/java/org/apache/cassandra/io/sstable/format/DataComponent.java

@@ -20,6 +20,7 @@ package org.apache.cassandra.io.sstable.format;
 
 import org.apache.cassandra.config.Config.FlushCompression;
 import org.apache.cassandra.db.compaction.OperationType;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.io.compress.CompressedSequentialWriter;
 import org.apache.cassandra.io.compress.ICompressor;
 import org.apache.cassandra.io.sstable.Descriptor;
@@ -38,7 +39,8 @@ public class DataComponent
                                                SequentialWriterOption options,
                                                MetadataCollector metadataCollector,
                                                OperationType operationType,
-                                               FlushCompression flushCompression)
+                                               FlushCompression flushCompression,
+                                               CompressionDictionaryManager compressionDictionaryManager)
     {
         if (metadata.params.compression.isEnabled())
         {
@@ -49,7 +51,8 @@ public class DataComponent
                                                   descriptor.fileFor(Components.DIGEST),
                                                   options,
                                                   compressionParams,
-                                                  metadataCollector);
+                                                  metadataCollector,
+                                                  compressionDictionaryManager);
         }
         else
         {

src/java/org/apache/cassandra/io/sstable/format/SSTableWriter.java

@@ -28,6 +28,7 @@ import java.util.Map;
 import java.util.Set;
 import java.util.function.Consumer;
 import java.util.function.Supplier;
+import javax.annotation.Nullable;
 
 import com.google.common.collect.ImmutableList;
 import com.google.common.collect.ImmutableSet;
@@ -37,6 +38,7 @@ import org.slf4j.LoggerFactory;
 
 import org.apache.cassandra.db.DecoratedKey;
 import org.apache.cassandra.db.SerializationHeader;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.db.lifecycle.ILifecycleTransaction;
 import org.apache.cassandra.db.rows.UnfilteredRowIterator;
 import org.apache.cassandra.dht.AbstractBounds;
@@ -443,6 +445,8 @@ public abstract class SSTableWriter extends SSTable implements Transactional
         private boolean transientSSTable;
         private SerializationHeader serializationHeader;
         private List<Index.Group> indexGroups;
+        @Nullable
+        private CompressionDictionaryManager compressionDictionaryManager;
 
         public B setMetadataCollector(MetadataCollector metadataCollector)
         {
@@ -521,6 +525,18 @@ public abstract class SSTableWriter extends SSTable implements Transactional
             return (B) this;
         }
 
+        public B setCompressionDictionaryManager(CompressionDictionaryManager compressionDictionaryManager)
+        {
+            this.compressionDictionaryManager = compressionDictionaryManager;
+            return (B) this;
+        }
+
+        @Nullable
+        public CompressionDictionaryManager getCompressionDictionaryManager()
+        {
+            return compressionDictionaryManager;
+        }
+
         public MetadataCollector getMetadataCollector()
         {
             return metadataCollector;

src/java/org/apache/cassandra/io/sstable/format/big/BigFormat.java

@@ -433,7 +433,8 @@ public class BigFormat extends AbstractSSTableFormat<BigTableReader, BigTableWri
 
     static class BigVersion extends Version
     {
-        public static final String current_version = DatabaseDescriptor.getStorageCompatibilityMode().isBefore(5) ? "nb" : "oa";
+        public static final String current_version = DatabaseDescriptor.getStorageCompatibilityMode().isBefore(5) ? "nb" :
+                                                     DatabaseDescriptor.getStorageCompatibilityMode().isBefore(6) ? "oa" : "pa";
         public static final String earliest_supported_version = "ma";
 
         // ma (3.0.0): swap bf hash order
@@ -448,6 +449,7 @@ public class BigFormat extends AbstractSSTableFormat<BigTableReader, BigTableWri
         // oa (5.0): improved min/max, partition level deletion presence marker, key range (CASSANDRA-18134)
         //           Long deletionTime to prevent TTL overflow
         //           token space coverage
+        // pa (6.0): compression dictionary metadata in CompressionInfo component
         //
         // NOTE: When adding a new version:
         //  - Please add it to LegacySSTableTest

src/java/org/apache/cassandra/io/sstable/format/big/BigSSTableReaderLoadingBuilder.java

@@ -26,6 +26,7 @@ import org.slf4j.LoggerFactory;
 
 import org.apache.cassandra.db.DecoratedKey;
 import org.apache.cassandra.db.SerializationHeader;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.io.compress.CompressionMetadata;
 import org.apache.cassandra.io.sstable.Downsampling;
 import org.apache.cassandra.io.sstable.KeyReader;
@@ -137,7 +138,8 @@ public class BigSSTableReaderLoadingBuilder extends SortedTableReaderLoadingBuil
                 }
             }
 
-            try (CompressionMetadata compressionMetadata = CompressionInfoComponent.maybeLoad(descriptor, components))
+            CompressionDictionaryManager compressionDictionaryManager = owner == null ? null : owner.compressionDictionaryManager();
+            try (CompressionMetadata compressionMetadata = CompressionInfoComponent.maybeLoad(descriptor, components, compressionDictionaryManager))
             {
                 builder.setDataFile(dataFileBuilder(builder.getStatsMetadata())
                                     .withCompressionMetadata(compressionMetadata)

src/java/org/apache/cassandra/io/sstable/format/big/BigTableWriter.java

@@ -389,7 +389,8 @@ public class BigTableWriter extends SortedTableWriter<BigFormatPartitionWriter,
                                                                     getIOOptions().writerOptions,
                                                                     getMetadataCollector(),
                                                                     ensuringInBuildInternalContext(operationType),
-                                                                    getIOOptions().flushCompression);
+                                                                    getIOOptions().flushCompression,
+                                                                    getCompressionDictionaryManager());
             this.dataWriterOpened = true;
             return dataWriter;
         }

src/java/org/apache/cassandra/io/sstable/format/bti/BtiFormat.java

@@ -286,11 +286,12 @@ public class BtiFormat extends AbstractSSTableFormat<BtiTableReader, BtiTableWri
 
     static class BtiVersion extends Version
     {
-        public static final String current_version = "da";
+        public static final String current_version = "ea";
         public static final String earliest_supported_version = "da";
 
         // versions aa-cz are not supported in OSS
-        // da (5.0): initial version of the BIT format
+        // da (5.0): initial version of the BTI format
+        // ea (6.0): compression dictionary metadata in CompressionInfo component
         // NOTE: when adding a new version, please add that to LegacySSTableTest, too.
 
         private final boolean isLatestVersion;

src/java/org/apache/cassandra/io/sstable/format/bti/BtiTableReaderLoadingBuilder.java

@@ -23,7 +23,9 @@ import java.io.IOException;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import org.apache.cassandra.db.ColumnFamilyStore;
 import org.apache.cassandra.db.DecoratedKey;
+import org.apache.cassandra.db.compression.CompressionDictionaryManager;
 import org.apache.cassandra.dht.IPartitioner;
 import org.apache.cassandra.io.compress.CompressionMetadata;
 import org.apache.cassandra.io.sstable.KeyReader;
@@ -38,6 +40,7 @@ import org.apache.cassandra.io.sstable.metadata.StatsMetadata;
 import org.apache.cassandra.io.sstable.metadata.ValidationMetadata;
 import org.apache.cassandra.io.util.FileHandle;
 import org.apache.cassandra.metrics.TableMetrics;
+import org.apache.cassandra.schema.Schema;
 import org.apache.cassandra.utils.FilterFactory;
 import org.apache.cassandra.utils.IFilter;
 import org.apache.cassandra.utils.Throwables;
@@ -68,8 +71,15 @@ public class BtiTableReaderLoadingBuilder extends SortedTableReaderLoadingBuilde
     {
         checkNotNull(statsMetadata);
 
+        ColumnFamilyStore cfs = Schema.instance.getColumnFamilyStoreInstance(tableMetadataRef.id);
+        CompressionDictionaryManager compressionDictionaryManager = null;
+        if (cfs != null)
+        {
+            compressionDictionaryManager = cfs.compressionDictionaryManager();
+        }
+
         try (PartitionIndex index = PartitionIndex.load(partitionIndexFileBuilder(), tableMetadataRef.getLocal().partitioner, false);
-             CompressionMetadata compressionMetadata = CompressionInfoComponent.maybeLoad(descriptor, components);
+             CompressionMetadata compressionMetadata = CompressionInfoComponent.maybeLoad(descriptor, components, compressionDictionaryManager);
              FileHandle dFile = dataFileBuilder(statsMetadata).withCompressionMetadata(compressionMetadata)
                                                               .withCrcCheckChance(() -> tableMetadataRef.getLocal().params.crcCheckChance)
                                                               .complete();
@@ -131,7 +141,7 @@ public class BtiTableReaderLoadingBuilder extends SortedTableReaderLoadingBuilde
                 }
             }
 
-            try (CompressionMetadata compressionMetadata = CompressionInfoComponent.maybeLoad(descriptor, components))
+            try (CompressionMetadata compressionMetadata = CompressionInfoComponent.maybeLoad(descriptor, components, owner == null ? null : owner.compressionDictionaryManager()))
             {
                 builder.setDataFile(dataFileBuilder(builder.getStatsMetadata())
                                     .withCompressionMetadata(compressionMetadata)

src/java/org/apache/cassandra/io/sstable/format/bti/BtiTableWriter.java

@@ -334,7 +334,8 @@ public class BtiTableWriter extends SortedTableWriter<BtiFormatPartitionWriter,
                                              getIOOptions().writerOptions,
                                              getMetadataCollector(),
                                              ensuringInBuildInternalContext(operationType),
-                                             getIOOptions().flushCompression);
+                                             getIOOptions().flushCompression,
+                                             getCompressionDictionaryManager());
         }
 
         @Override

src/java/org/apache/cassandra/net/MessagingService.java

@@ -478,21 +478,28 @@ public class MessagingService extends MessagingServiceMBeanImpl implements Messa
     public <RSP> Future<RSP> sendWithResponse(InetAddressAndPort to, Message<?> msg)
     {
         Promise<RSP> future = AsyncPromise.uncancellable();
-        MessagingService.instance().sendWithCallback(msg, to,
-                                                     new RequestCallback<RSP>()
-                                                     {
-                                                         @Override
-                                                         public void onResponse(Message<RSP> msg)
-                                                         {
-                                                             future.setSuccess(msg.payload);
-                                                         }
+        RequestCallback<RSP> callback = new RequestCallback<RSP>()
+        {
+            @Override
+            public void onResponse(Message<RSP> msg)
+            {
+                future.setSuccess(msg.payload);
+            }
 
-                                                         @Override
-                                                         public void onFailure(InetAddressAndPort from, RequestFailure failure)
-                                                         {
-                                                             future.setFailure(new RuntimeException(failure.toString()));
-                                                         }
-                                                     });
+            @Override
+            public void onFailure(InetAddressAndPort from, RequestFailure failure)
+            {
+                future.setFailure(new RuntimeException(failure.toString()));
+            }
+        };
+        try
+        {
+            MessagingService.instance().sendWithCallback(msg, to, callback);
+        }
+        catch (Throwable e) // catch any exception during sending the message and wrap it inside feture to have unified exception handling
+        {
+            future.setFailure(e);
+        }
 
         return future;
     }

src/java/org/apache/cassandra/net/Verb.java

@@ -44,6 +44,8 @@ import org.apache.cassandra.db.SnapshotCommand;
 import org.apache.cassandra.db.TruncateRequest;
 import org.apache.cassandra.db.TruncateResponse;
 import org.apache.cassandra.db.TruncateVerbHandler;
+import org.apache.cassandra.db.compression.CompressionDictionaryUpdateMessage;
+import org.apache.cassandra.db.compression.CompressionDictionaryUpdateVerbHandler;
 import org.apache.cassandra.db.virtual.VirtualMutation;
 import org.apache.cassandra.exceptions.RequestFailure;
 import org.apache.cassandra.gms.GossipDigestAck;
@@ -375,6 +377,9 @@ public enum Verb
     ACCORD_FETCH_TOPOLOGY_RSP       (169, P0, shortTimeout, FETCH_METADATA,     () -> accordEmbedded(FetchTopologies.responseSerializer),        RESPONSE_HANDLER),
     ACCORD_FETCH_TOPOLOGY_REQ       (170, P0, shortTimeout, FETCH_METADATA,     () -> accordEmbedded(FetchTopologies.serializer),                () -> FetchTopologies.handler,           ACCORD_FETCH_TOPOLOGY_RSP),
 
+    DICTIONARY_UPDATE_RSP           (171, P1, rpcTimeout,   REQUEST_RESPONSE,  () -> NoPayload.serializer,                  RESPONSE_HANDLER                             ),
+    DICTIONARY_UPDATE_REQ           (172, P1, rpcTimeout, MISC, () -> CompressionDictionaryUpdateMessage.serializer, () -> CompressionDictionaryUpdateVerbHandler.instance, DICTIONARY_UPDATE_RSP ),
+
     // generic failure response
     FAILURE_RSP            (99,  P0, noTimeout,       REQUEST_RESPONSE,  () -> RequestFailure.serializer,            RESPONSE_HANDLER                             ),
 
@@ -682,4 +687,4 @@ class VerbTimeouts
 class ResponseHandlerSupplier
 {
     static final Supplier<IVerbHandler<?>> RESPONSE_HANDLER = () -> ResponseVerbHandler.instance;
-}
+}

src/java/org/apache/cassandra/schema/CompressionParams.java

@@ -160,15 +160,31 @@ public final class CompressionParams
         return new CompressionParams(LZ4Compressor.create(Collections.emptyMap()), chunkLength, maxCompressedLength, calcMinCompressRatio(chunkLength, maxCompressedLength), Collections.emptyMap());
     }
 
+    @VisibleForTesting
     public static CompressionParams zstd()
     {
-        return zstd(DEFAULT_CHUNK_LENGTH);
+        return zstd(DEFAULT_CHUNK_LENGTH, false);
     }
 
+    @VisibleForTesting
     public static CompressionParams zstd(Integer chunkLength)
     {
-        ZstdCompressor compressor = ZstdCompressor.create(Collections.emptyMap());
-        return new CompressionParams(compressor, chunkLength, Integer.MAX_VALUE, DEFAULT_MIN_COMPRESS_RATIO, Collections.emptyMap());
+        return zstd(chunkLength, false);
+    }
+
+    @VisibleForTesting
+    public static CompressionParams zstd(Integer chunkLength, boolean useDictionary)
+    {
+        return zstd(chunkLength, useDictionary, Collections.emptyMap());
+    }
+
+    @VisibleForTesting
+    public static CompressionParams zstd(Integer chunkLength, boolean useDictionary, Map<String, String> options)
+    {
+        ICompressor compressor = useDictionary
+                                 ? ZstdDictionaryCompressor.create(options)
+                                 : ZstdCompressor.create(options);
+        return new CompressionParams(compressor, chunkLength, Integer.MAX_VALUE, DEFAULT_MIN_COMPRESS_RATIO, options);
     }
 
     @VisibleForTesting
@@ -223,6 +239,18 @@ public final class CompressionParams
         return sstableCompressor != null;
     }
 
+    /**
+     * Checks if dictionary compression is enabled for this configuration.
+     * Dictionary compression is enabled when both compression is enabled and
+     * the compressor supports dictionary-based compression.
+     *
+     * @return {@code true} if dictionary compression is enabled, {@code false} otherwise.
+     */
+    public boolean isDictionaryCompressionEnabled()
+    {
+        return isEnabled() && sstableCompressor instanceof IDictionaryCompressor;
+    }
+
     /**
      * Returns the SSTable compressor.
      * @return the SSTable compressor or {@code null} if compression is disabled.

src/java/org/apache/cassandra/schema/SystemDistributedKeyspace.java

@@ -30,6 +30,8 @@ import java.util.Set;
 import java.util.UUID;
 import java.util.concurrent.TimeUnit;
 
+import javax.annotation.Nullable;
+
 import com.google.common.annotations.VisibleForTesting;
 import com.google.common.base.Joiner;
 import com.google.common.collect.ImmutableMap;
@@ -49,6 +51,7 @@ import org.apache.cassandra.db.ConsistencyLevel;
 import org.apache.cassandra.db.Keyspace;
 import org.apache.cassandra.dht.Range;
 import org.apache.cassandra.dht.Token;
+import org.apache.cassandra.db.compression.CompressionDictionary;
 import org.apache.cassandra.locator.InetAddressAndPort;
 import org.apache.cassandra.repair.CommonRange;
 import org.apache.cassandra.repair.messages.RepairOption;
@@ -56,7 +59,6 @@ import org.apache.cassandra.utils.FBUtilities;
 import org.apache.cassandra.utils.TimeUUID;
 
 import static java.lang.String.format;
-
 import static org.apache.cassandra.utils.ByteBufferUtil.bytes;
 
 public final class SystemDistributedKeyspace
@@ -83,10 +85,11 @@ public final class SystemDistributedKeyspace
      * gen 5: add ttl and TWCS to repair_history tables
      * gen 6: add denylist table
      * gen 7: add auto_repair_history and auto_repair_priority tables for AutoRepair feature
+     * gen 8: add compression_dictionaries for dictionary-based compression algorithms (e.g. zstd)
      *
      * // TODO: TCM - how do we evolve these tables?
      */
-    public static final long GENERATION = 7;
+    public static final long GENERATION = 8;
 
     public static final String REPAIR_HISTORY = "repair_history";
 
@@ -100,7 +103,12 @@ public final class SystemDistributedKeyspace
 
     public static final String AUTO_REPAIR_PRIORITY = "auto_repair_priority";
 
-    public static final Set<String> TABLE_NAMES = ImmutableSet.of(REPAIR_HISTORY, PARENT_REPAIR_HISTORY, VIEW_BUILD_STATUS, PARTITION_DENYLIST_TABLE, AUTO_REPAIR_HISTORY, AUTO_REPAIR_PRIORITY);
+    public static final String COMPRESSION_DICTIONARIES = "compression_dictionaries";
+
+    public static final Set<String> TABLE_NAMES = ImmutableSet.of(REPAIR_HISTORY, PARENT_REPAIR_HISTORY,
+                                                                  VIEW_BUILD_STATUS, PARTITION_DENYLIST_TABLE,
+                                                                  AUTO_REPAIR_HISTORY, AUTO_REPAIR_PRIORITY,
+                                                                  COMPRESSION_DICTIONARIES);
 
     public static final String REPAIR_HISTORY_CQL = "CREATE TABLE IF NOT EXISTS %s ("
                                                      + "keyspace_name text,"
@@ -185,6 +193,20 @@ public final class SystemDistributedKeyspace
     private static final TableMetadata AutoRepairPriorityTable =
             parse(AUTO_REPAIR_PRIORITY, "Auto repair priority for each group", AUTO_REPAIR_PRIORITY_CQL).build();
 
+    public static final String COMPRESSION_DICTIONARIES_CQL = "CREATE TABLE IF NOT EXISTS %s (" +
+                                                              "keyspace_name text," +
+                                                              "table_name text," +
+                                                              "kind text," +
+                                                              "dict_id bigint," +
+                                                              "dict blob," +
+                                                              "dict_length int," +
+                                                              "dict_checksum int," +
+                                                              "PRIMARY KEY ((keyspace_name, table_name), dict_id)) " +
+                                                              "WITH CLUSTERING ORDER BY (dict_id DESC)"; // in order to retrieve the latest dictionary; the contract is the newer the dictionary the larger the dict_id
+
+    private static final TableMetadata CompressionDictionariesTable =
+        parse(COMPRESSION_DICTIONARIES, "Compression dictionaries for applicable tables", COMPRESSION_DICTIONARIES_CQL).build();
+
     private static TableMetadata.Builder parse(String table, String description, String cql)
     {
         return CreateTableStatement.parse(format(cql, table), SchemaConstants.DISTRIBUTED_KEYSPACE_NAME)
@@ -197,7 +219,10 @@ public final class SystemDistributedKeyspace
     {
         return KeyspaceMetadata.create(SchemaConstants.DISTRIBUTED_KEYSPACE_NAME,
                                        KeyspaceParams.simple(Math.max(DEFAULT_RF, DatabaseDescriptor.getDefaultKeyspaceRF())),
-                                       Tables.of(RepairHistory, ParentRepairHistory, ViewBuildStatus, PartitionDenylistTable, AutoRepairHistoryTable, AutoRepairPriorityTable));
+                                       Tables.of(RepairHistory, ParentRepairHistory,
+                                                 ViewBuildStatus, PartitionDenylistTable,
+                                                 AutoRepairHistoryTable, AutoRepairPriorityTable,
+                                                 CompressionDictionariesTable));
     }
 
     public static void startParentRepair(TimeUUID parent_id, String keyspaceName, String[] cfnames, RepairOption options)
@@ -382,20 +407,101 @@ public final class SystemDistributedKeyspace
         forceBlockingFlush(VIEW_BUILD_STATUS, ColumnFamilyStore.FlushReason.INTERNALLY_FORCED);
     }
 
-    private static void processSilent(String fmtQry, String... values)
+    /**
+     * Stores a compression dictionary for a given keyspace and table in the distributed system keyspace.
+     *
+     * @param keyspaceName the keyspace name to associate with the dictionary
+     * @param tableName the table name to associate with the dictionary
+     * @param dictionary the compression dictionary to store
+     */
+    public static void storeCompressionDictionary(String keyspaceName, String tableName, CompressionDictionary dictionary)
     {
+        byte[] dict = dictionary.rawDictionary();
+        int checksum = CompressionDictionary.calculateChecksum((byte) dictionary.kind().ordinal(), dictionary.dictId().id, dict);
+        String query = "INSERT INTO %s.%s (keyspace_name, table_name, kind, dict_id, dict, dict_length, dict_checksum) VALUES ('%s', '%s', '%s', %s, ?, %s, %s)";
+        String fmtQuery = format(query,
+                                 SchemaConstants.DISTRIBUTED_KEYSPACE_NAME,
+                                 COMPRESSION_DICTIONARIES,
+                                 keyspaceName,
+                                 tableName,
+                                 dictionary.kind(),
+                                 dictionary.dictId().id,
+                                 dict.length,
+                                 checksum);
+        noThrow(fmtQuery,
+                () -> QueryProcessor.process(fmtQuery, ConsistencyLevel.ONE,
+                                             Collections.singletonList(ByteBuffer.wrap(dict))));
+    }
+
+    /**
+     * Retrieves the latest compression dictionary for a given keyspace and table.
+     *
+     * @param keyspaceName the keyspace name to retrieve the dictionary for
+     * @param tableName the table name to retrieve the dictionary for
+     * @return the latest compression dictionary for the specified keyspace and table,
+     *         or null if no dictionary exists or if an error occurs during retrieval
+     */
+    @Nullable
+    public static CompressionDictionary retrieveLatestCompressionDictionary(String keyspaceName, String tableName)
+    {
+        String query = "SELECT kind, dict_id, dict, dict_length, dict_checksum FROM %s.%s WHERE keyspace_name='%s' AND table_name='%s' LIMIT 1";
+        String fmtQuery = format(query, SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, COMPRESSION_DICTIONARIES, keyspaceName, tableName);
         try
         {
+            UntypedResultSet.Row row = QueryProcessor.execute(fmtQuery, ConsistencyLevel.ONE).one();
+            return CompressionDictionary.createFromRow(row);
+        }
+        catch (Exception e)
+        {
+            return null;
+        }
+    }
+
+    /**
+     * Retrieves a specific compression dictionary for a given keyspace and table.
+     *
+     * @param keyspaceName the keyspace name to retrieve the dictionary for
+     * @param tableName the table name to retrieve the dictionary for
+     * @param dictionaryId the dictionary id to retrieve the dictionary for
+     * @return the compression dictionary identified by the specified keyspace, table and dictionaryId,
+     *         or null if no dictionary exists or if an error occurs during retrieval
+     */
+    public static CompressionDictionary retrieveCompressionDictionary(String keyspaceName, String tableName, CompressionDictionary.DictId dictionaryId)
+    {
+        String query = "SELECT kind, dict_id, dict, dict_length, dict_checksum FROM %s.%s WHERE keyspace_name='%s' AND table_name='%s' AND dict_id=%s";
+        String fmtQuery = format(query, SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, COMPRESSION_DICTIONARIES, keyspaceName, tableName, dictionaryId.id);
+        try
+        {
+            UntypedResultSet.Row row = QueryProcessor.execute(fmtQuery, ConsistencyLevel.ONE).one();
+            return CompressionDictionary.createFromRow(row);
+        }
+        catch (Exception e)
+        {
+            return null;
+        }
+    }
+
+    private static void processSilent(String fmtQry, String... values)
+    {
+        noThrow(fmtQry, () -> {
             List<ByteBuffer> valueList = new ArrayList<>(values.length);
             for (String v : values)
             {
                 valueList.add(bytes(v));
             }
             QueryProcessor.process(fmtQry, ConsistencyLevel.ANY, valueList);
+        });
+    }
+
+    private static void noThrow(String fmtQry, Runnable queryExec)
+    {
+        try
+        {
+            queryExec.run();
         }
         catch (Throwable t)
         {
-            logger.error("Error executing query "+fmtQry, t);
+            logger.error("Error executing query " + fmtQry, t);
         }
     }
 

src/java/org/apache/cassandra/tools/NodeProbe.java

@@ -43,6 +43,7 @@ import java.util.concurrent.ExecutionException;
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.TimeoutException;
 import javax.annotation.Nullable;
+import javax.management.InstanceNotFoundException;
 import javax.management.JMX;
 import javax.management.MBeanServerConnection;
 import javax.management.MalformedObjectNameException;
@@ -88,10 +89,12 @@ import org.apache.cassandra.batchlog.BatchlogManagerMBean;
 import org.apache.cassandra.db.ColumnFamilyStoreMBean;
 import org.apache.cassandra.db.compaction.CompactionManager;
 import org.apache.cassandra.db.compaction.CompactionManagerMBean;
-import org.apache.cassandra.db.virtual.CIDRFilteringMetricsTable;
-import org.apache.cassandra.db.virtual.CIDRFilteringMetricsTableMBean;
+import org.apache.cassandra.db.compression.CompressionDictionaryManagerMBean;
+import org.apache.cassandra.db.compression.TrainingState;
 import org.apache.cassandra.db.guardrails.Guardrails;
 import org.apache.cassandra.db.guardrails.GuardrailsMBean;
+import org.apache.cassandra.db.virtual.CIDRFilteringMetricsTable;
+import org.apache.cassandra.db.virtual.CIDRFilteringMetricsTableMBean;
 import org.apache.cassandra.fql.FullQueryLoggerOptions;
 import org.apache.cassandra.fql.FullQueryLoggerOptionsCompositeData;
 import org.apache.cassandra.gms.FailureDetector;
@@ -2682,6 +2685,68 @@ public class NodeProbe implements AutoCloseable
     {
         autoRepairProxy.setMixedMajorVersionRepairEnabled(enabled);
     }
+
+    /**
+     * Triggers compression dictionary training for the specified table.
+     * Samples chunks from existing SSTables and trains a dictionary.
+     *
+     * @param keyspace the keyspace name
+     * @param table the table name
+     * @param force force the dictionary training even if there are not enough samples
+     * @throws IOException if there's an error accessing the MBean
+     * @throws IllegalArgumentException if table doesn't support dictionary compression
+     */
+    public void trainCompressionDictionary(String keyspace, String table, boolean force) throws IOException
+    {
+        CompressionDictionaryManagerMBean proxy = getDictionaryManagerProxy(keyspace, table);
+        try
+        {
+            proxy.train(force);
+        }
+        catch (Exception e)
+        {
+            if (e.getCause() instanceof InstanceNotFoundException)
+            {
+                String message = String.format("Table %s.%s does not exist or does not support dictionary compression",
+                                               keyspace, table);
+                throw new IOException(message);
+            }
+            else
+            {
+                throw new IOException(e.getMessage());
+            }
+        }
+    }
+
+    /**
+     * Gets the compression dictionary training state for the specified table.
+     * Returns an atomic snapshot of training status, progress, and failure details.
+     *
+     * @param keyspace the keyspace name
+     * @param table the table name
+     * @return the current training state
+     * @throws IOException if there's an error accessing the MBean
+     */
+    public TrainingState getCompressionDictionaryTrainingState(String keyspace, String table) throws IOException
+    {
+        CompositeData compositeData = getDictionaryManagerProxy(keyspace, table).getTrainingState();
+        return TrainingState.fromCompositeData(compositeData);
+    }
+
+    private CompressionDictionaryManagerMBean getDictionaryManagerProxy(String keyspace, String table) throws IOException
+    {
+        // Construct table-specific MBean name
+        String mbeanName = CompressionDictionaryManagerMBean.MBEAN_NAME + ",keyspace=" + keyspace + ",table=" + table;
+        try
+        {
+            ObjectName objectName = new ObjectName(mbeanName);
+            return JMX.newMBeanProxy(mbeanServerConn, objectName, CompressionDictionaryManagerMBean.class);
+        }
+        catch (MalformedObjectNameException e)
+        {
+            throw new IOException("Invalid keyspace or table name", e);
+        }
+    }
 }
 
 class ColumnFamilyStoreMBeanIterator implements Iterator<Map.Entry<String, ColumnFamilyStoreMBean>>

src/java/org/apache/cassandra/tools/SSTableMetadataViewer.java

@@ -322,7 +322,7 @@ public class SSTableMetadataViewer
         CompactionMetadata compaction = statsComponent.compactionMetadata();
         SerializationHeader.Component header = statsComponent.serializationHeader();
         Class<? extends ICompressor> compressorClass = null;
-        try (CompressionMetadata compression = CompressionInfoComponent.loadIfExists(descriptor))
+        try (CompressionMetadata compression = CompressionInfoComponent.loadIfExists(descriptor, null))
         {
             compressorClass = compression != null ? compression.compressor().getClass() : null;
         }

src/java/org/apache/cassandra/tools/nodetool/CompressionDictionary.java

@@ -0,0 +1,122 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.cassandra.tools.nodetool;
+
+import java.io.PrintStream;
+import java.util.concurrent.TimeUnit;
+
+import com.google.common.util.concurrent.Uninterruptibles;
+
+import org.apache.cassandra.db.compression.ICompressionDictionaryTrainer.TrainingStatus;
+import org.apache.cassandra.db.compression.TrainingState;
+import org.apache.cassandra.tools.NodeProbe;
+import org.apache.cassandra.utils.Clock;
+import picocli.CommandLine.Command;
+import picocli.CommandLine.Option;
+import picocli.CommandLine.Parameters;
+
+@Command(name = "compressiondictionary",
+         description = "Manage compression dictionaries",
+         subcommands = { CompressionDictionary.Train.class })
+public class CompressionDictionary
+{
+    @Command(name = "train",
+             description = "Manually trigger compression dictionary training for a table. If no SSTables are available, the memtable will be flushed first.")
+    public static class Train extends AbstractCommand
+    {
+        @Parameters(index = "0", description = "The keyspace name", arity = "1")
+        private String keyspace;
+
+        @Parameters(index = "1", description = "The table name", arity = "1")
+        private String table;
+
+        @Option(names = { "-f", "--force" }, description = "Force the dictionary training even if there are not enough samples")
+        private boolean force = false;
+
+        @Override
+        public void execute(NodeProbe probe)
+        {
+            PrintStream out = probe.output().out;
+            PrintStream err = probe.output().err;
+
+            try
+            {
+                out.printf("Starting compression dictionary training for %s.%s...%n", keyspace, table);
+                out.printf("Training from existing SSTables (flushing first if needed)%n");
+
+                probe.trainCompressionDictionary(keyspace, table, force);
+
+                // Wait for training completion (10 minutes timeout for SSTable-based training)
+                out.println("Sampling from existing SSTables and training.");
+                long maxWaitMillis = TimeUnit.MINUTES.toMillis(10);
+                long startTime = Clock.Global.currentTimeMillis();
+
+                while (Clock.Global.currentTimeMillis() - startTime < maxWaitMillis)
+                {
+                    TrainingState trainingState = probe.getCompressionDictionaryTrainingState(keyspace, table);
+                    TrainingStatus status = trainingState.getStatus();
+                    displayProgress(trainingState, startTime, out, status);
+                    if (TrainingStatus.COMPLETED == status)
+                    {
+                        out.printf("%nTraining completed successfully for %s.%s%n", keyspace, table);
+                        return;
+                    }
+                    else if (TrainingStatus.FAILED == status)
+                    {
+                        err.printf("%nTraining failed for %s.%s%n", keyspace, table);
+                        try
+                        {
+                            String failureMessage = trainingState.getFailureMessage();
+                            if (failureMessage != null && !failureMessage.isEmpty())
+                            {
+                                err.printf("Reason: %s%n", failureMessage);
+                            }
+                        }
+                        catch (Exception e)
+                        {
+                            // If we can't get the failure message, just continue without it
+                        }
+                        System.exit(1);
+                    }
+
+                    Uninterruptibles.sleepUninterruptibly(1, TimeUnit.SECONDS);
+                }
+
+                err.printf("%nTraining did not complete within expected timeframe (10 minutes).%n");
+                System.exit(1);
+            }
+            catch (Exception e)
+            {
+                err.printf("Failed to trigger training: %s%n", e.getMessage());
+                System.exit(1);
+            }
+        }
+
+        private static void displayProgress(TrainingState trainingState, long startTime, PrintStream out, TrainingStatus status)
+        {
+            // Display meaningful statistics
+            long sampleCount = trainingState.getSampleCount();
+            long totalSampleSize = trainingState.getTotalSampleSize();
+            long elapsedSeconds = (Clock.Global.currentTimeMillis() - startTime) / 1000;
+            double sampleSizeMB = totalSampleSize / (1024.0 * 1024.0);
+
+            out.printf("\rStatus: %s | Samples: %d | Size: %.2f MiB | Elapsed: %ds",
+                       status, sampleCount, sampleSizeMB, elapsedSeconds);
+        }
+    }
+}

src/java/org/apache/cassandra/tools/nodetool/NodetoolCommand.java

@@ -207,6 +207,7 @@ import static org.apache.cassandra.tools.nodetool.Help.printTopCommandUsage;
                          TableStats.class,
                          TopPartitions.class,
                          TpStats.class,
+                         CompressionDictionary.class,
                          TruncateHints.class,
                          UpdateCIDRGroup.class,
                          UpgradeSSTable.class,

src/java/org/apache/cassandra/utils/StorageCompatibilityMode.java

@@ -35,6 +35,13 @@ public enum StorageCompatibilityMode
      */
     CASSANDRA_4(4),
 
+    /**
+     * Similar to CASSANDRA_4.
+     * The new features in 6.0 are
+     * - ZSTD dictionary-based compression. Once SSTables are compressed with dictionary, they cannot be rolled back.
+     */
+    CASSANDRA_5(5),
+
     /**
      * Use the storage formats of the current version, but disabling features that are not compatible with any
      * not-upgraded nodes in the cluster. Use this during rolling upgrades to a new major Cassandra version. Once all

test/microbench/org/apache/cassandra/test/microbench/ZstdDictionaryCompressorBenchBase.java

@@ -0,0 +1,231 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.test.microbench;
+
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.util.Map;
+import java.util.Random;
+import java.util.UUID;
+
+import com.github.luben.zstd.ZstdDictTrainer;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.TearDown;
+
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import org.apache.cassandra.db.compression.ZstdCompressionDictionary;
+import org.apache.cassandra.io.compress.ZstdDictionaryCompressor;
+
+// The bench takes over 20 minutes to finish
+@State(Scope.Benchmark)
+public abstract class ZstdDictionaryCompressorBenchBase
+{
+    @Param({"4096", "16384", "65536"})
+    protected int dataSize;
+
+    @Param({"CASSANDRA_LIKE", "COMPRESSIBLE", "MIXED"})
+    protected DataType dataType;
+
+    @Param({"0", "65536"})
+    protected int dictionarySize;
+
+    @Param({"3", "5", "7"})
+    protected int compressionLevel;
+
+    protected byte[] inputData;
+    protected ByteBuffer inputBuffer;
+    protected ByteBuffer compressedBuffer;
+    protected ByteBuffer decompressedBuffer;
+    protected ZstdDictionaryCompressor compressor;
+    protected ZstdDictionaryCompressor noDictCompressor;
+    protected ZstdCompressionDictionary dictionary;
+
+    public enum DataType
+    {
+        CASSANDRA_LIKE, COMPRESSIBLE, MIXED
+    }
+
+    @Setup(Level.Trial)
+    public void setupTrial()
+    {
+        DatabaseDescriptor.daemonInitialization();
+    }
+
+    @Setup(Level.Iteration)
+    public void setupIteration() throws IOException
+    {
+        Random random = new Random(42);
+
+        // Generate test data based on type
+        inputData = generateTestData(dataType, dataSize, random);
+
+        // Create direct ByteBuffers (required by ZSTD)
+        inputBuffer = ByteBuffer.allocateDirect(dataSize);
+        inputBuffer.put(inputData);
+        inputBuffer.flip();
+
+        // Allocate buffers with extra space for compression overhead
+        int maxCompressedSize = dataSize + 1024;
+        compressedBuffer = ByteBuffer.allocateDirect(maxCompressedSize);
+        decompressedBuffer = ByteBuffer.allocateDirect(dataSize);
+
+        // Create dictionary if needed
+        if (dictionarySize != 0)
+        {
+            dictionary = createDictionary(dataType, dictionarySize, random);
+            Map<String, String> options = Map.of("compression_level", String.valueOf(compressionLevel));
+            compressor = ZstdDictionaryCompressor.create(options).getOrCopyWithDictionary(dictionary);
+        }
+        else
+        {
+            Map<String, String> options = Map.of("compression_level", String.valueOf(compressionLevel));
+            compressor = ZstdDictionaryCompressor.create(options);
+        }
+
+        // Always create a no-dictionary compressor for comparison
+        Map<String, String> options = Map.of("compression_level", String.valueOf(compressionLevel));
+        noDictCompressor = ZstdDictionaryCompressor.create(options);
+    }
+
+    @TearDown(Level.Iteration)
+    public void tearDown()
+    {
+        if (dictionary != null)
+        {
+            dictionary.close();
+            dictionary = null;
+        }
+        ZstdDictionaryCompressor.invalidateCache();
+    }
+
+    protected byte[] generateTestData(DataType type, int size, Random random)
+    {
+        byte[] data = new byte[size];
+
+        switch (type)
+        {
+            case CASSANDRA_LIKE:
+                generateCassandraLikeData(data, random);
+                break;
+
+            case COMPRESSIBLE:
+                generateCompressibleData(data, random);
+                break;
+
+            case MIXED:
+                generateMixedData(data, random);
+                break;
+        }
+
+        return data;
+    }
+
+    private void generateCassandraLikeData(byte[] data, Random random)
+    {
+        StringBuilder sb = new StringBuilder();
+        String[] patterns = {
+            "user_id_", "timestamp_", "session_", "event_type_",
+            "metadata_", "value_", "status_", "location_"
+        };
+
+        while (sb.length() < data.length)
+        {
+            String pattern = patterns[random.nextInt(patterns.length)];
+            sb.append(pattern).append(UUID.randomUUID().toString()).append("|");
+            sb.append("timestamp:").append(System.currentTimeMillis() + random.nextInt(86400000)).append("|");
+            sb.append("value:").append(random.nextDouble()).append("|");
+            sb.append("count:").append(random.nextInt(1000)).append("\n");
+        }
+
+        byte[] generated = sb.substring(0, Math.min(data.length, sb.length())).getBytes();
+        System.arraycopy(generated, 0, data, 0, generated.length);
+
+        // Fill remaining space with random data if needed
+        if (generated.length < data.length)
+        {
+            byte[] remaining = new byte[data.length - generated.length];
+            random.nextBytes(remaining);
+            System.arraycopy(remaining, 0, data, generated.length, remaining.length);
+        }
+    }
+
+    private void generateCompressibleData(byte[] data, Random random)
+    {
+        String pattern = "The quick brown fox jumps over the lazy dog. This is a highly compressible pattern that repeats. ";
+        byte[] patternBytes = pattern.getBytes();
+
+        for (int i = 0; i < data.length; i++)
+        {
+            data[i] = patternBytes[i % patternBytes.length];
+        }
+
+        // Add some randomness (10%)
+        for (int i = 0; i < data.length / 10; i++)
+        {
+            data[random.nextInt(data.length)] = (byte) random.nextInt(256);
+        }
+    }
+
+    private void generateMixedData(byte[] data, Random random)
+    {
+        int quarter = data.length / 4;
+
+        // 25% random
+        random.nextBytes(data);
+
+        // 25% compressible
+        byte[] compressible = new byte[quarter];
+        generateCompressibleData(compressible, random);
+        System.arraycopy(compressible, 0, data, quarter, quarter);
+
+        // 50% Cassandra-like
+        byte[] cassandraLike = new byte[data.length - 2 * quarter];
+        generateCassandraLikeData(cassandraLike, random);
+        System.arraycopy(cassandraLike, 0, data, 2 * quarter, cassandraLike.length);
+    }
+
+    private ZstdCompressionDictionary createDictionary(DataType dataType, int dictSize, Random random)
+    {
+        // Generate training samples
+        byte[][] samples = new byte[100][];
+        int totalSampleSize = 0;
+        for (int i = 0; i < samples.length; i++)
+        {
+            samples[i] = generateTestData(dataType, Math.min(1024, dataSize), random);
+            totalSampleSize += samples[i].length;
+        }
+
+        // Train dictionary
+        ZstdDictTrainer trainer = new ZstdDictTrainer(totalSampleSize, dictSize);
+        for (byte[] sample : samples)
+        {
+            trainer.addSample(sample);
+        }
+
+        byte[] dictData = trainer.trainSamples();
+        DictId dictId = new DictId(Kind.ZSTD, 0);
+        return new ZstdCompressionDictionary(dictId, dictData);
+    }
+}

test/microbench/org/apache/cassandra/test/microbench/ZstdDictionaryCompressorRatioBench.java

@@ -0,0 +1,147 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.test.microbench;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+
+import org.apache.cassandra.config.DatabaseDescriptor;
+
+// This is not really a bench, but share common utilties from the base class.
+// Running the class output compression ratio and dictionary effectiveness of different configurations
+public class ZstdDictionaryCompressorRatioBench extends ZstdDictionaryCompressorBenchBase
+{
+    private static class CompressionResult
+    {
+        final String configuration;
+        final double compressionRatio;
+        final double dictionaryEffectiveness;
+
+        CompressionResult(String configuration, double compressionRatio, double dictionaryEffectiveness)
+        {
+            this.configuration = configuration;
+            this.compressionRatio = compressionRatio;
+            this.dictionaryEffectiveness = dictionaryEffectiveness;
+        }
+    }
+
+    private CompressionResult measureCompressionRatio() throws IOException
+    {
+        // Compress with current compressor
+        inputBuffer.rewind();
+        compressedBuffer.clear();
+        compressor.compress(inputBuffer, compressedBuffer);
+        int compressedSize = compressedBuffer.position();
+
+        // Compress without dictionary for comparison
+        inputBuffer.rewind();
+        compressedBuffer.clear();
+        noDictCompressor.compress(inputBuffer, compressedBuffer);
+        int noDictCompressedSize = compressedBuffer.position();
+
+        // Calculate ratios
+        double compressionRatio = (double) inputBuffer.limit() / compressedSize;
+        double dictionaryEffectiveness = (double) noDictCompressedSize / compressedSize;
+
+        // Create configuration string
+        String config = String.format("%s_%s_L%d_Chunk%dKiB",
+                                      dataType,
+                                      dictionarySize == 0 ? "NoDict" : "WithDict",
+                                      compressionLevel,
+                                      dataSize / 1024);
+
+        return new CompressionResult(config, compressionRatio, dictionaryEffectiveness);
+    }
+
+    public static void main(String[] args) throws Exception
+    {
+        DatabaseDescriptor.daemonInitialization();
+
+        List<CompressionResult> allResults = new ArrayList<>();
+
+        // Define test parameters
+        DataType[] dataTypes = {DataType.CASSANDRA_LIKE, DataType.COMPRESSIBLE, DataType.MIXED};
+        int[] dictionarySizes = {0, 65536};
+        int[] compressionLevels = {3, 5, 7};
+        int[] dataSizes = {4096, 16384, 65536};
+
+        System.out.println("Running ZSTD Dictionary Compressor Ratio Measurements...");
+        System.out.println("Total configurations: " + (dataTypes.length * dictionarySizes.length * compressionLevels.length * dataSizes.length));
+
+        int configCount = 0;
+        for (DataType dataType : dataTypes)
+        {
+            for (int dictionarySize : dictionarySizes)
+            {
+                for (int compressionLevel : compressionLevels)
+                {
+                    for (int dataSize : dataSizes)
+                    {
+                        configCount++;
+                        ZstdDictionaryCompressorRatioBench bench = new ZstdDictionaryCompressorRatioBench();
+                        bench.dataType = dataType;
+                        bench.dictionarySize = dictionarySize;
+                        bench.compressionLevel = compressionLevel;
+                        bench.dataSize = dataSize;
+
+                        try
+                        {
+                            bench.setupIteration();
+                            CompressionResult result = bench.measureCompressionRatio();
+                            allResults.add(result);
+                            bench.tearDown();
+                        }
+                        catch (Exception e)
+                        {
+                            System.err.println("Failed to process configuration: " + e.getMessage());
+                            e.printStackTrace();
+                        }
+                    }
+                }
+            }
+        }
+
+        // Print consolidated results
+        printConsolidatedResults(allResults);
+    }
+
+    private static void printConsolidatedResults(List<CompressionResult> results)
+    {
+        StringBuilder report = new StringBuilder();
+        report.append("\n").append("=".repeat(100)).append("\n");
+        report.append("ZSTD DICTIONARY COMPRESSOR RATIO RESULTS").append("\n");
+        report.append("=".repeat(100)).append("\n");
+        report.append(String.format("%-50s %-20s %-20s%n", "Configuration", "Compression Ratio", "Dictionary Effectiveness"));
+        report.append("-".repeat(100)).append("\n");
+
+        for (CompressionResult entry : results)
+        {
+            report.append(String.format("%-50s %-20.3f %-20.3f%n",
+                entry.configuration, entry.compressionRatio, entry.dictionaryEffectiveness));
+        }
+
+        report.append("=".repeat(100)).append("\n");
+        report.append("Compression Ratio: Original Size / Compressed Size (higher is better)").append("\n");
+        report.append("Dictionary Effectiveness: Non-Dict Size / Dict Size (higher is better)").append("\n");
+        report.append("=".repeat(100)).append("\n");
+
+        System.out.print(report.toString());
+    }
+}

test/microbench/org/apache/cassandra/test/microbench/ZstdDictionaryCompressorThroughputBench.java

@@ -0,0 +1,65 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.test.microbench;
+
+import java.io.IOException;
+import java.util.concurrent.TimeUnit;
+
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+@BenchmarkMode({Mode.Throughput, Mode.AverageTime})
+@OutputTimeUnit(TimeUnit.MILLISECONDS)
+@Warmup(iterations = 1, time = 1, timeUnit = TimeUnit.SECONDS)
+@Measurement(iterations = 2, time = 2, timeUnit = TimeUnit.SECONDS)
+@Fork(value = 1, jvmArgsAppend = "-Xmx1G")
+public class ZstdDictionaryCompressorThroughputBench extends ZstdDictionaryCompressorBenchBase
+{
+    @Benchmark
+    public void compressionThroughput(Blackhole bh) throws IOException
+    {
+        inputBuffer.rewind();
+        compressedBuffer.clear();
+
+        compressor.compress(inputBuffer, compressedBuffer);
+        bh.consume(compressedBuffer.position());
+    }
+
+    @Benchmark
+    public void decompressionThroughput(Blackhole bh) throws IOException
+    {
+        // First compress the data
+        inputBuffer.rewind();
+        compressedBuffer.clear();
+        compressor.compress(inputBuffer, compressedBuffer);
+
+        // Then decompress it
+        compressedBuffer.flip();
+        decompressedBuffer.clear();
+        compressor.uncompress(compressedBuffer, decompressedBuffer);
+
+        bh.consume(decompressedBuffer.position());
+    }
+}

test/resources/nodetool/help/compressiondictionary

@@ -0,0 +1,44 @@
+NAME
+        nodetool compressiondictionary - Manage compression dictionaries
+
+SYNOPSIS
+        nodetool [(-h <host> | --host <host>)] [(-p <port> | --port <port>)]
+                [(-pp | --print-port)] [(-pw <password> | --password <password>)]
+                [(-pwf <passwordFilePath> | --password-file <passwordFilePath>)]
+                [(-u <username> | --username <username>)] compressiondictionary
+                <command> [<args>]
+
+        nodetool [(-h <host> | --host <host>)] [(-p <port> | --port <port>)]
+                [(-pp | --print-port)] [(-pw <password> | --password <password>)]
+                [(-pwf <passwordFilePath> | --password-file <passwordFilePath>)]
+                [(-u <username> | --username <username>)] compressiondictionary train
+                [(-f | --force)] [--] <keyspace> <table>
+
+OPTIONS
+        -h <host>, --host <host>
+            Node hostname or ip address
+
+        -p <port>, --port <port>
+            Remote jmx agent port number
+
+        -pp, --print-port
+            Operate in 4.0 mode with hosts disambiguated by port number
+
+        -pw <password>, --password <password>
+            Remote jmx agent password
+
+        -pwf <passwordFilePath>, --password-file <passwordFilePath>
+            Path to the JMX password file
+
+        -u <username>, --username <username>
+            Remote jmx agent username
+
+COMMANDS
+        With no arguments, Display help information
+
+        train
+            Manually trigger compression dictionary training for a table. If no
+            SSTables are available, the memtable will be flushed first.
+
+            With --force option, Force the dictionary training even if there are not
+            enough samples

test/resources/nodetool/help/compressiondictionary$train

@@ -0,0 +1,44 @@
+NAME
+        nodetool compressiondictionary train - Manually trigger compression
+        dictionary training for a table. If no SSTables are available, the
+        memtable will be flushed first.
+
+SYNOPSIS
+        nodetool [(-h <host> | --host <host>)] [(-p <port> | --port <port>)]
+                [(-pp | --print-port)] [(-pw <password> | --password <password>)]
+                [(-pwf <passwordFilePath> | --password-file <passwordFilePath>)]
+                [(-u <username> | --username <username>)] compressiondictionary train
+                [(-f | --force)] [--] <keyspace> <table>
+
+OPTIONS
+        -f, --force
+            Force the dictionary training even if there are not enough samples
+
+        -h <host>, --host <host>
+            Node hostname or ip address
+
+        -p <port>, --port <port>
+            Remote jmx agent port number
+
+        -pp, --print-port
+            Operate in 4.0 mode with hosts disambiguated by port number
+
+        -pw <password>, --password <password>
+            Remote jmx agent password
+
+        -pwf <passwordFilePath>, --password-file <passwordFilePath>
+            Path to the JMX password file
+
+        -u <username>, --username <username>
+            Remote jmx agent username
+
+        --
+            This option can be used to separate command-line options from the
+            list of argument, (useful when arguments might be mistaken for
+            command-line options
+
+        <keyspace>
+            The keyspace name
+
+        <table>
+            The table name

test/resources/nodetool/help/nodetool

@@ -21,6 +21,7 @@ The most commonly used nodetool commands are:
     compact                             Force a (major) compaction on one or more tables or user-defined compaction on given SSTables
     compactionhistory                   Print history of compaction
     compactionstats                     Print statistics on compactions
+    compressiondictionary               Manage compression dictionaries
     consensus_admin                     List and mark ranges as migrating between consensus protocols
     datapaths                           Print all directories where data of tables are stored
     decommission                        Decommission the *node I am connecting to*

test/unit/org/apache/cassandra/db/compression/CompressionDictionaryCacheTest.java

@@ -0,0 +1,438 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicInteger;
+import java.util.concurrent.atomic.AtomicReference;
+
+import org.junit.After;
+import org.junit.Before;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import com.github.luben.zstd.ZstdDictTrainer;
+import org.apache.cassandra.config.DatabaseDescriptor;
+
+import static org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import static org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import static org.assertj.core.api.Assertions.assertThat;
+
+public class CompressionDictionaryCacheTest
+{
+    private static final String TEST_PATTERN = "The quick brown fox jumps over the lazy dog. ";
+
+    private CompressionDictionaryCache cache;
+    private ZstdCompressionDictionary testDict1;
+    private ZstdCompressionDictionary testDict2;
+    private ZstdCompressionDictionary testDict3;
+
+    @BeforeClass
+    public static void setUpClass()
+    {
+        DatabaseDescriptor.daemonInitialization();
+    }
+
+    @Before
+    public void setUp()
+    {
+        cache = new CompressionDictionaryCache();
+        testDict1 = createTestDictionary(1);
+        testDict2 = createTestDictionary(2);
+        testDict3 = createTestDictionary(3);
+    }
+
+    @After
+    public void tearDown()
+    {
+        if (cache != null)
+        {
+            cache.close();
+        }
+
+        // Close dictionaries if not already closed
+        closeQuietly(testDict1);
+        closeQuietly(testDict2);
+        closeQuietly(testDict3);
+    }
+
+    // Basic cache operations tests
+
+    @Test
+    public void testGetCurrentInitiallyNull()
+    {
+        assertThat(cache.getCurrent())
+        .as("Current dictionary should be null initially")
+        .isNull();
+    }
+
+    @Test
+    public void testAddAndGet()
+    {
+        cache.add(testDict1);
+
+        CompressionDictionary retrieved = cache.get(testDict1.dictId());
+        assertThat(retrieved)
+        .as("Should retrieve the same dictionary instance")
+        .isSameAs(testDict1);
+    }
+
+    @Test
+    public void testGetNonExistentDictionary()
+    {
+        DictId nonExistentId = new DictId(Kind.ZSTD, 999);
+        assertThat(cache.get(nonExistentId))
+        .as("Should return null for non-existent dictionary")
+        .isNull();
+    }
+
+    @Test
+    public void testAddMultipleDictionaries()
+    {
+        cache.add(testDict1);
+        cache.add(testDict2);
+        cache.add(testDict3);
+
+        assertThat(cache.get(testDict1.dictId())).isSameAs(testDict1);
+        assertThat(cache.get(testDict2.dictId())).isSameAs(testDict2);
+        assertThat(cache.get(testDict3.dictId())).isSameAs(testDict3);
+    }
+
+    @Test
+    public void testSetCurrentWithNewerDictionary()
+    {
+        cache.add(testDict1);
+        cache.add(testDict2);
+
+        assertThat(cache.getCurrent())
+        .as("Should update to newer dictionary")
+        .isSameAs(testDict2);
+
+        // Both should be in cache
+        assertThat(cache.get(testDict1.dictId())).isSameAs(testDict1);
+        assertThat(cache.get(testDict2.dictId())).isSameAs(testDict2);
+    }
+
+    @Test
+    public void testSetCurrentWithOlderDictionary()
+    {
+        cache.add(testDict2);
+        cache.add(testDict1); // older dictionary
+
+        assertThat(cache.getCurrent())
+        .as("Should keep newer dictionary as current")
+        .isSameAs(testDict2);
+
+        // Both should be in cache
+        assertThat(cache.get(testDict1.dictId())).isSameAs(testDict1);
+        assertThat(cache.get(testDict2.dictId())).isSameAs(testDict2);
+    }
+
+    @Test
+    public void testSetCurrentWithSameIdDictionary()
+    {
+        ZstdCompressionDictionary sameDictCopy = createTestDictionary(2);
+
+        cache.add(testDict2);
+        cache.add(sameDictCopy);
+
+        // Should not update since ID is the same (not newer)
+        assertThat(cache.getCurrent())
+        .as("Should keep original dictionary as current")
+        .isSameAs(testDict2);
+
+        sameDictCopy.close();
+    }
+
+    @Test
+    public void testSetCurrentWithNull()
+    {
+        cache.add(testDict1);
+        cache.add(null);
+
+        // Should not change current dictionary
+        assertThat(cache.getCurrent())
+        .as("Should keep existing dictionary as current")
+        .isSameAs(testDict1);
+    }
+
+    @Test
+    public void testCacheClose()
+    {
+        cache.add(testDict1);
+        cache.add(testDict2);
+
+        assertThat(cache.getCurrent())
+        .as("Current should not be null before close")
+        .isNotNull();
+        assertThat(cache.get(testDict1.dictId()))
+        .as("Cache should contain dict1 before close")
+        .isNotNull();
+
+        cache.close();
+
+        assertThat(cache.getCurrent())
+        .as("Current should be null after close")
+        .isNull();
+        assertThat(cache.get(testDict1.dictId()))
+        .as("Cache should not contain dict1 after close")
+        .isNull();
+        assertThat(cache.get(testDict2.dictId()))
+        .as("Cache should not contain dict2 after close")
+        .isNull();
+    }
+
+    @Test
+    public void testCloseIdempotent()
+    {
+        cache.add(testDict1);
+
+        // Close multiple times should not cause issues
+        cache.close();
+        cache.close();
+        cache.close();
+
+        assertThat(cache.getCurrent())
+        .as("Current should remain null")
+        .isNull();
+        assertThat(cache.get(testDict1.dictId()))
+        .as("Cache should remain empty")
+        .isNull();
+    }
+
+    @Test
+    public void testConcurrentAccess() throws InterruptedException
+    {
+        int threadCount = 10;
+        int operationsPerThread = 100;
+        ExecutorService executor = Executors.newFixedThreadPool(threadCount);
+        CountDownLatch startLatch = new CountDownLatch(1);
+        CountDownLatch doneLatch = new CountDownLatch(threadCount);
+        AtomicInteger successCount = new AtomicInteger(0);
+        AtomicReference<Exception> errorRef = new AtomicReference<>();
+
+        // Pre-populate cache
+        cache.add(testDict1);
+        cache.add(testDict2);
+
+        for (int i = 0; i < threadCount; i++)
+        {
+            executor.submit(() -> {
+                try
+                {
+                    startLatch.await();
+
+                    for (int j = 0; j < operationsPerThread; j++)
+                    {
+                        // Mix of read operations
+                        CompressionDictionary current = cache.getCurrent();
+                        cache.get(testDict1.dictId());
+                        cache.get(testDict2.dictId());
+
+                        // Verify consistency
+                        if (current != null && current.dictId().equals(testDict2.dictId()))
+                        {
+                            successCount.incrementAndGet();
+                        }
+                    }
+                }
+                catch (Exception e)
+                {
+                    errorRef.set(e);
+                }
+                finally
+                {
+                    doneLatch.countDown();
+                }
+            });
+        }
+
+        startLatch.countDown(); // Start all threads
+        assertThat(doneLatch.await(10, TimeUnit.SECONDS))
+        .as("Threads should complete within timeout")
+        .isTrue();
+
+        executor.shutdown();
+
+        assertThat(errorRef.get())
+        .as("No errors should occur during concurrent access")
+        .isNull();
+        assertThat(successCount.get())
+        .as("Should have successful read operations")
+        .isGreaterThan(0);
+    }
+
+    @Test
+    public void testConcurrentSetCurrent() throws InterruptedException
+    {
+        int threadCount = 5;
+        ExecutorService executor = Executors.newFixedThreadPool(threadCount);
+        CountDownLatch startLatch = new CountDownLatch(1);
+        CountDownLatch doneLatch = new CountDownLatch(threadCount);
+
+        // Create multiple dictionaries with different IDs
+        ZstdCompressionDictionary[] dicts = new ZstdCompressionDictionary[threadCount];
+        for (int i = 0; i < threadCount; i++)
+        {
+            dicts[i] = createTestDictionary(100 + i); // High IDs to ensure newer
+        }
+
+        for (int i = 0; i < threadCount; i++)
+        {
+            ZstdCompressionDictionary dict = dicts[i];
+            executor.submit(() -> {
+                try
+                {
+                    startLatch.await();
+                    cache.add(dict);
+                }
+                catch (Exception e)
+                {
+                    // Ignore - testing thread safety
+                }
+                finally
+                {
+                    doneLatch.countDown();
+                }
+            });
+        }
+
+        startLatch.countDown();
+        assertThat(doneLatch.await(5, TimeUnit.SECONDS))
+        .as("Threads should complete within timeout")
+        .isTrue();
+
+        executor.shutdown();
+
+        // Verify that a current dictionary was set and it's one of our test dictionaries
+        CompressionDictionary current = cache.getCurrent();
+        assertThat(current)
+        .as("A current dictionary should be set")
+        .isNotNull();
+        assertThat(current.dictId().id)
+        .as("Current dictionary should be one of the test dictionaries")
+        .isBetween(100L, 100L + threadCount);
+
+        // Clean up
+        for (ZstdCompressionDictionary dict : dicts)
+        {
+            closeQuietly(dict);
+        }
+    }
+
+    @Test
+    public void testGetCurrentRefreshesCacheEntry() throws InterruptedException
+    {
+        int expireSeconds = 1;
+        try (CompressionDictionaryCache shortLivedCache = new CompressionDictionaryCache(10, expireSeconds))
+        {
+            shortLivedCache.add(testDict1);
+            assertThat(shortLivedCache.getCurrent())
+            .as("Current dictionary should be set")
+            .isSameAs(testDict1);
+            assertThat(shortLivedCache.get(testDict1.dictId()))
+            .as("Dictionary should be in cache")
+            .isNotNull();
+
+            // Access getCurrent() repeatedly for slightly longer than the expiration time
+            int iterations = expireSeconds * 2 + 1;
+            for (int i = 0; i < iterations; i++)
+            {
+                Thread.sleep(900); // Sleep a bit less than expireSeconds between accesses
+
+                CompressionDictionary current = shortLivedCache.getCurrent();
+                assertThat(current)
+                .as("Current dictionary should remain accessible after %d seconds", i + 1)
+                .isSameAs(testDict1);
+
+                assertThat(shortLivedCache.get(testDict1.dictId()))
+                .as("Dictionary should still be in cache after %d seconds", i + 1)
+                .isNotNull();
+            }
+
+            assertThat(shortLivedCache.getCurrent())
+            .as("Current dictionary should remain accessible due to getCurrent() refreshing the cache")
+            .isSameAs(testDict1);
+        }
+    }
+
+    @Test
+    public void testGetCurrentReturnsNullAfterExpiration() throws InterruptedException
+    {
+        int expireSeconds = 1;
+        try (CompressionDictionaryCache shortLivedCache = new CompressionDictionaryCache(10, expireSeconds))
+        {
+            shortLivedCache.add(testDict1);
+            assertThat(shortLivedCache.getCurrent())
+            .as("Current dictionary should be set")
+            .isSameAs(testDict1);
+
+            // Wait for expiration without accessing getCurrent()
+            Thread.sleep((expireSeconds + 1) * 1000);
+
+            // Current should now return null since the entry expired and we only store the DictId
+            assertThat(shortLivedCache.getCurrent())
+            .as("Current dictionary should be null after expiration when not accessed")
+            .isNull();
+        }
+    }
+
+    private static ZstdCompressionDictionary createTestDictionary(long id)
+    {
+        try
+        {
+            // Create simple dictionary
+            ZstdDictTrainer trainer = new ZstdDictTrainer(10 * 1024, 1024, 3);
+
+            // Add samples
+            byte[] sample = TEST_PATTERN.getBytes();
+            for (int i = 0; i < 100; i++)
+            {
+                trainer.addSample(sample);
+            }
+
+            byte[] dictBytes = trainer.trainSamples();
+            DictId dictId = new DictId(Kind.ZSTD, id);
+
+            return new ZstdCompressionDictionary(dictId, dictBytes);
+        }
+        catch (Exception e)
+        {
+            throw new RuntimeException("Failed to create test dictionary", e);
+        }
+    }
+
+    private static void closeQuietly(AutoCloseable resource)
+    {
+        if (resource != null)
+        {
+            try
+            {
+                resource.close();
+            }
+            catch (Exception e)
+            {
+                // Ignore
+            }
+        }
+    }
+}

test/unit/org/apache/cassandra/db/compression/CompressionDictionaryEventHandlerTest.java

@@ -0,0 +1,237 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.Collections;
+import java.util.Map;
+import java.util.Set;
+import java.util.UUID;
+import java.util.concurrent.ConcurrentHashMap;
+import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicReference;
+
+import org.junit.After;
+import org.junit.Before;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import org.apache.cassandra.SchemaLoader;
+import org.apache.cassandra.ServerTestUtils;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.db.DecoratedKey;
+import org.apache.cassandra.db.Keyspace;
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import org.apache.cassandra.locator.InetAddressAndPort;
+import org.apache.cassandra.net.MessagingService;
+import org.apache.cassandra.net.Verb;
+import org.apache.cassandra.schema.CompressionParams;
+import org.apache.cassandra.schema.KeyspaceParams;
+import org.apache.cassandra.schema.Schema;
+import org.apache.cassandra.schema.TableId;
+import org.apache.cassandra.schema.TableMetadata;
+import org.apache.cassandra.tcm.membership.NodeAddresses;
+import org.apache.cassandra.tcm.membership.NodeId;
+import org.apache.cassandra.tcm.transformations.Register;
+import org.apache.cassandra.tcm.transformations.UnsafeJoin;
+import org.apache.cassandra.utils.ByteBufferUtil;
+import org.apache.cassandra.utils.FBUtilities;
+
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatNoException;
+
+public class CompressionDictionaryEventHandlerTest
+{
+    private static final String TEST_NAME = "compression_dict_event_handler_test_";
+    private static final String KEYSPACE = TEST_NAME + "keyspace";
+    private static final String TABLE = "test_table";
+    private static final DictId TEST_DICTIONARY_ID = new DictId(Kind.ZSTD, 12345L);
+
+    private static TableMetadata tableMetadata;
+    private static ColumnFamilyStore cfs;
+
+    private CompressionDictionaryEventHandler eventHandler;
+    private ZstdCompressionDictionary testDictionary;
+
+    @BeforeClass
+    public static void setUpClass() throws Exception
+    {
+        ServerTestUtils.prepareServerNoRegister();
+
+        // Create a table with dictionary compression enabled
+        CompressionParams compressionParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                     Map.of("compression_level", "3"));
+
+        TableMetadata.Builder tableBuilder = TableMetadata.builder(KEYSPACE, TABLE)
+                                                          .addPartitionKeyColumn("pk", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                          .addRegularColumn("data", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                          .compression(compressionParams);
+
+        SchemaLoader.createKeyspace(KEYSPACE,
+                                    KeyspaceParams.simple(1),
+                                    tableBuilder);
+
+        tableMetadata = Schema.instance.getTableMetadata(KEYSPACE, TABLE);
+        cfs = Keyspace.open(KEYSPACE).getColumnFamilyStore(TABLE);
+
+        // Register some nodes for cluster testing
+        InetAddressAndPort ep1 = InetAddressAndPort.getByName("127.0.0.2:9042");
+        InetAddressAndPort ep2 = InetAddressAndPort.getByName("127.0.0.3:9042");
+        InetAddressAndPort ep3 = FBUtilities.getBroadcastAddressAndPort();
+
+        NodeId node1 = Register.register(new NodeAddresses(UUID.randomUUID(), ep1, ep1, ep1));
+        NodeId node2 = Register.register(new NodeAddresses(UUID.randomUUID(), ep2, ep2, ep2));
+        NodeId node3 = Register.register(new NodeAddresses(UUID.randomUUID(), ep3, ep3, ep3));
+
+        // Simple token distribution for testing
+        UnsafeJoin.unsafeJoin(node1, Collections.singleton(key(tableMetadata, 1).getToken()));
+        UnsafeJoin.unsafeJoin(node2, Collections.singleton(key(tableMetadata, 2).getToken()));
+        UnsafeJoin.unsafeJoin(node3, Collections.singleton(key(tableMetadata, 3).getToken()));
+    }
+
+    @Before
+    public void setUp()
+    {
+        MessagingService.instance().inboundSink.clear();
+        MessagingService.instance().outboundSink.clear();
+        testDictionary = createTestDictionary();
+        eventHandler = new CompressionDictionaryEventHandler(cfs, new CompressionDictionaryCache());
+    }
+
+    @After
+    public void tearDown()
+    {
+        if (testDictionary != null)
+        {
+            testDictionary.close();
+        }
+        MessagingService.instance().inboundSink.clear();
+        MessagingService.instance().outboundSink.clear();
+    }
+
+    @Test
+    public void testOnNewDictionaryTrained() throws InterruptedException
+    {
+        // Expect messages to 2 other nodes (excluding self)
+        CountDownLatch messageSentLatch = new CountDownLatch(2);
+        Set<InetAddressAndPort> receivers = ConcurrentHashMap.newKeySet(2);
+        AtomicReference<CompressionDictionaryUpdateMessage> capturedMessage = new AtomicReference<>();
+
+        // Capture outbound messages
+        MessagingService.instance().outboundSink.add((message, to) -> {
+            if (message.verb() == Verb.DICTIONARY_UPDATE_REQ)
+            {
+                capturedMessage.set((CompressionDictionaryUpdateMessage) message.payload);
+                receivers.add(to);
+                messageSentLatch.countDown();
+            }
+            return false; // Don't actually send
+        });
+
+        eventHandler.onNewDictionaryTrained(TEST_DICTIONARY_ID);
+
+        // Wait for message to be processed
+        assertThat(messageSentLatch.await(5, TimeUnit.SECONDS))
+        .as("Dictionary update notification should be sent")
+        .isTrue();
+
+        assertThat(receivers)
+        .as("Should not send notification to self")
+        .hasSize(2)
+        .doesNotContain(FBUtilities.getBroadcastAddressAndPort());
+
+        CompressionDictionaryUpdateMessage message = capturedMessage.get();
+        assertThat(message)
+        .as("Message should be captured")
+        .isNotNull();
+        assertThat(message.tableId)
+        .as("Message should contain correct table ID")
+        .isEqualTo(tableMetadata.id);
+        assertThat(message.dictionaryId)
+        .as("Message should contain correct dictionary ID")
+        .isEqualTo(TEST_DICTIONARY_ID);
+    }
+
+    @Test
+    public void testMessageSerialization()
+    {
+        TableId testTableId = tableMetadata.id;
+        CompressionDictionaryUpdateMessage message = new CompressionDictionaryUpdateMessage(testTableId, TEST_DICTIONARY_ID);
+
+        assertThat(message.tableId)
+        .as("Message should contain correct table ID")
+        .isEqualTo(testTableId);
+        assertThat(message.dictionaryId)
+        .as("Message should contain correct dictionary ID")
+        .isEqualTo(TEST_DICTIONARY_ID);
+        assertThat(CompressionDictionaryUpdateMessage.serializer)
+        .as("Message should have serializer")
+        .isNotNull();
+    }
+
+    @Test
+    public void testMessageSerializationRoundTrip() throws Exception
+    {
+        TableId testTableId = tableMetadata.id;
+        CompressionDictionaryUpdateMessage originalMessage = new CompressionDictionaryUpdateMessage(testTableId, TEST_DICTIONARY_ID);
+
+        // Serialize
+        org.apache.cassandra.io.util.DataOutputBuffer out = new org.apache.cassandra.io.util.DataOutputBuffer();
+        CompressionDictionaryUpdateMessage.serializer.serialize(originalMessage, out, MessagingService.current_version);
+
+        // Deserialize
+        org.apache.cassandra.io.util.DataInputBuffer in = new org.apache.cassandra.io.util.DataInputBuffer(out.getData());
+        CompressionDictionaryUpdateMessage deserializedMessage =
+        CompressionDictionaryUpdateMessage.serializer.deserialize(in, MessagingService.current_version);
+
+        assertThat(deserializedMessage.tableId)
+        .as("Deserialized table ID should match")
+        .isEqualTo(originalMessage.tableId);
+        assertThat(deserializedMessage.dictionaryId)
+        .as("Deserialized dictionary ID should match")
+        .isEqualTo(originalMessage.dictionaryId);
+    }
+
+    @Test
+    public void testSendNotificationRobustness()
+    {
+        // Test that sending notifications doesn't throw even if messaging fails
+        MessagingService.instance().outboundSink.add((message, to) -> {
+            if (message.verb() == Verb.DICTIONARY_UPDATE_REQ)
+            {
+                throw new RuntimeException("Simulated messaging failure");
+            }
+            return false;
+        });
+
+        assertThatNoException().isThrownBy(() -> eventHandler.onNewDictionaryTrained(TEST_DICTIONARY_ID));
+    }
+
+    private static ZstdCompressionDictionary createTestDictionary()
+    {
+        byte[] dictBytes = "test dictionary data for event handler testing".getBytes();
+        return new ZstdCompressionDictionary(TEST_DICTIONARY_ID, dictBytes);
+    }
+
+    private static DecoratedKey key(TableMetadata metadata, int key)
+    {
+        return metadata.partitioner.decorateKey(ByteBufferUtil.bytes(key));
+    }
+}

test/unit/org/apache/cassandra/db/compression/CompressionDictionaryIntegrationTest.java

@@ -0,0 +1,224 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.Collections;
+import java.util.concurrent.TimeUnit;
+
+import org.junit.Before;
+import org.junit.Test;
+
+import org.apache.cassandra.config.Config;
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.config.DataStorageSpec;
+import org.apache.cassandra.cql3.CQLTester;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.db.Keyspace;
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import org.apache.cassandra.db.compression.ICompressionDictionaryTrainer.TrainingStatus;
+import org.apache.cassandra.schema.CompressionParams;
+import org.apache.cassandra.utils.Clock;
+
+import static org.apache.cassandra.Util.spinUntilTrue;
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatNoException;
+import static org.assertj.core.api.Assertions.assertThatThrownBy;
+
+public class CompressionDictionaryIntegrationTest extends CQLTester
+{
+    private static final String REPEATED_DATA = "The quick brown fox jumps over the lazy dog. This text repeats for better compression. ";
+
+    @Before
+    public void configureDatabaseDescriptor()
+    {
+        Config config = DatabaseDescriptor.getRawConfig();
+        config.compression_dictionary_training_sampling_rate = 1.0f;
+        config.compression_dictionary_training_max_total_sample_size = new DataStorageSpec.IntKibibytesBound("128KiB");
+        config.compression_dictionary_training_max_dictionary_size = new DataStorageSpec.IntKibibytesBound("10KiB");
+        config.flush_compression = Config.FlushCompression.table;
+        DatabaseDescriptor.setConfig(config);
+    }
+
+    @Test
+    public void testEnableDisableDictionaryCompression()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+        CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+
+        // Insert data and flush to create SSTables
+        for (int i = 0; i < 100; i++)
+        {
+            execute("INSERT INTO %s (id, data) VALUES (?, ?)", i, REPEATED_DATA + " " + i);
+        }
+        flush();
+
+        assertThatNoException()
+        .as("Should allow manual training")
+        .isThrownBy(() -> manager.train(false));
+
+        // Disable dictionary compression
+        CompressionParams nonDictParams = CompressionParams.lz4();
+        manager.maybeReloadFromSchema(nonDictParams);
+
+        assertThatThrownBy(() -> manager.train(false))
+        .as("Should disallow manual training when using lz4")
+        .isInstanceOf(UnsupportedOperationException.class)
+        .hasMessageContaining("does not support dictionary compression");
+
+        // Re-enable dictionary compression
+        CompressionParams dictParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                              Collections.singletonMap("compression_level", "3"));
+        manager.maybeReloadFromSchema(dictParams);
+
+        // Insert more data for the re-enabled compression
+        for (int i = 100; i < 200; i++)
+        {
+            execute("INSERT INTO %s (id, data) VALUES (?, ?)", i, REPEATED_DATA + " " + i);
+        }
+        flush();
+
+        assertThatNoException()
+        .as("Should allow manual training after switching back to dictionary compression")
+        .isThrownBy(() -> manager.train(false));
+    }
+
+    @Test
+    public void testCompressionParameterChanges()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+        CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+        ICompressionDictionaryTrainer trainer = manager.trainer();
+        assertThat(trainer).isNotNull();
+        assertThat(trainer.kind()).isEqualTo(Kind.ZSTD);
+
+        // Change compression level - should create new trainer
+        CompressionParams newParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                             Collections.singletonMap("compression_level", "5"));
+        manager.maybeReloadFromSchema(newParams);
+        ICompressionDictionaryTrainer newTrainer = manager.trainer();
+        assertThat(newTrainer.kind()).isEqualTo(Kind.ZSTD);
+        assertThat(newTrainer)
+        .as("Should create a different trainer instance when compression level is changed")
+        .isNotSameAs(trainer);
+    }
+
+    @Test
+    public void testResourceCleanupOnClose() throws Exception
+    {
+        createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+        ColumnFamilyStore cfs = getCurrentColumnFamilyStore();
+        CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+
+        // Add test dictionary
+        ZstdCompressionDictionary testDict = createTestDictionary();
+        manager.add(testDict);
+
+        assertThat(testDict.selfRef().globalCount())
+        .as("Dictionary's reference count should be 1 after adding to cache")
+        .isOne();
+
+        assertThat(manager.getCurrent())
+        .as("Should have current dictionary before close")
+        .isNotNull();
+
+        manager.close();
+
+        assertThat(manager.trainer()).isNull();
+        // Dictionary's reference count should be 0 after closing manager
+        // Dictionary is closed in a separate thread. Wait a bit for the reference count to be updated.
+        spinUntilTrue(() -> testDict.selfRef().globalCount() == 0, 1, TimeUnit.SECONDS);
+        assertThat(testDict.rawDictionary())
+        .as("The raw dictionary bytes should still be accessible")
+        .isNotNull();
+    }
+
+    private static ZstdCompressionDictionary createTestDictionary()
+    {
+        byte[] dictBytes = (REPEATED_DATA + " dictionary training data").getBytes();
+        DictId dictId = new DictId(Kind.ZSTD, Clock.Global.currentTimeMillis());
+        return new ZstdCompressionDictionary(dictId, dictBytes);
+    }
+
+    @Test
+    public void testSSTableBasedTraining()
+    {
+        DatabaseDescriptor.setFlushCompression(Config.FlushCompression.table);
+        String table = createTable("CREATE TABLE %s (pk text PRIMARY KEY, data text) " +
+                                   "WITH compression = {'class': 'ZstdDictionaryCompressor', 'chunk_length_in_kb' : 4}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+        CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+
+        // Insert compressible data and flush to create SSTables
+        for (int i = 0; i < 1000; i++)
+        {
+            execute("INSERT INTO %s (pk, data) VALUES (?, ?)",
+                    "key" + i,
+                    REPEATED_DATA + " row " + i);
+            if (i % 200 == 0)
+                flush();
+        }
+        flush();
+
+        // Verify we have SSTables
+        assertThat(cfs.getLiveSSTables())
+        .as("Should have created SSTables")
+        .hasSizeGreaterThan(0);
+
+        // Train from existing SSTables
+        manager.train(true);
+
+        // Training should complete quickly since we're reading from existing SSTables
+        spinUntilTrue(() -> TrainingState.fromCompositeData(manager.getTrainingState()).status == TrainingStatus.COMPLETED, 10);
+
+        // Verify dictionary was trained and is available
+        spinUntilTrue(() -> manager.getCurrent() != null, 2);
+
+        CompressionDictionary currentDict = manager.getCurrent();
+        assertThat(currentDict).isNotNull();
+        assertThat(currentDict.kind())
+        .as("Dictionary should be ZSTD type")
+        .isEqualTo(Kind.ZSTD);
+
+        assertThat(currentDict.rawDictionary().length)
+        .as("Dictionary should have content")
+        .isGreaterThan(0);
+
+        // Verify we can still read the data
+        assertRows(execute("SELECT pk, data FROM %s WHERE pk = ?", "key0"),
+                   row("key0", REPEATED_DATA + " row 0"));
+    }
+
+    @Test
+    public void testSSTableBasedTrainingWithoutSSTables()
+    {
+        String table = createTable("CREATE TABLE %s (pk text PRIMARY KEY, data text) " +
+                                   "WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+        CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+
+        // Try to train without any SSTables
+        assertThatThrownBy(() -> manager.train(false))
+        .as("Should fail when no SSTables are available")
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("No SSTables available for training");
+    }
+}

test/unit/org/apache/cassandra/db/compression/CompressionDictionaryManagerMBeanTest.java

@@ -0,0 +1,156 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.Map;
+
+import org.junit.Before;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import org.apache.cassandra.SchemaLoader;
+import org.apache.cassandra.ServerTestUtils;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.db.Keyspace;
+import org.apache.cassandra.schema.CompressionParams;
+import org.apache.cassandra.schema.KeyspaceParams;
+import org.apache.cassandra.schema.TableMetadata;
+import org.apache.cassandra.utils.MBeanWrapper;
+import org.apache.cassandra.utils.MBeanWrapper.OnException;
+
+import static org.assertj.core.api.Assertions.assertThat;
+
+public class CompressionDictionaryManagerMBeanTest
+{
+    private static final String KEYSPACE_WITH_DICT = "keyspace_mbean_test";
+    private static final String TABLE = "test_table";
+
+    private static ColumnFamilyStore cfsWithDict;
+
+    @BeforeClass
+    public static void setUpClass() throws Exception
+    {
+        ServerTestUtils.prepareServer();
+        CompressionParams compressionParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                     Map.of("compression_level", "3"));
+        TableMetadata.Builder tableBuilder = TableMetadata.builder(KEYSPACE_WITH_DICT, TABLE)
+                                                          .addPartitionKeyColumn("pk", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                          .addRegularColumn("data", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                          .compression(compressionParams);
+        SchemaLoader.createKeyspace(KEYSPACE_WITH_DICT,
+                                    KeyspaceParams.simple(1),
+                                    tableBuilder);
+        cfsWithDict = Keyspace.open(KEYSPACE_WITH_DICT).getColumnFamilyStore(TABLE);
+    }
+
+    // Ensure no mbean is registered at the begining of the test
+    @Before
+    public void cleanup()
+    {
+        String mbeanName = CompressionDictionaryManager.mbeanName(KEYSPACE_WITH_DICT, TABLE);
+        MBeanWrapper.instance.unregisterMBean(mbeanName, OnException.IGNORE);
+    }
+
+    @Test
+    public void testMBeanRegisteredWhenBookkeepingEnabled()
+    {
+        String mbeanName = CompressionDictionaryManager.mbeanName(KEYSPACE_WITH_DICT, TABLE);
+        // Create manager with bookkeeping enabled
+        try (CompressionDictionaryManager manager = new CompressionDictionaryManager(cfsWithDict, true))
+        {
+            // Verify MBean is registered
+            assertThat(MBeanWrapper.instance.isRegistered(mbeanName))
+            .as("MBean should be registered when bookkeeping is enabled")
+            .isTrue();
+        }
+        // Closing manager should unregister the mbean; Verify it is unregistered
+        assertThat(MBeanWrapper.instance.isRegistered(mbeanName))
+        .as("MBean should be unregistered after unregisterMbean() call")
+        .isFalse();
+    }
+
+    @Test
+    public void testMBeanNotRegisteredWhenBookkeepingDisabled()
+    {
+        // Create manager with bookkeeping disabled
+        try (CompressionDictionaryManager manager = new CompressionDictionaryManager(cfsWithDict, false))
+        {
+            // Verify MBean is NOT registered
+            String mbeanName = CompressionDictionaryManager.mbeanName(KEYSPACE_WITH_DICT, TABLE);;
+            assertThat(MBeanWrapper.instance.isRegistered(mbeanName))
+            .as("MBean should not be registered when bookkeeping is disabled")
+            .isFalse();
+        }
+        // Closing manager should not throw due to mbean not registered
+    }
+
+    @Test
+    public void testMBeanUnregisteredOnCFSInvalidation()
+    {
+        String testKeyspace = "test_invalidation_mbean_ks";
+        String testTable = "test_invalidation_mbean_table";
+
+        CompressionParams compressionParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                     Map.of("compression_level", "3"));
+
+        TableMetadata.Builder tableBuilder = TableMetadata.builder(testKeyspace, testTable)
+                                                          .addPartitionKeyColumn("pk", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                          .addRegularColumn("data", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                          .compression(compressionParams);
+
+        SchemaLoader.createKeyspace(testKeyspace,
+                                    KeyspaceParams.simple(1),
+                                    tableBuilder);
+
+        ColumnFamilyStore cfs = Keyspace.open(testKeyspace).getColumnFamilyStore(testTable);
+
+        String mbeanName = CompressionDictionaryManager.mbeanName(testKeyspace, testTable);
+
+        // Verify MBean is registered (CFS registers it during creation)
+        assertThat(MBeanWrapper.instance.isRegistered(mbeanName))
+        .as("MBean should be registered after CFS creation")
+        .isTrue();
+
+        // Invalidate the CFS (which should unregister the MBean)
+        cfs.invalidate(true, true);
+
+        // Verify MBean is unregistered
+        assertThat(MBeanWrapper.instance.isRegistered(mbeanName))
+        .as("MBean should be unregistered after CFS invalidation")
+        .isFalse();
+    }
+
+    @Test
+    public void testMBeanStatisticsMethods()
+    {
+        // Create manager with bookkeeping enabled
+        try (CompressionDictionaryManager manager = new CompressionDictionaryManager(cfsWithDict, true))
+        {
+            TrainingState state = TrainingState.fromCompositeData(manager.getTrainingState());
+            // Test statistics methods directly on the manager (which implements the MBean interface)
+            assertThat(state.getSampleCount())
+            .as("Sample count should be non-negative")
+            .isGreaterThanOrEqualTo(0);
+
+            assertThat(state.getTotalSampleSize())
+            .as("Total sample size should be non-negative")
+            .isGreaterThanOrEqualTo(0);
+        }
+    }
+}

test/unit/org/apache/cassandra/db/compression/CompressionDictionaryManagerTest.java

@@ -0,0 +1,258 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.nio.ByteBuffer;
+import java.util.Map;
+
+import org.junit.After;
+import org.junit.Before;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import org.apache.cassandra.SchemaLoader;
+import org.apache.cassandra.ServerTestUtils;
+import org.apache.cassandra.config.CassandraRelevantProperties;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.db.Keyspace;
+import org.apache.cassandra.db.compression.ICompressionDictionaryTrainer.TrainingStatus;
+import org.apache.cassandra.schema.CompressionParams;
+import org.apache.cassandra.schema.KeyspaceParams;
+import org.apache.cassandra.schema.TableMetadata;
+
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatNoException;
+import static org.assertj.core.api.Assertions.assertThatThrownBy;
+
+public class CompressionDictionaryManagerTest
+{
+    private static final String KEYSPACE_WITH_DICT = "keyspace_with_dict";
+    private static final String KEYSPACE_WITHOUT_DICT = "keyspace_without_dict";
+    private static final String TABLE = "test_table";
+
+    private static ColumnFamilyStore cfsWithDict;
+    private static ColumnFamilyStore cfsWithoutDict;
+
+    private CompressionDictionaryManager managerWithDict;
+    private CompressionDictionaryManager managerWithoutDict;
+
+    @BeforeClass
+    public static void setUpClass() throws Exception
+    {
+        CassandraRelevantProperties.ORG_APACHE_CASSANDRA_DISABLE_MBEAN_REGISTRATION.setBoolean(true);
+        ServerTestUtils.prepareServerNoRegister();
+
+        // Create table with dictionary compression enabled
+        CompressionParams compressionParamsWithDict = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                             Map.of("compression_level", "3"));
+
+        TableMetadata.Builder tableBuilderWithDict = TableMetadata.builder(KEYSPACE_WITH_DICT, TABLE)
+                                                                  .addPartitionKeyColumn("pk", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                                  .addRegularColumn("data", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                                  .compression(compressionParamsWithDict);
+
+        // Create table without dictionary compression
+        CompressionParams compressionParamsWithoutDict = CompressionParams.lz4();
+
+        TableMetadata.Builder tableBuilderWithoutDict = TableMetadata.builder(KEYSPACE_WITHOUT_DICT, TABLE)
+                                                                     .addPartitionKeyColumn("pk", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                                     .addRegularColumn("data", org.apache.cassandra.db.marshal.UTF8Type.instance)
+                                                                     .compression(compressionParamsWithoutDict);
+
+        SchemaLoader.createKeyspace(KEYSPACE_WITH_DICT,
+                                    KeyspaceParams.simple(1),
+                                    tableBuilderWithDict);
+
+        SchemaLoader.createKeyspace(KEYSPACE_WITHOUT_DICT,
+                                    KeyspaceParams.simple(1),
+                                    tableBuilderWithoutDict);
+
+        cfsWithDict = Keyspace.open(KEYSPACE_WITH_DICT).getColumnFamilyStore(TABLE);
+        cfsWithoutDict = Keyspace.open(KEYSPACE_WITHOUT_DICT).getColumnFamilyStore(TABLE);
+    }
+
+    @Before
+    public void setUp()
+    {
+        managerWithDict = new CompressionDictionaryManager(cfsWithDict, true);
+        managerWithoutDict = new CompressionDictionaryManager(cfsWithoutDict, true);
+    }
+
+    @After
+    public void tearDown() throws Exception
+    {
+        if (managerWithDict != null)
+        {
+            managerWithDict.close();
+        }
+        if (managerWithoutDict != null)
+        {
+            managerWithoutDict.close();
+        }
+    }
+
+    @Test
+    public void testManagerInitializationWithDictionaryCompression()
+    {
+        assertThat(managerWithDict)
+        .as("Manager should be created successfully for dictionary-enabled table")
+        .isNotNull();
+
+        // Manager should start in a valid state
+        TrainingState trainingState = TrainingState.fromCompositeData(managerWithDict.getTrainingState());
+        assertThat(trainingState.getStatus())
+        .as("Training status should be valid")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+    }
+
+    @Test
+    public void testManagerInitializationWithoutDictionaryCompression()
+    {
+        assertThat(managerWithoutDict)
+        .as("Manager should be created successfully for non-dictionary table")
+        .isNotNull();
+
+        // Should report NOT_STARTED since no trainer is created
+        TrainingState trainingState = TrainingState.fromCompositeData(managerWithoutDict.getTrainingState());
+        assertThat(trainingState.getStatus())
+        .as("Should report NOT_STARTED for non-dictionary tables")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+    }
+
+    @Test
+    public void testMaybeReloadFromSchemaEnableDictionaryCompression()
+    {
+        // Start with manager for non-dictionary table
+        TrainingState initialTrainingState = TrainingState.fromCompositeData(managerWithoutDict.getTrainingState());
+        assertThat(initialTrainingState.getStatus())
+        .as("Initially should not be training")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+
+        // Enable dictionary compression by switching to dict params
+        CompressionParams dictParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                              Map.of("compression_level", "3"));
+
+        managerWithoutDict.maybeReloadFromSchema(dictParams);
+
+        // Should now have a trainer
+        assertThat(managerWithoutDict.trainer())
+        .as("Should have a trainer after enabling dictionary compression")
+        .isNotNull();
+    }
+
+    @Test
+    public void testMaybeReloadFromSchemaDisableDictionaryCompression()
+    {
+        // Verify we have a trainer initially
+        assertThat(managerWithDict.trainer()).isNotNull();
+
+        // Disable dictionary compression
+        CompressionParams nonDictParams = CompressionParams.lz4();
+        managerWithDict.maybeReloadFromSchema(nonDictParams);
+
+        // Should disable training
+        assertThat(managerWithDict.trainer())
+        .as("Should not have trainer when dictionary compression is disabled")
+        .isNull();
+    }
+
+    @Test
+    public void testTrainerCompatibilityCheck()
+    {
+        ICompressionDictionaryTrainer initialTrainer = managerWithDict.trainer();
+        assertThat(initialTrainer).isNotNull();
+
+        // Change compression level - should create new trainer
+        CompressionParams differentLevelParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                        Map.of("compression_level", "5"));
+        managerWithDict.maybeReloadFromSchema(differentLevelParams);
+        ICompressionDictionaryTrainer newTrainer = managerWithDict.trainer();
+
+        // Should have a different trainer instance
+        assertThat(newTrainer)
+        .as("Should create new trainer when compression level changes")
+        .isNotSameAs(initialTrainer);
+    }
+
+    @Test
+    public void testAddSample()
+    {
+        ByteBuffer sample = ByteBuffer.wrap("test sample data".getBytes());
+        ByteBuffer emptyBuffer = ByteBuffer.allocate(0);
+
+        // Should not throw for dictionary-enabled table
+        assertThatNoException().isThrownBy(() -> managerWithDict.addSample(sample));
+        assertThatNoException().isThrownBy(() -> managerWithDict.addSample(null));
+        assertThatNoException().isThrownBy(() -> managerWithDict.addSample(emptyBuffer));
+
+        // Should not throw for non-dictionary table (graceful handling)
+        assertThatNoException().isThrownBy(() -> managerWithoutDict.addSample(sample));
+        assertThatNoException().isThrownBy(() -> managerWithoutDict.addSample(null));
+        assertThatNoException().isThrownBy(() -> managerWithoutDict.addSample(emptyBuffer));
+    }
+
+    @Test
+    public void testTrainManualWithNonDictionaryTable()
+    {
+        assertThatThrownBy(() -> managerWithoutDict.train(false))
+        .isInstanceOf(UnsupportedOperationException.class)
+        .hasMessageContaining("does not support dictionary compression");
+    }
+
+    @Test
+    public void testTrainManualWithDictionaryTable()
+    {
+        // Should throw because no SSTables exist
+        assertThatThrownBy(() -> managerWithDict.train(false))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("No SSTables available for training");
+    }
+
+    @Test
+    public void testSchemaChangeWorkflow()
+    {
+        // Start with non-dictionary table
+        TrainingState initialTrainingState = TrainingState.fromCompositeData(managerWithoutDict.getTrainingState());
+        assertThat(initialTrainingState.getStatus()).isEqualTo(TrainingStatus.NOT_STARTED);
+        assertThat(managerWithoutDict.trainer()).isNull();
+
+        // Enable dictionary compression
+        CompressionParams dictParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                              Map.of("compression_level", "3"));
+        managerWithoutDict.maybeReloadFromSchema(dictParams);
+
+        // Should now support training
+        assertThat(managerWithoutDict.trainer()).isNotNull();
+
+        // Change compression level
+        CompressionParams newDictParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                 Map.of("compression_level", "5"));
+        managerWithoutDict.maybeReloadFromSchema(newDictParams);
+
+        // Should still support training with new parameters
+        assertThat(managerWithoutDict.trainer()).isNotNull();
+
+        // Disable dictionary compression
+        CompressionParams nonDictParams = CompressionParams.lz4();
+        managerWithoutDict.maybeReloadFromSchema(nonDictParams);
+
+        // Should disable training
+        assertThat(managerWithoutDict.trainer()).isNull();
+    }
+}

test/unit/org/apache/cassandra/db/compression/CompressionDictionarySchedulerTest.java

@@ -0,0 +1,128 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.HashSet;
+import java.util.Set;
+
+import org.junit.After;
+import org.junit.Before;
+import org.junit.Test;
+
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.cql3.CQLTester;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.db.Keyspace;
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+
+import static org.apache.cassandra.Util.spinUntilTrue;
+import static org.assertj.core.api.Assertions.assertThat;
+
+public class CompressionDictionarySchedulerTest extends CQLTester
+{
+    private CompressionDictionaryScheduler scheduler;
+    private ICompressionDictionaryCache cache;
+
+    @Before
+    public void setUp()
+    {
+        cache = new CompressionDictionaryCache();
+        // Disable compaction to make the sstable sampling deterministic; to avoid excluding sstables get compacted away.
+        disableCompaction(KEYSPACE);
+    }
+
+    @After
+    public void tearDown()
+    {
+        if (scheduler != null)
+        {
+            scheduler.close();
+        }
+    }
+
+    @Test
+    public void testScheduleSSTableBasedTrainingWithNoSSTables()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) " +
+                                   "WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+        scheduler = new CompressionDictionaryScheduler(KEYSPACE, table, cache, true);
+
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+        CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+
+        Set<SSTableReader> sstables = new HashSet<>();
+        CompressionDictionaryTrainingConfig config = createSampleAllTrainingConfig(cfs);
+
+        // Should not throw, but task will complete quickly with no SSTables
+        scheduler.scheduleSSTableBasedTraining(manager.trainer(), sstables, config, true);
+        spinUntilTrue(() -> !scheduler.isManualTrainingRunning());
+        assertThat(manager.getCurrent()).isNull();
+    }
+
+    @Test
+    public void testScheduleSSTableBasedTrainingWithSSTables()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) " +
+                                   "WITH compression = {'class': 'ZstdDictionaryCompressor', 'chunk_length_in_kb': '4'}");
+        scheduler = new CompressionDictionaryScheduler(KEYSPACE, table, cache, true);
+
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+        CompressionDictionaryManager manager = cfs.compressionDictionaryManager();
+
+        createSSTables();
+
+        Set<SSTableReader> sstables = cfs.getLiveSSTables();
+        assertThat(sstables).isNotEmpty();
+
+        CompressionDictionaryTrainingConfig config = createSampleAllTrainingConfig(cfs);
+        manager.trainer().start(true);
+
+        assertThat(manager.getCurrent()).as("There should be no dictionary at this step").isNull();
+        scheduler.scheduleSSTableBasedTraining(manager.trainer(), sstables, config, true);
+
+        // Task should be scheduled
+        assertThat(scheduler.isManualTrainingRunning()).isTrue();
+        // A dictionary should be trained
+        spinUntilTrue(() -> manager.getCurrent() != null);
+    }
+
+    private void createSSTables()
+    {
+        for (int file = 0; file < 10; file++)
+        {
+            int batchSize = 1000;
+            for (int i = 0; i < batchSize; i++)
+            {
+                int index = i + file * batchSize;
+                execute("INSERT INTO %s (id, data) VALUES (?, ?)", index, "test data " + index);
+            }
+            flush();
+        }
+    }
+
+    private static CompressionDictionaryTrainingConfig createSampleAllTrainingConfig(ColumnFamilyStore cfs) {
+        return CompressionDictionaryTrainingConfig
+               .builder()
+               .maxDictionarySize(DatabaseDescriptor.getCompressionDictionaryTrainingMaxDictionarySize())
+               .maxTotalSampleSize(DatabaseDescriptor.getCompressionDictionaryTrainingMaxTotalSampleSize())
+               .samplingRate(1.0f)
+               .chunkSize(cfs.metadata().params.compression.chunkLength())
+               .build();
+    }
+}

test/unit/org/apache/cassandra/db/compression/CompressionDictionaryTrainingConfigTest.java

@@ -0,0 +1,67 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import org.junit.Test;
+
+import static org.assertj.core.api.Assertions.assertThat;
+
+public class CompressionDictionaryTrainingConfigTest
+{
+    @Test
+    public void testBuilderDefaults()
+    {
+        CompressionDictionaryTrainingConfig config = CompressionDictionaryTrainingConfig.builder().build();
+
+        assertThat(config.maxDictionarySize)
+        .as("Default max dictionary size should be 64KB")
+        .isEqualTo(65536);
+        assertThat(config.maxTotalSampleSize)
+        .as("Default max total sample size should be 10MB")
+        .isEqualTo(10 * 1024 * 1024);
+        assertThat(config.samplingRate)
+        .as("Default sampling rate should be 100 (1%)")
+        .isEqualTo(100);
+    }
+
+    @Test
+    public void testCalculatedThresholds()
+    {
+        int dictSize = 16 * 1024; // 16KB
+        int sampleSize = 2 * 1024 * 1024; // 2MB
+        float samplingRate = 0.005f; // 0.5%
+
+        CompressionDictionaryTrainingConfig config = CompressionDictionaryTrainingConfig.builder()
+                                                                                        .maxDictionarySize(dictSize)
+                                                                                        .maxTotalSampleSize(sampleSize)
+                                                                                        .samplingRate(samplingRate)
+                                                                                        .build();
+
+        // Verify all calculated values are consistent
+        assertThat(config.maxDictionarySize).isEqualTo(dictSize);
+        assertThat(config.maxTotalSampleSize).isEqualTo(sampleSize);
+        assertThat(config.acceptableTotalSampleSize).isEqualTo(sampleSize / 10 * 8);
+        assertThat(config.samplingRate).isEqualTo(Math.round(1 / samplingRate));
+
+        // Verify relationship between max and acceptable sample sizes
+        assertThat(config.acceptableTotalSampleSize)
+        .as("Acceptable sample size should be less than or equal to max")
+        .isLessThanOrEqualTo(config.maxTotalSampleSize);
+    }
+}

test/unit/org/apache/cassandra/db/compression/SSTableChunkSamplerTest.java

@@ -0,0 +1,236 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.util.List;
+import java.util.Set;
+
+import org.junit.Test;
+
+import org.apache.cassandra.cql3.CQLTester;
+import org.apache.cassandra.db.ColumnFamilyStore;
+import org.apache.cassandra.db.Keyspace;
+import org.apache.cassandra.db.compression.SSTableChunkSampler.SSTableChunkInfo;
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatThrownBy;
+import static org.mockito.Mockito.RETURNS_DEEP_STUBS;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.when;
+
+public class SSTableChunkSamplerTest extends CQLTester
+{
+    @Test
+    public void testSSTableChunkInfoForCompressedSSTable()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'LZ4Compressor', 'chunk_length_in_kb': '64'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+
+        // Insert data and flush to create an SSTable
+        for (int i = 0; i < 100; i++)
+        {
+            execute("INSERT INTO %s (id, data) VALUES (?, ?)", i, "test data " + i);
+        }
+        flush();
+
+        Set<SSTableReader> sstables = cfs.getLiveSSTables();
+        assertThat(sstables).isNotEmpty();
+
+        SSTableReader sstable = sstables.iterator().next();
+        CompressionDictionaryTrainingConfig config = CompressionDictionaryTrainingConfig.builder()
+                                                                                        .chunkSize(64 * 1024)
+                                                                                        .build();
+
+        SSTableChunkInfo info = new SSTableChunkInfo(sstable, config);
+
+        assertThat(info.isCompressed).isTrue();
+        assertThat(info.chunkCount).isGreaterThan(0);
+        assertThat(info.dataLength).isGreaterThan(0);
+        assertThat(info.chunkSize).isEqualTo(64 * 1024);
+        assertThat(info.metadata).isNotNull();
+    }
+
+    @Test
+    public void testSSTableChunkInfoForUncompressedSSTable()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'enabled': 'false'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+
+        // Insert data and flush to create an uncompressed SSTable
+        for (int i = 0; i < 100; i++)
+        {
+            execute("INSERT INTO %s (id, data) VALUES (?, ?)", i, "test data " + i);
+        }
+        flush();
+
+        Set<SSTableReader> sstables = cfs.getLiveSSTables();
+        assertThat(sstables).isNotEmpty();
+
+        SSTableReader sstable = sstables.iterator().next();
+        CompressionDictionaryTrainingConfig config = CompressionDictionaryTrainingConfig.builder()
+                                                                                        .chunkSize(64 * 1024)
+                                                                                        .build();
+
+        SSTableChunkInfo info = new SSTableChunkInfo(sstable, config);
+
+        assertThat(info.isCompressed).isFalse();
+        assertThat(info.chunkCount).isGreaterThan(0);
+        assertThat(info.dataLength).isGreaterThan(0);
+        assertThat(info.chunkSize).isEqualTo(64 * 1024);
+        assertThat(info.metadata).isNull();
+    }
+
+    @Test
+    public void testCalculateTargetChunkCount()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'enabled': 'false'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+
+        // Create multiple SSTables
+        for (int batch = 0; batch < 3; batch++)
+        {
+            for (int i = 0; i < 100; i++)
+            {
+                execute("INSERT INTO %s (id, data) VALUES (?, ?)", batch * 100 + i, "test data " + i);
+            }
+            flush();
+        }
+
+        CompressionDictionaryTrainingConfig config = CompressionDictionaryTrainingConfig.builder()
+                                                                                        .maxTotalSampleSize(10 * 1024 * 1024) // 10MB
+                                                                                        .chunkSize(64 * 1024)
+                                                                                        .build();
+
+        Set<SSTableReader> sstables = cfs.getLiveSSTables();
+        assertThat(sstables).hasSizeGreaterThanOrEqualTo(3);
+
+        List<SSTableChunkInfo> sstableInfos = SSTableChunkSampler.buildSSTableInfos(sstables, config);
+        long totalChunks = sstableInfos.stream().mapToLong(info -> info.chunkCount).sum();
+        long targetChunkCount = SSTableChunkSampler.calculateTargetChunkCount(sstableInfos, totalChunks, config);
+
+        // Target should be based on maxTotalSampleSize divided by average chunk size
+        assertThat(targetChunkCount).isGreaterThan(0);
+        long totalDataSize = sstableInfos.stream().mapToLong(info -> info.dataLength).sum();
+        int averageChunkSize = (int) (totalDataSize / totalChunks);
+        long expectedTarget = config.maxTotalSampleSize / averageChunkSize;
+        assertThat(targetChunkCount).isEqualTo(expectedTarget);
+    }
+
+    @Test
+    public void testSelectRandomChunkIndices()
+    {
+        // test scenarios: select small portion, large portion and all
+        for (int expectedChunkCount : List.of(10, 80, 100))
+        {
+            Set<Long> selected = SSTableChunkSampler.selectRandomChunkIndices(100, expectedChunkCount);
+
+            assertThat(selected).hasSize(expectedChunkCount);
+            assertThat(selected).allMatch(idx -> idx >= 0 && idx < 100);
+        }
+    }
+
+    @Test
+    public void testSelectRandomChunkIndicesDistribution()
+    {
+        // Test that selection is reasonably distributed
+        int totalChunks = 100;
+        int runs = 1000;
+        int[] hitCount = new int[totalChunks];
+
+        // Run many selections and count how often each chunk is selected
+        for (int i = 0; i < runs; i++)
+        {
+            Set<Long> selected = SSTableChunkSampler.selectRandomChunkIndices(totalChunks, 10);
+            for (long idx : selected)
+            {
+                hitCount[(int) idx]++;
+            }
+        }
+
+        // Each chunk should be selected approximately 10% of the time (10 out of 100)
+        // So in 1000 runs, expect ~100 hits per chunk
+        // Allow for variance - between 50 and 150 hits
+        for (int count : hitCount)
+        {
+            assertThat(count).isBetween(50, 150);
+        }
+    }
+
+    @Test
+    public void testSampleFromSSTablesWithTrainerNotReady()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'LZ4Compressor'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+
+        // Insert data and flush to create an SSTable
+        for (int i = 0; i < 100; i++)
+        {
+            execute("INSERT INTO %s (id, data) VALUES (?, ?)", i, "test data " + i);
+        }
+        flush();
+
+        Set<SSTableReader> sstables = cfs.getLiveSSTables();
+        assertThat(sstables).isNotEmpty();
+
+        CompressionDictionaryTrainingConfig config = CompressionDictionaryTrainingConfig.builder()
+                                                                                        .chunkSize(64 * 1024)
+                                                                                        .build();
+
+        // Create a mock trainer that is not ready to sample
+        ICompressionDictionaryTrainer trainer = mock(ICompressionDictionaryTrainer.class, RETURNS_DEEP_STUBS);
+        when(trainer.shouldSample()).thenReturn(false);
+        when(trainer.getTrainingState().getStatus()).thenReturn(ICompressionDictionaryTrainer.TrainingStatus.NOT_STARTED);
+
+        // Should throw IllegalStateException when trainer is not ready
+        assertThatThrownBy(() -> SSTableChunkSampler.sampleFromSSTables(sstables, trainer, config))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Trainer is not ready to accept samples");
+    }
+
+    @Test
+    public void testReadChunkThrowsOnInvalidPosition()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'enabled': 'false'}");
+        ColumnFamilyStore cfs = Keyspace.open(keyspace()).getColumnFamilyStore(table);
+
+        // Insert data and flush to create an uncompressed SSTable
+        for (int i = 0; i < 100; i++)
+        {
+            execute("INSERT INTO %s (id, data) VALUES (?, ?)", i, "test data " + i);
+        }
+        flush();
+
+        Set<SSTableReader> sstables = cfs.getLiveSSTables();
+        assertThat(sstables).isNotEmpty();
+
+        SSTableReader sstable = sstables.iterator().next();
+        CompressionDictionaryTrainingConfig config = CompressionDictionaryTrainingConfig.builder()
+                                                                                        .chunkSize(64 * 1024)
+                                                                                        .build();
+
+        SSTableChunkInfo info = new SSTableChunkInfo(sstable, config);
+
+        // Try to read at a position beyond the data length - should throw IOException
+        long invalidPosition = info.dataLength + 1000;
+        assertThatThrownBy(() -> SSTableChunkSampler.readUncompressedChunk(info, invalidPosition))
+        .isInstanceOf(java.io.IOException.class)
+        .hasMessageContaining("Invalid read size");
+    }
+}

test/unit/org/apache/cassandra/db/compression/ZstdCompressionDictionaryTest.java

@@ -0,0 +1,392 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.io.ByteArrayInputStream;
+import java.io.ByteArrayOutputStream;
+import java.io.DataInputStream;
+import java.io.DataOutputStream;
+import java.io.IOException;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicInteger;
+
+import org.junit.Before;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import com.github.luben.zstd.ZstdDictCompress;
+import com.github.luben.zstd.ZstdDictDecompress;
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import org.apache.cassandra.io.compress.ZstdCompressorBase;
+import org.apache.cassandra.utils.concurrent.Ref;
+
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatThrownBy;
+
+public class ZstdCompressionDictionaryTest
+{
+    private static final byte[] SAMPLE_DICT_DATA = createSampleDictionaryData();
+    private static final DictId SAMPLE_DICT_ID = new DictId(Kind.ZSTD, 123456789L);
+
+    private ZstdCompressionDictionary dictionary;
+
+    @BeforeClass
+    public static void setUpClass()
+    {
+        DatabaseDescriptor.daemonInitialization();
+    }
+
+    @Before
+    public void setUp()
+    {
+        dictionary = new ZstdCompressionDictionary(SAMPLE_DICT_ID, SAMPLE_DICT_DATA);
+    }
+
+    @Test
+    public void testEqualsAndHashCode()
+    {
+        ZstdCompressionDictionary dictionary2 = new ZstdCompressionDictionary(SAMPLE_DICT_ID, SAMPLE_DICT_DATA);
+        ZstdCompressionDictionary differentIdDict = new ZstdCompressionDictionary(
+            new DictId(Kind.ZSTD, 987654321L), SAMPLE_DICT_DATA);
+
+        assertThat(dictionary)
+        .as("Dictionaries with same ID should be equal")
+        .isEqualTo(dictionary2);
+
+        assertThat(dictionary.hashCode())
+        .as("Hash codes should be equal for same ID")
+        .isEqualTo(dictionary2.hashCode());
+
+        assertThat(dictionary)
+        .as("Dictionaries with different IDs should not be equal")
+        .isNotEqualTo(differentIdDict);
+
+        dictionary2.close();
+        differentIdDict.close();
+    }
+
+    @Test
+    public void testDictionaryForCompression()
+    {
+        int compressionLevel = 3;
+        ZstdDictCompress compressDict = dictionary.dictionaryForCompression(compressionLevel);
+
+        assertThat(compressDict)
+        .as("Compression dictionary should not be null")
+        .isNotNull();
+
+        // Calling again should return the same cached instance
+        ZstdDictCompress compressDict2 = dictionary.dictionaryForCompression(compressionLevel);
+        assertThat(compressDict2)
+        .as("Second call should return cached instance")
+        .isSameAs(compressDict);
+    }
+
+    @Test
+    public void testDictionaryForCompressionMultipleLevels()
+    {
+        ZstdDictCompress level1 = dictionary.dictionaryForCompression(1);
+        ZstdDictCompress level3 = dictionary.dictionaryForCompression(3);
+        ZstdDictCompress level6 = dictionary.dictionaryForCompression(6);
+
+        assertThat(level1)
+        .as("Level 1 compression dictionary should not be null")
+        .isNotNull();
+
+        assertThat(level3)
+        .as("Level 3 compression dictionary should not be null")
+        .isNotNull();
+
+        assertThat(level6)
+        .as("Level 6 compression dictionary should not be null")
+        .isNotNull();
+
+        assertThat(level1)
+        .as("Different compression levels should have different instances")
+        .isNotSameAs(level3);
+
+        assertThat(level3)
+        .as("Different compression levels should have different instances")
+        .isNotSameAs(level6);
+    }
+
+    @Test
+    public void testDictionaryForDecompression()
+    {
+        ZstdDictDecompress decompressDict = dictionary.dictionaryForDecompression();
+
+        assertThat(decompressDict)
+        .as("Decompression dictionary should not be null")
+        .isNotNull();
+
+        ZstdDictDecompress decompressDict2 = dictionary.dictionaryForDecompression();
+        assertThat(decompressDict2)
+        .as("Second call should return cached instance")
+        .isSameAs(decompressDict);
+    }
+
+    @Test
+    public void testInvalidCompressionLevel()
+    {
+        // Test with various invalid compression levels
+        assertThatThrownBy(() -> dictionary.dictionaryForCompression(ZstdCompressorBase.FAST_COMPRESSION_LEVEL - 1))
+        .as("Negative compression level should throw exception")
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("is invalid");
+
+        assertThatThrownBy(() -> dictionary.dictionaryForCompression(100))
+        .as("Too high compression level should throw exception")
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("is invalid");
+    }
+
+    @Test
+    public void testDictionaryClose()
+    {
+        // Access some dictionaries first
+        dictionary.dictionaryForCompression(3);
+        dictionary.dictionaryForDecompression();
+
+        dictionary.close();
+
+        assertThatThrownBy(() -> dictionary.dictionaryForCompression(3))
+        .as("Should throw exception when accessing closed dictionary")
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Dictionary has been closed");
+
+        assertThatThrownBy(() -> dictionary.dictionaryForDecompression())
+        .as("Should throw exception when accessing closed dictionary")
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Dictionary has been closed");
+    }
+
+    @Test
+    public void testTryRef()
+    {
+        Ref<ZstdCompressionDictionary> ref = dictionary.tryRef();
+
+        assertThat(ref)
+        .as("tryRef should return non-null reference")
+        .isNotNull();
+
+        assertThat(ref.get())
+        .as("Reference should point to same dictionary")
+        .isSameAs(dictionary);
+
+        ref.release();
+    }
+
+    @Test
+    public void testMultipleReferences()
+    {
+        Ref<ZstdCompressionDictionary> ref1 = dictionary.ref();
+        Ref<ZstdCompressionDictionary> ref2 = dictionary.ref();
+        Ref<ZstdCompressionDictionary> ref3 = dictionary.tryRef();
+
+        assertThat(ref1.get())
+        .as("All references should point to same dictionary")
+        .isSameAs(dictionary);
+
+        assertThat(ref2.get())
+        .as("All references should point to same dictionary")
+        .isSameAs(dictionary);
+
+        assertThat(ref3.get())
+        .as("All references should point to same dictionary")
+        .isSameAs(dictionary);
+
+        // Dictionary should still be accessible
+        assertThat(dictionary.dictionaryForCompression(3))
+        .as("Dictionary should still be accessible with multiple refs")
+        .isNotNull();
+
+        ref1.release();
+        ref2.release();
+        ref3.release();
+    }
+
+    @Test
+    public void testReferenceAfterClose()
+    {
+        dictionary.close();
+
+        assertThatThrownBy(() -> dictionary.ref())
+        .as("Should not be able to get reference after close")
+        .isInstanceOf(AssertionError.class);
+
+        Ref<ZstdCompressionDictionary> tryRef = dictionary.tryRef();
+        assertThat(tryRef)
+        .as("tryRef should return null after close")
+        .isNull();
+    }
+
+    @Test
+    public void testConcurrentAccess() throws Exception
+    {
+        ExecutorService executor = Executors.newFixedThreadPool(4);
+        AtomicInteger successCount = new AtomicInteger(0);
+        int numTasks = 100;
+
+        try
+        {
+            Future<?>[] futures = new Future[numTasks];
+
+            for (int i = 0; i < numTasks; i++)
+            {
+                final int level = (i % 6) + 1; // Compression levels 1-6
+                futures[i] = executor.submit(() -> {
+                    try
+                    {
+                        Ref<ZstdCompressionDictionary> ref = dictionary.ref();
+                        ZstdDictCompress compressDict = ref.get().dictionaryForCompression(level);
+                        ZstdDictDecompress decompressDict = ref.get().dictionaryForDecompression();
+
+                        assertThat(compressDict).isNotNull();
+                        assertThat(decompressDict).isNotNull();
+
+                        successCount.incrementAndGet();
+                        ref.release();
+                    }
+                    catch (Exception e)
+                    {
+                        throw new RuntimeException(e);
+                    }
+                });
+            }
+
+            // Wait for all tasks to complete
+            for (Future<?> future : futures)
+            {
+                future.get(5, TimeUnit.SECONDS);
+            }
+
+            assertThat(successCount.get())
+            .as("All concurrent accesses should succeed")
+            .isEqualTo(numTasks);
+        }
+        finally
+        {
+            executor.shutdown();
+            executor.awaitTermination(5, TimeUnit.SECONDS);
+        }
+    }
+
+    @Test
+    public void testSerializeDeserialize() throws IOException
+    {
+        ByteArrayOutputStream baos = new ByteArrayOutputStream();
+        DataOutputStream dos = new DataOutputStream(baos);
+
+        dictionary.serialize(dos);
+        dos.flush();
+
+        byte[] serializedData = baos.toByteArray();
+        assertThat(serializedData.length)
+        .as("Serialized data should not be empty")
+        .isGreaterThan(0);
+
+        // Deserialize
+        ByteArrayInputStream bais = new ByteArrayInputStream(serializedData);
+        DataInputStream dis = new DataInputStream(bais);
+
+        CompressionDictionary deserializedDict = CompressionDictionary.deserialize(dis, null);
+
+        assertThat(deserializedDict)
+        .as("Deserialized dictionary should not be null")
+        .isNotNull();
+
+        assertThat(deserializedDict.dictId())
+        .as("Deserialized dictionary ID should match")
+        .isEqualTo(dictionary.dictId());
+
+        assertThat(deserializedDict.kind())
+        .as("Deserialized dictionary kind should match")
+        .isEqualTo(dictionary.kind());
+
+        assertThat(deserializedDict.rawDictionary())
+        .as("Deserialized dictionary data should match")
+        .isEqualTo(dictionary.rawDictionary());
+    }
+
+    @Test
+    public void testSerializeDeserializeWithManager() throws Exception
+    {
+        ByteArrayOutputStream baos = new ByteArrayOutputStream();
+        DataOutputStream dos = new DataOutputStream(baos);
+
+        dictionary.serialize(dos);
+        dos.flush();
+
+        byte[] serializedData = baos.toByteArray();
+
+        // First deserialization should create and cache the dictionary
+        ByteArrayInputStream bais1 = new ByteArrayInputStream(serializedData);
+        DataInputStream dis1 = new DataInputStream(bais1);
+        CompressionDictionary dict1 = CompressionDictionary.deserialize(dis1, null);
+
+        // Second deserialization should return cached instance
+        ByteArrayInputStream bais2 = new ByteArrayInputStream(serializedData);
+        DataInputStream dis2 = new DataInputStream(bais2);
+        CompressionDictionary dict2 = CompressionDictionary.deserialize(dis2, null);
+
+        assertThat(dict1)
+        .as("Both deserializations should return identical dictionary")
+        .isNotNull()
+        .isEqualTo(dict2);
+
+        dict1.close();
+        dict2.close();
+    }
+
+    @Test
+    public void testDeserializeCorruptedData() throws IOException
+    {
+        ByteArrayOutputStream baos = new ByteArrayOutputStream();
+        DataOutputStream dos = new DataOutputStream(baos);
+
+        // Write corrupted data (wrong checksum)
+        dos.writeByte(Kind.ZSTD.ordinal());
+        dos.writeLong(SAMPLE_DICT_ID.id);
+        dos.writeInt(SAMPLE_DICT_DATA.length);
+        dos.write(SAMPLE_DICT_DATA);
+        dos.writeInt(0xDEADBEEF); // Wrong checksum
+        dos.flush();
+
+        byte[] corruptedData = baos.toByteArray();
+        ByteArrayInputStream bais = new ByteArrayInputStream(corruptedData);
+        DataInputStream dis = new DataInputStream(bais);
+
+        assertThatThrownBy(() -> CompressionDictionary.deserialize(dis, null))
+        .as("Should throw exception for corrupted data")
+        .isInstanceOf(IOException.class)
+        .hasMessageContaining("checksum does not match");
+    }
+
+    private static byte[] createSampleDictionaryData()
+    {
+        // Create sample dictionary data that could be used for compression
+        String sampleText = "The quick brown fox jumps over the lazy dog. ";
+        return sampleText.repeat(100).getBytes();
+    }
+}

test/unit/org/apache/cassandra/db/compression/ZstdDictionaryTrainerTest.java

@@ -0,0 +1,772 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db.compression;
+
+import java.nio.ByteBuffer;
+import java.util.Map;
+import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicReference;
+import java.util.function.Consumer;
+
+import org.apache.cassandra.utils.concurrent.Future;
+
+import org.junit.After;
+import org.junit.Before;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compression.ICompressionDictionaryTrainer.TrainingStatus;
+import org.apache.cassandra.schema.CompressionParams;
+import org.apache.cassandra.utils.Clock;
+
+import static org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatThrownBy;
+
+public class ZstdDictionaryTrainerTest
+{
+    private static final String TEST_KEYSPACE = "test_ks";
+    private static final String TEST_TABLE = "test_table";
+    private static final String SAMPLE_DATA = "The quick brown fox jumps over the lazy dog. ";
+    private static final int COMPRESSION_LEVEL = 3;
+
+    private CompressionDictionaryTrainingConfig testConfig;
+    private ZstdDictionaryTrainer trainer;
+    private Consumer<CompressionDictionary> mockCallback;
+    private AtomicReference<CompressionDictionary> callbackResult;
+
+    @BeforeClass
+    public static void setUpClass()
+    {
+        DatabaseDescriptor.daemonInitialization();
+    }
+
+    @Before
+    public void setUp()
+    {
+        testConfig = CompressionDictionaryTrainingConfig.builder()
+                                                        .maxDictionarySize(1024)          // Small for testing
+                                                        .maxTotalSampleSize(10 * 1024)    // 10KB total
+                                                        .samplingRate(1)                  // 100% sampling for predictable tests
+                                                        .build();
+
+        callbackResult = new AtomicReference<>();
+        mockCallback = callbackResult::set;
+
+        trainer = new ZstdDictionaryTrainer(TEST_KEYSPACE, TEST_TABLE, testConfig, COMPRESSION_LEVEL);
+        trainer.setDictionaryTrainedListener(mockCallback);
+    }
+
+    @After
+    public void tearDown() throws Exception
+    {
+        if (trainer != null)
+        {
+            trainer.close();
+        }
+
+        // Clean up any dictionary created in callback
+        CompressionDictionary dict = callbackResult.get();
+        if (dict != null)
+        {
+            dict.close();
+            callbackResult.set(null);
+        }
+    }
+
+    @Test
+    public void testTrainerInitialState()
+    {
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Initial status should be NOT_STARTED")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+        assertThat(trainer.isReady())
+        .as("Should not be ready initially")
+        .isFalse();
+        assertThat(trainer.kind())
+        .as("Should return ZSTD kind")
+        .isEqualTo(Kind.ZSTD);
+    }
+
+    @Test
+    public void testTrainerStart()
+    {
+        // Auto start depends on configuration - test both scenarios
+        boolean started = trainer.start(false);
+        if (started)
+        {
+            assertThat(trainer.getTrainingState().getStatus())
+            .as("Status should be SAMPLING if auto-start enabled")
+            .isEqualTo(TrainingStatus.SAMPLING);
+        }
+        else
+        {
+            assertThat(trainer.getTrainingState().getStatus())
+            .as("Status should remain NOT_STARTED if auto-start disabled")
+            .isEqualTo(TrainingStatus.NOT_STARTED);
+        }
+    }
+
+    @Test
+    public void testTrainerStartManual()
+    {
+        assertThat(trainer.start(true))
+        .as("Manual training should start successfully")
+        .isTrue();
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should be SAMPLING after start")
+        .isEqualTo(TrainingStatus.SAMPLING);
+        assertThat(trainer.isReady())
+        .as("Should not be ready immediately after start")
+        .isFalse();
+    }
+
+    @Test
+    public void testTrainerStartMultipleTimes()
+    {
+        assertThat(trainer.start(true))
+        .as("First start (manual training) should succeed")
+        .isTrue();
+        Object firstTrainer = trainer.trainer();
+        assertThat(firstTrainer).isNotNull();
+        assertThat(trainer.start(true))
+        .as("Second start (manual training) should suceed and reset")
+        .isTrue();
+        Object secondTrainer = trainer.trainer();
+        assertThat(secondTrainer).isNotNull().isNotSameAs(firstTrainer);
+        assertThat(trainer.start(false))
+        .as("Third start (not manual training) should fail")
+        .isFalse();
+    }
+
+    @Test
+    public void testTrainerCloseIdempotent()
+    {
+        trainer.start(true);
+        trainer.close();
+        trainer.close(); // Should not throw
+        trainer.close(); // Should not throw
+
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should remain NOT_STARTED after multiple closes")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+    }
+
+    @Test
+    public void testTrainerReset()
+    {
+        trainer.start(true);
+        addSampleData(1000); // Add some samples
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Should have samples before reset")
+        .isGreaterThan(0);
+
+        trainer.reset();
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should be NOT_STARTED after reset")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Sample count should be 0 after reset")
+        .isEqualTo(0);
+        assertThat(trainer.isReady())
+        .as("Should not be ready after reset")
+        .isFalse();
+    }
+
+    @Test
+    public void testStartAfterClose()
+    {
+        trainer.start(true);
+        trainer.close();
+
+        assertThat(trainer.start(true))
+        .as("Should not start after close")
+        .isFalse();
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should remain NOT_STARTED")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+    }
+
+    @Test
+    public void testShouldSample()
+    {
+        trainer.start(true);
+        // With sampling rate 1 (100%), should always return true
+        for (int i = 0; i < 10; i++)
+        {
+            assertThat(trainer.shouldSample())
+            .as("Should sample with rate 1")
+            .isTrue();
+        }
+    }
+
+    @Test
+    public void testShouldSampleWithLowRate()
+    {
+        // Test with lower sampling rate
+        CompressionDictionaryTrainingConfig lowSamplingConfig =
+        CompressionDictionaryTrainingConfig.builder()
+                                           .maxDictionarySize(1024)
+                                           .maxTotalSampleSize(10 * 1024)
+                                           .samplingRate(0.001f) // 0.1% sampling
+                                           .build();
+
+        try (ZstdDictionaryTrainer lowSamplingTrainer = new ZstdDictionaryTrainer(TEST_KEYSPACE, TEST_TABLE,
+                                                                                  lowSamplingConfig, COMPRESSION_LEVEL))
+        {
+            lowSamplingTrainer.setDictionaryTrainedListener(mockCallback);
+            // With very low sampling rate, should mostly return false
+            int sampleCount = 0;
+            int iterations = 1000;
+            for (int i = 0; i < iterations; i++)
+            {
+                if (lowSamplingTrainer.shouldSample())
+                {
+                    sampleCount++;
+                }
+            }
+
+            // Should be roughly 0.1% (1 out of 1000), allow some variance
+            assertThat(sampleCount)
+            .as("Sample rate should be low")
+            .isLessThan(iterations / 10);
+        }
+    }
+
+    @Test
+    public void testAddSample()
+    {
+        trainer.start(true);
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Initial sample count should be 0")
+        .isEqualTo(0);
+
+        ByteBuffer sample = ByteBuffer.wrap(SAMPLE_DATA.getBytes());
+        trainer.addSample(sample);
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Sample count should be 1 after adding one sample")
+        .isEqualTo(1);
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should be SAMPLING")
+        .isEqualTo(TrainingStatus.SAMPLING);
+        assertThat(trainer.isReady())
+        .as("Should not be ready with single small sample")
+        .isFalse();
+    }
+
+    @Test
+    public void testAddSampleBeforeStart()
+    {
+        // Should not accept samples before start
+        ByteBuffer sample = ByteBuffer.wrap(SAMPLE_DATA.getBytes());
+        trainer.addSample(sample);
+
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should remain NOT_STARTED")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+        assertThat(trainer.isReady())
+        .as("Should not be ready")
+        .isFalse();
+    }
+
+    @Test
+    public void testAddSampleAfterClose()
+    {
+        trainer.start(true);
+        trainer.close();
+
+        ByteBuffer sample = ByteBuffer.wrap(SAMPLE_DATA.getBytes());
+        trainer.addSample(sample);
+
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should remain NOT_STARTED after close")
+        .isEqualTo(TrainingStatus.NOT_STARTED);
+        assertThat(trainer.isReady())
+        .as("Should not be ready after close")
+        .isFalse();
+    }
+
+    @Test
+    public void testAddNullSample()
+    {
+        trainer.start(true);
+        trainer.addSample(null); // Should not throw
+
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should remain SAMPLING")
+        .isEqualTo(TrainingStatus.SAMPLING);
+        assertThat(trainer.isReady())
+        .as("Should not be ready with null sample")
+        .isFalse();
+    }
+
+    @Test
+    public void testAddEmptySample()
+    {
+        trainer.start(true);
+        ByteBuffer empty = ByteBuffer.allocate(0);
+        trainer.addSample(empty); // Should not throw
+
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should remain SAMPLING")
+        .isEqualTo(TrainingStatus.SAMPLING);
+        assertThat(trainer.isReady())
+        .as("Should not be ready with empty sample")
+        .isFalse();
+    }
+
+    @Test
+    public void testIsReady()
+    {
+        trainer.start(true);
+        assertThat(trainer.isReady())
+        .as("Should not be ready initially")
+        .isFalse();
+
+        addSampleData(testConfig.acceptableTotalSampleSize / 2);
+        assertThat(trainer.isReady())
+        .as("Should not be ready with insufficient samples")
+        .isFalse();
+
+        addSampleData(testConfig.acceptableTotalSampleSize);
+        assertThat(trainer.isReady())
+        .as("Should be ready after enough samples")
+        .isTrue();
+
+        trainer.close();
+
+        assertThat(trainer.isReady())
+        .as("Should not be ready when closed")
+        .isFalse();
+    }
+
+    @Test
+    public void testTrainDictionaryWithInsufficientSampleCount()
+    {
+        trainer.start(true);
+
+        // Add sufficient data size but only 5 samples (less than minimum 11)
+        for (int i = 0; i < 5; i++)
+        {
+            ByteBuffer largeSample = ByteBuffer.wrap(new byte[testConfig.acceptableTotalSampleSize / 5]);
+            trainer.addSample(largeSample);
+        }
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Should have 5 samples")
+        .isEqualTo(5);
+        assertThat(trainer.isReady())
+        .as("Should not be ready with insufficient sample count")
+        .isFalse();
+
+        // Trying to train without force should fail with detailed message
+        assertThatThrownBy(() -> trainer.trainDictionary(false))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Trainer is not ready")
+        .hasMessageContaining("insufficient samples collected")
+        .hasMessageContaining("have 5/11 samples")
+        .hasMessageContaining("Use --force to train anyway");
+
+        // Force training should fail with insufficient samples (below absolute minimum)
+        assertThatThrownBy(() -> trainer.trainDictionary(true))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Insufficient samples for training: 5 (minimum required: 11)");
+    }
+
+    @Test
+    public void testTrainDictionaryWithSufficientSampleCount()
+    {
+        trainer.start(true);
+
+        // Add 15 samples with sufficient total size
+        for (int i = 0; i < 15; i++)
+        {
+            ByteBuffer sample = ByteBuffer.wrap(new byte[testConfig.acceptableTotalSampleSize / 15 + 1]);
+            trainer.addSample(sample);
+        }
+
+        assertThat(trainer.getTrainingState().getSampleCount()).isEqualTo(15);
+        assertThat(trainer.isReady()).isTrue();
+
+        // Training should succeed
+        CompressionDictionary dictionary = trainer.trainDictionary(false);
+        assertThat(dictionary).as("Dictionary should be created").isNotNull();
+        assertThat(trainer.getTrainingState().getStatus()).isEqualTo(TrainingStatus.COMPLETED);
+    }
+
+    @Test
+    public void testTrainDictionaryAsync() throws Exception
+    {
+        Future<CompressionDictionary> future = startTraining(true, false, testConfig.acceptableTotalSampleSize);
+        CompressionDictionary dictionary = future.get(5, TimeUnit.SECONDS);
+
+        assertThat(dictionary).as("Dictionary should not be null").isNotNull();
+        assertThat(trainer.getTrainingState().getStatus()).as("Status should be COMPLETED").isEqualTo(TrainingStatus.COMPLETED);
+
+        // Verify callback was called
+        assertThat(callbackResult.get()).as("Callback should have been called").isNotNull();
+        assertThat(callbackResult.get().dictId()).as("Callback should receive same dictionary").isEqualTo(dictionary.dictId());
+    }
+
+    @Test
+    public void testTrainDictionaryAsyncForce() throws Exception
+    {
+        // Don't add enough samples
+        Future<CompressionDictionary> future = startTraining(true, true, 512);
+        CompressionDictionary dictionary = future.get(1, TimeUnit.SECONDS);
+        assertThat(dictionary)
+        .as("Forced async training should produce dictionary")
+        .isNotNull();
+    }
+
+    @Test
+    public void testTrainDictionaryAsyncForceFailsWithNoData() throws Exception
+    {
+        AtomicReference<CompressionDictionary> dictRef = new AtomicReference<>();
+        Future<CompressionDictionary> result = startTraining(true, true, 0)
+                                                          .addCallback((dict, t) -> dictRef.set(dict));
+
+        assertThat(result.isDone() && result.cause() != null)
+        .as("Result should be completed exceptionally")
+        .isTrue();
+        assertThat(trainer.getTrainingState().getStatus())
+        .as("Status should be FAILED")
+        .isEqualTo(TrainingStatus.FAILED);
+        assertThat(dictRef.get())
+        .as("Dictionary reference should be null")
+        .isNull();
+    }
+
+    @Test
+    public void testDictionaryTrainedListener()
+    {
+        trainer.start(true);
+        addSampleData(testConfig.acceptableTotalSampleSize);
+
+        // Train dictionary synchronously - callback should be called
+        CompressionDictionary dictionary = trainer.trainDictionary(false);
+
+        // Verify callback was invoked with the dictionary
+        assertThat(callbackResult.get()).as("Callback should have been called").isNotNull();
+        assertThat(callbackResult.get().dictId().id)
+        .as("Callback should receive correct dictionary ID")
+        .isEqualTo(dictionary.dictId().id);
+        assertThat(callbackResult.get().kind())
+        .as("Callback should receive correct dictionary kind")
+        .isEqualTo(dictionary.kind());
+    }
+
+    @Test
+    public void testMonotonicDictionaryIds()
+    {
+        long now = Clock.Global.currentTimeMillis();
+        long id1 = ZstdDictionaryTrainer.makeDictionaryId(now, 100L);
+        long hourLater= now + TimeUnit.HOURS.toMillis(1);
+        long id2 = ZstdDictionaryTrainer.makeDictionaryId(hourLater, 200L);
+        long id3 = ZstdDictionaryTrainer.makeDictionaryId(now, 200L);
+
+        assertThat(id2)
+        .as("Dictionary IDs should be monotonic over time")
+        .isGreaterThan(id1)
+        .isGreaterThan(id3);
+
+        assertThat(id3).isNotEqualTo(id1).isNotEqualTo(id2);
+    }
+
+    @Test
+    public void testIsCompatibleWith()
+    {
+        CompressionParams compatibleParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                    Map.of("compression_level", "3"));
+
+        assertThat(trainer.isCompatibleWith(compatibleParams))
+        .as("Should be compatible with same compression level")
+        .isTrue();
+
+
+        CompressionParams incompatibleParams = CompressionParams.lz4();
+
+        assertThat(trainer.isCompatibleWith(incompatibleParams))
+        .as("Should not be compatible with different compressor")
+        .isFalse();
+
+        CompressionParams differentLevelParams = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true,
+                                                                        Map.of("compression_level", "4"));
+
+        assertThat(trainer.isCompatibleWith(differentLevelParams))
+        .as("Should not be compatible with different compression level")
+        .isFalse();
+
+        CompressionParams disabledParams = CompressionParams.noCompression();
+
+        assertThat(trainer.isCompatibleWith(disabledParams))
+        .as("Should not be compatible with disabled compression")
+        .isFalse();
+    }
+
+    @Test
+    public void testUpdateSamplingRate()
+    {
+        trainer.start(true);
+
+        // Test updating to different valid sampling rates
+        trainer.updateSamplingRate(10);
+
+        // With sampling rate 10 (10%), should mostly return false
+        int sampleCount = 0;
+        int iterations = 1000;
+        for (int i = 0; i < iterations; i++)
+        {
+            if (trainer.shouldSample())
+            {
+                sampleCount++;
+            }
+        }
+
+        // Should be roughly 10% (1 out of 10), allow some variance
+        assertThat(sampleCount)
+        .as("Sample rate should be approximately 10%")
+        .isGreaterThan(iterations / 20) // at least 5%
+        .isLessThan(iterations / 5);    // at most 20%
+
+        // Test updating to 100% sampling
+        trainer.updateSamplingRate(1);
+
+        // Should always sample now
+        for (int i = 0; i < 10; i++)
+        {
+            assertThat(trainer.shouldSample())
+            .as("Should always sample with rate 1")
+            .isTrue();
+        }
+    }
+
+    @Test
+    public void testUpdateSamplingRateValidation()
+    {
+        trainer.start(true);
+
+        // Test invalid sampling rates
+        assertThatThrownBy(() -> trainer.updateSamplingRate(0))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("Sampling rate must be positive");
+
+        assertThatThrownBy(() -> trainer.updateSamplingRate(-1))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("Sampling rate must be positive");
+
+        assertThatThrownBy(() -> trainer.updateSamplingRate(-100))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("Sampling rate must be positive");
+    }
+
+    @Test
+    public void testUpdateSamplingRateBeforeStart()
+    {
+        // Should be able to update sampling rate even before start
+        trainer.updateSamplingRate(5);
+
+        trainer.start(true);
+
+        // Verify the updated rate is used after start
+        int sampleCount = 0;
+        int iterations = 1000;
+        for (int i = 0; i < iterations; i++)
+        {
+            if (trainer.shouldSample())
+            {
+                sampleCount++;
+            }
+        }
+
+        // Should be roughly 20% (1 out of 5), allow some variance
+        assertThat(sampleCount)
+        .as("Sample rate should be approximately 20%")
+        .isGreaterThan(iterations / 10) // at least 10%
+        .isLessThan(iterations / 2);    // at most 50%
+    }
+
+    @Test
+    public void testTrainDictionaryNotInitialized()
+    {
+        // Try to train without starting
+        assertThatThrownBy(() -> trainer.trainDictionary(false))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Trainer is not ready")
+        .hasMessageContaining("trainer not initialized")
+        .hasMessageContaining("call start() first");
+    }
+
+    @Test
+    public void testTrainDictionaryClosed()
+    {
+        trainer.start(true);
+        addSampleData(testConfig.acceptableTotalSampleSize);
+        trainer.close();
+
+        // Try to train after closing
+        assertThatThrownBy(() -> trainer.trainDictionary(false))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Trainer is not ready")
+        .hasMessageContaining("trainer is closed");
+    }
+
+    @Test
+    public void testTrainDictionaryInsufficientSampleSize()
+    {
+        trainer.start(true);
+
+        // Add enough samples (15) but with insufficient total size
+        for (int i = 0; i < 15; i++)
+        {
+            ByteBuffer smallSample = ByteBuffer.wrap(new byte[10]);
+            trainer.addSample(smallSample);
+        }
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Should have 15 samples")
+        .isEqualTo(15);
+        assertThat(trainer.getTrainingState().getTotalSampleSize())
+        .as("Total sample size should be small")
+        .isLessThan(testConfig.acceptableTotalSampleSize);
+        assertThat(trainer.isReady())
+        .as("Should not be ready with insufficient sample size")
+        .isFalse();
+
+        // Trying to train without force should fail with detailed message
+        assertThatThrownBy(() -> trainer.trainDictionary(false))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Trainer is not ready")
+        .hasMessageContaining("insufficient sample size")
+        .hasMessageContaining("have 150 bytes/8 KiB")
+        .hasMessageContaining("Use --force to train anyway");
+    }
+
+    @Test
+    public void testTrainDictionaryInsufficientBothSampleCountAndSize()
+    {
+        trainer.start(true);
+
+        // Add only 3 samples with small size
+        for (int i = 0; i < 3; i++)
+        {
+            ByteBuffer smallSample = ByteBuffer.wrap(new byte[10]);
+            trainer.addSample(smallSample);
+        }
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Should have 3 samples")
+        .isEqualTo(3);
+        assertThat(trainer.isReady())
+        .as("Should not be ready with insufficient samples and size")
+        .isFalse();
+
+        // Trying to train without force should fail with detailed message showing both issues
+        assertThatThrownBy(() -> trainer.trainDictionary(false))
+        .isInstanceOf(IllegalStateException.class)
+        .hasMessageContaining("Trainer is not ready")
+        .hasMessageContaining("insufficient samples collected")
+        .hasMessageContaining("have 3/11 samples, 30 bytes/8 KiB")
+        .hasMessageContaining("Use --force to train anyway");
+    }
+
+    private Future<CompressionDictionary> startTraining(boolean manualTraining, boolean forceTrain, int sampleSize) throws Exception
+    {
+        trainer.start(manualTraining);
+        if (sampleSize > 0)
+        {
+            addSampleData(sampleSize);
+        }
+
+        if (forceTrain)
+        {
+            assertThat(trainer.isReady())
+            .as("Trainer should not be ready to train due to lack of samples")
+            .isFalse();
+        }
+
+        CountDownLatch latch = new CountDownLatch(1);
+        Future<CompressionDictionary> future = trainer.trainDictionaryAsync(forceTrain)
+                                                                 .addCallback((dict, throwable) -> latch.countDown());
+        assertThat(latch.await(10, TimeUnit.SECONDS))
+        .as("Training should complete within timeout")
+        .isTrue();
+        return future;
+    }
+
+    private void addSampleData(int totalSize)
+    {
+        byte[] sampleBytes = SAMPLE_DATA.getBytes();
+        int samplesNeeded = (totalSize + sampleBytes.length - 1) / sampleBytes.length; // Round up
+
+        for (int i = 0; i < samplesNeeded; i++)
+        {
+            ByteBuffer sample = ByteBuffer.wrap(sampleBytes);
+            trainer.addSample(sample);
+        }
+    }
+
+    @Test
+    public void testStatisticsMethods()
+    {
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Initial sample count should be 0")
+        .isEqualTo(0);
+
+        assertThat(trainer.getTrainingState().getTotalSampleSize())
+        .as("Initial total sample size should be 0")
+        .isEqualTo(0);
+
+        // Start training
+        trainer.start(true);
+
+        // Add some samples
+        byte[] sampleBytes = SAMPLE_DATA.getBytes();
+        int sampleSize = sampleBytes.length;
+        int numSamples = 5;
+
+        for (int i = 0; i < numSamples; i++)
+        {
+            trainer.addSample(ByteBuffer.wrap(sampleBytes));
+        }
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Sample count should be updated after adding samples")
+        .isEqualTo(numSamples);
+
+        assertThat(trainer.getTrainingState().getTotalSampleSize())
+        .as("Total sample size should match number of samples times sample size")
+        .isEqualTo((long) numSamples * sampleSize);
+
+        trainer.reset();
+
+        assertThat(trainer.getTrainingState().getSampleCount())
+        .as("Sample count should be 0 after reset")
+        .isEqualTo(0);
+
+        assertThat(trainer.getTrainingState().getTotalSampleSize())
+        .as("Total sample size should be 0 after reset")
+        .isEqualTo(0);
+    }
+}

test/unit/org/apache/cassandra/io/compress/CQLCompressionTest.java

@@ -81,11 +81,11 @@ public class CQLCompressionTest extends CQLTester
     {
         createTable("create table %s (id int primary key, uh text) with compression = {'class':'ZstdCompressor', 'compression_level':-22}");
         assertTrue(((ZstdCompressor)getCurrentColumnFamilyStore().metadata().params.compression.getSstableCompressor()).getClass().equals(ZstdCompressor.class));
-        assertEquals(((ZstdCompressor)getCurrentColumnFamilyStore().metadata().params.compression.getSstableCompressor()).getCompressionLevel(), -22);
+        assertEquals(((ZstdCompressor)getCurrentColumnFamilyStore().metadata().params.compression.getSstableCompressor()).compressionLevel(), -22);
 
         createTable("create table %s (id int primary key, uh text) with compression = {'class':'ZstdCompressor'}");
         assertTrue(((ZstdCompressor)getCurrentColumnFamilyStore().metadata().params.compression.getSstableCompressor()).getClass().equals(ZstdCompressor.class));
-        assertEquals(((ZstdCompressor)getCurrentColumnFamilyStore().metadata().params.compression.getSstableCompressor()).getCompressionLevel(), ZstdCompressor.DEFAULT_COMPRESSION_LEVEL);
+        assertEquals(((ZstdCompressor)getCurrentColumnFamilyStore().metadata().params.compression.getSstableCompressor()).compressionLevel(), ZstdCompressor.DEFAULT_COMPRESSION_LEVEL);
     }
 
     @Test(expected = ConfigurationException.class)

test/unit/org/apache/cassandra/io/compress/CompressedRandomAccessReaderTest.java

@@ -314,4 +314,4 @@ public class CompressedRandomAccessReaderTest
         file.write(checksum);
         SyncUtil.sync(file.getFD());
     }
-}
+}

test/unit/org/apache/cassandra/io/compress/CompressedSequentialWriterTest.java

@@ -395,4 +395,4 @@ public class CompressedSequentialWriterTest extends SequentialWriterTest
         }
     }
 
-}
+}

test/unit/org/apache/cassandra/io/compress/CompressionMetadataTest.java

@@ -42,7 +42,8 @@ public class CompressionMetadataTest
                                        memory,
                                        memory.size(),
                                        dataLength,
-                                       compressedFileLength);
+                                       compressedFileLength,
+                                       null);
     }
 
     @Test
@@ -75,4 +76,4 @@ public class CompressionMetadataTest
         assertThat(copy.isCleanedUp()).isTrue();
         assertThatExceptionOfType(AssertionError.class).isThrownBy(memory::size);
     }
-}
+}

test/unit/org/apache/cassandra/io/compress/ZstdCompressorTest.java

@@ -36,7 +36,7 @@ public class ZstdCompressorTest
     public void emptyConfigurationUsesDefaultCompressionLevel()
     {
         ZstdCompressor compressor = ZstdCompressor.create(Collections.emptyMap());
-        assertEquals(ZstdCompressor.DEFAULT_COMPRESSION_LEVEL, compressor.getCompressionLevel());
+        assertEquals(ZstdCompressor.DEFAULT_COMPRESSION_LEVEL, compressor.compressionLevel());
     }
 
     @Test(expected = IllegalArgumentException.class)

test/unit/org/apache/cassandra/io/compress/ZstdDictionaryCompressorTest.java

@@ -0,0 +1,392 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.io.compress;
+
+import com.github.luben.zstd.Zstd;
+import com.github.luben.zstd.ZstdDictTrainer;
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compression.ZstdCompressionDictionary;
+import org.junit.AfterClass;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.util.Collections;
+import java.util.Map;
+import java.util.Random;
+
+import static org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import static org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatThrownBy;
+import static org.junit.Assert.fail;
+
+public class ZstdDictionaryCompressorTest
+{
+    private static final int TEST_DATA_SIZE = 1024;
+    private static final String REPEATED_PATTERN = "The quick brown fox jumps over the lazy dog. ";
+
+    private static byte[] testData;
+    private static byte[] compressibleData;
+    private static ZstdCompressionDictionary testDictionary;
+
+    @BeforeClass
+    public static void setup()
+    {
+        DatabaseDescriptor.daemonInitialization();
+        testData = new byte[TEST_DATA_SIZE];
+        new Random(42).nextBytes(testData);
+
+        // Generate compressible data
+        StringBuilder sb = new StringBuilder();
+        while (sb.length() < TEST_DATA_SIZE)
+        {
+            sb.append(REPEATED_PATTERN);
+        }
+        compressibleData = sb.substring(0, TEST_DATA_SIZE).getBytes();
+        testDictionary = createTestDictionary();
+    }
+
+    @AfterClass
+    public static void tearDown()
+    {
+        if (testDictionary != null)
+        {
+            testDictionary.close();
+        }
+    }
+
+    @Test
+    public void testCreateWithOptions()
+    {
+        Map<String, String> options = Map.of(ZstdCompressor.COMPRESSION_LEVEL_OPTION_NAME, "5");
+
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(options);
+        assertThat(compressor).isNotNull();
+        assertThat(compressor.compressionLevel()).isEqualTo(5);
+        assertThat(compressor.dictionary()).isNull(); // No dictionary should be set
+    }
+
+    @Test
+    public void testCreateWithEmptyOptions()
+    {
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(Collections.emptyMap());
+        assertThat(compressor).isNotNull();
+        assertThat(compressor.compressionLevel()).isEqualTo(ZstdCompressor.DEFAULT_COMPRESSION_LEVEL);
+    }
+
+    @Test
+    public void testCreateWithDictionary()
+    {
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(testDictionary);
+        assertThat(compressor).isNotNull();
+        assertThat(compressor.compressionLevel()).isEqualTo(ZstdCompressor.DEFAULT_COMPRESSION_LEVEL);
+        assertThat(compressor.dictionary()).isSameAs(testDictionary);
+    }
+
+    @Test
+    public void testCreateWithInvalidCompressionLevel()
+    {
+        String invalidLevel = String.valueOf(Zstd.maxCompressionLevel() + 1);
+        Map<String, String> options = Map.of(ZstdCompressor.COMPRESSION_LEVEL_OPTION_NAME, invalidLevel);
+
+        assertThatThrownBy(() -> ZstdDictionaryCompressor.create(options))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessage(ZstdCompressor.COMPRESSION_LEVEL_OPTION_NAME + '=' + invalidLevel + " is invalid");
+    }
+
+    @Test
+    public void testCompressDecompressWithDictionary() throws IOException
+    {
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(testDictionary);
+
+        ByteBuffer input = ByteBuffer.allocateDirect(compressibleData.length);
+        input.put(compressibleData);
+        input.flip();
+
+        ByteBuffer compressed = ByteBuffer.allocateDirect(compressor.initialCompressedBufferLength(compressibleData.length));
+
+        // Compress
+        compressor.compress(input, compressed);
+        compressed.flip();
+
+        assertThat(compressed.remaining())
+        .as("Data should be compressed")
+        .isLessThan(compressibleData.length);
+
+        // Decompress
+        ByteBuffer decompressed = ByteBuffer.allocateDirect(compressibleData.length);
+        compressed.rewind();
+        compressor.uncompress(compressed, decompressed);
+        decompressed.flip();
+
+        // Verify roundtrip
+        byte[] result = new byte[decompressed.remaining()];
+        decompressed.get(result);
+        assertThat(result).isEqualTo(compressibleData);
+    }
+
+    @Test
+    public void testCompressDecompressWithoutDictionary() throws IOException
+    {
+        // Test fallback behavior when no dictionary is provided
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(Collections.emptyMap());
+
+        ByteBuffer input = ByteBuffer.allocateDirect(testData.length);
+        input.put(testData);
+        input.flip();
+
+        ByteBuffer compressed = ByteBuffer.allocateDirect(compressor.initialCompressedBufferLength(testData.length));
+
+        // Compress
+        compressor.compress(input, compressed);
+        compressed.flip();
+
+        // Decompress
+        ByteBuffer decompressed = ByteBuffer.allocateDirect(testData.length);
+        compressed.rewind();
+        compressor.uncompress(compressed, decompressed);
+        decompressed.flip();
+
+        // Verify roundtrip
+        byte[] result = new byte[decompressed.remaining()];
+        decompressed.get(result);
+        assertThat(result).isEqualTo(testData);
+    }
+
+    @Test
+    public void testCompressDecompressByteArray() throws IOException
+    {
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(testDictionary);
+
+        // Test byte array compression/decompression using direct buffers
+        ByteBuffer input = ByteBuffer.allocateDirect(compressibleData.length);
+        input.put(compressibleData);
+        input.flip();
+
+        ByteBuffer output = ByteBuffer.allocateDirect(compressor.initialCompressedBufferLength(compressibleData.length));
+
+        compressor.compress(input, output);
+        int compressedLength = output.position();
+
+        // Extract compressed data to byte array for array-based decompression test
+        byte[] compressed = new byte[compressedLength];
+        output.flip();
+        output.get(compressed);
+
+        // Decompress using byte array method
+        byte[] decompressed = new byte[compressibleData.length];
+        int decompressedLength = compressor.uncompress(compressed, 0, compressedLength, decompressed, 0);
+
+        assertThat(decompressedLength).isEqualTo(compressibleData.length);
+        assertThat(decompressed).isEqualTo(compressibleData);
+    }
+
+    @Test
+    public void testDictionaryCompressionImprovement()
+    {
+        // Test that dictionary compression provides better compression ratio
+        ZstdDictionaryCompressor dictCompressor = ZstdDictionaryCompressor.create(testDictionary);
+        ZstdDictionaryCompressor noDictCompressor = ZstdDictionaryCompressor.create(Collections.emptyMap());
+
+        ByteBuffer input1 = ByteBuffer.allocateDirect(compressibleData.length);
+        input1.put(compressibleData);
+        input1.flip();
+
+        ByteBuffer input2 = ByteBuffer.allocateDirect(compressibleData.length);
+        input2.put(compressibleData);
+        input2.flip();
+
+        ByteBuffer dictCompressed = ByteBuffer.allocateDirect(dictCompressor.initialCompressedBufferLength(compressibleData.length));
+        ByteBuffer noDictCompressed = ByteBuffer.allocateDirect(noDictCompressor.initialCompressedBufferLength(compressibleData.length));
+
+        try
+        {
+            dictCompressor.compress(input1, dictCompressed);
+            noDictCompressor.compress(input2, noDictCompressed);
+
+            dictCompressed.flip();
+            noDictCompressed.flip();
+
+            // Dictionary compression should achieve better compression ratio for repetitive data
+            assertThat(dictCompressed.remaining())
+            .as("Dictionary compression should achieve better compression ratio")
+            .isLessThanOrEqualTo(noDictCompressed.remaining());
+        }
+        catch (IOException e)
+        {
+            fail("Compression should not fail: " + e.getMessage());
+        }
+    }
+
+    @Test
+    public void testCompressorCaching()
+    {
+        // Test that same dictionary returns same compressor instance
+        ZstdDictionaryCompressor compressor1 = ZstdDictionaryCompressor.create(testDictionary);
+        ZstdDictionaryCompressor compressor2 = ZstdDictionaryCompressor.create(testDictionary);
+
+        assertThat(compressor1)
+        .as("Same dictionary should return cached compressor instance")
+        .isSameAs(compressor2);
+    }
+
+    @Test
+    public void testGetOrCopyWithDictionary()
+    {
+        ZstdDictionaryCompressor originalCompressor = ZstdDictionaryCompressor.create(Collections.emptyMap());
+        ZstdDictionaryCompressor dictCompressor = originalCompressor.getOrCopyWithDictionary(testDictionary);
+
+        assertThat(dictCompressor)
+        .as("Should return different compressor instance")
+        .isNotSameAs(originalCompressor);
+        assertThat(dictCompressor.dictionary())
+        .as("Should have the provided dictionary")
+        .isSameAs(testDictionary);
+        assertThat(dictCompressor.compressionLevel())
+        .as("Should preserve compression level")
+        .isEqualTo(originalCompressor.compressionLevel());
+    }
+
+    @Test
+    public void testGetOrCopyWithSameDictionary()
+    {
+        ZstdDictionaryCompressor originalCompressor = ZstdDictionaryCompressor.create(testDictionary);
+        ZstdDictionaryCompressor sameCompressor = originalCompressor.getOrCopyWithDictionary(testDictionary);
+
+        assertThat(sameCompressor)
+        .as("Same dictionary should return same compressor")
+        .isSameAs(originalCompressor);
+    }
+
+    @Test
+    public void testClosedDictionaryHandling()
+    {
+        ZstdDictionaryCompressor.invalidateCache();
+        ZstdCompressionDictionary closedDict = createTestDictionary();
+        closedDict.close();
+
+        // This should throw IllegalStateException
+        assertThatThrownBy(() -> ZstdDictionaryCompressor.create(closedDict))
+        .isInstanceOf(IllegalStateException.class);
+    }
+
+    @Test
+    public void testCompressionWithNullDictionary() throws IOException
+    {
+        // Test that null dictionary falls back to standard compression
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create((ZstdCompressionDictionary) null);
+
+        ByteBuffer input = ByteBuffer.allocateDirect(testData.length);
+        input.put(testData);
+        input.flip();
+
+        ByteBuffer compressed = ByteBuffer.allocateDirect(compressor.initialCompressedBufferLength(testData.length));
+
+        // Should not throw exception, should fall back to standard Zstd
+        compressor.compress(input, compressed);
+        compressed.flip();
+
+        ByteBuffer decompressed = ByteBuffer.allocateDirect(testData.length);
+        compressed.rewind();
+        compressor.uncompress(compressed, decompressed);
+        decompressed.flip();
+
+        byte[] result = new byte[decompressed.remaining()];
+        decompressed.get(result);
+        assertThat(result)
+        .as("Null dictionary should fall back to standard compression")
+        .isEqualTo(testData);
+    }
+
+    @Test
+    public void testDecompressionFailureHandling()
+    {
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(testDictionary);
+
+        // Create invalid compressed data
+        byte[] invalidData = new byte[10];
+        new Random().nextBytes(invalidData);
+
+        byte[] output = new byte[100];
+
+        assertThatThrownBy(() -> compressor.uncompress(invalidData, 0, invalidData.length, output, 0))
+        .isInstanceOf(IOException.class)
+        .hasMessageContaining("Decompression failed");
+    }
+
+    @Test
+    public void testAcceptableDictionaryKind()
+    {
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(Collections.emptyMap());
+        assertThat(compressor.acceptableDictionaryKind())
+        .as("Should accept ZSTD dictionary kind")
+        .isEqualTo(Kind.ZSTD);
+    }
+
+    @Test
+    public void testEmptyDataCompression() throws IOException
+    {
+        ZstdDictionaryCompressor compressor = ZstdDictionaryCompressor.create(testDictionary);
+
+        byte[] emptyData = new byte[0];
+        ByteBuffer input = ByteBuffer.allocateDirect(emptyData.length + 1); // Allocate at least 1 byte for direct buffer
+        input.put(emptyData);
+        input.flip();
+
+        ByteBuffer compressed = ByteBuffer.allocateDirect(Math.max(1, compressor.initialCompressedBufferLength(0)));
+
+        compressor.compress(input, compressed);
+        compressed.flip();
+
+        ByteBuffer decompressed = ByteBuffer.allocateDirect(1); // Allocate at least 1 byte for direct buffer
+        compressed.rewind();
+        compressor.uncompress(compressed, decompressed);
+
+        assertThat(decompressed.position())
+        .as("Should have written nothing for empty data")
+        .isEqualTo(0);
+    }
+
+    private static ZstdCompressionDictionary createTestDictionary()
+    {
+        try
+        {
+            int sampleSize = 100 * 1024;
+            int dictSize = 6 * 1024;
+            // Create a simple dictionary from repetitive data
+            ZstdDictTrainer trainer = new ZstdDictTrainer(sampleSize, dictSize, 3);
+
+            for (int i = 0; i < 1000; i++)
+            {
+                trainer.addSample(compressibleData);
+            }
+
+            byte[] dictBytes = trainer.trainSamples();
+            DictId dictId = new DictId(Kind.ZSTD, 1);
+
+            return new ZstdCompressionDictionary(dictId, dictBytes);
+        }
+        catch (Exception e)
+        {
+            throw new RuntimeException("Failed to create test dictionary", e);
+        }
+    }
+}

test/unit/org/apache/cassandra/io/sstable/ScrubTest.java

@@ -559,7 +559,7 @@ public class ScrubTest
 
         if (compression)
         { // overwrite with garbage the compression chunks from key1 to key2
-            CompressionMetadata compData = CompressionInfoComponent.load(sstable.descriptor);
+            CompressionMetadata compData = CompressionInfoComponent.load(sstable.descriptor, null);
 
             CompressionMetadata.Chunk chunk1 = compData.chunkFor(
             sstable.getPosition(PartitionPosition.ForKey.get(key1, sstable.getPartitioner()), SSTableReader.Operator.EQ));

test/unit/org/apache/cassandra/schema/CompressionParamsTest.java

@@ -0,0 +1,74 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.schema;
+
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import org.apache.cassandra.config.DatabaseDescriptor;
+
+import static org.assertj.core.api.Assertions.assertThat;
+
+public class CompressionParamsTest
+{
+    @BeforeClass
+    public static void beforeClass()
+    {
+        DatabaseDescriptor.daemonInitialization();
+    }
+
+    @Test
+    public void testIsDictionaryCompressionEnabled()
+    {
+        CompressionParams noCompression = CompressionParams.noCompression();
+        assertThat(noCompression.isDictionaryCompressionEnabled())
+        .as("No compression should not enable dictionary compression")
+        .isFalse();
+
+        CompressionParams regularZstd = CompressionParams.zstd();
+        assertThat(regularZstd.isDictionaryCompressionEnabled())
+        .as("Regular Zstd compression should not enable dictionary compression")
+        .isFalse();
+
+        CompressionParams zstdDictionary = CompressionParams.zstd(CompressionParams.DEFAULT_CHUNK_LENGTH, true);
+        assertThat(zstdDictionary.isDictionaryCompressionEnabled())
+        .as("Zstd dictionary compression should enable dictionary compression")
+        .isTrue();
+
+        CompressionParams lz4 = CompressionParams.lz4();
+        assertThat(lz4.isDictionaryCompressionEnabled())
+        .as("LZ4 compression should not enable dictionary compression")
+        .isFalse();
+
+        CompressionParams snappy = CompressionParams.snappy();
+        assertThat(snappy.isDictionaryCompressionEnabled())
+        .as("Snappy compression should not enable dictionary compression")
+        .isFalse();
+
+        CompressionParams deflate = CompressionParams.deflate();
+        assertThat(deflate.isDictionaryCompressionEnabled())
+        .as("Deflate compression should not enable dictionary compression")
+        .isFalse();
+
+        CompressionParams noop = CompressionParams.noop();
+        assertThat(noop.isDictionaryCompressionEnabled())
+        .as("Noop compression should not enable dictionary compression")
+        .isFalse();
+    }
+}

test/unit/org/apache/cassandra/schema/SystemDistributedKeyspaceCompressionDictionaryTest.java

@@ -0,0 +1,236 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.schema;
+
+import java.util.List;
+import java.util.Set;
+
+import org.junit.Before;
+import org.junit.Test;
+
+import org.apache.cassandra.cql3.CQLTester;
+import org.apache.cassandra.cql3.QueryProcessor;
+import org.apache.cassandra.db.compression.CompressionDictionary;
+import org.apache.cassandra.db.compression.CompressionDictionary.DictId;
+import org.apache.cassandra.db.compression.CompressionDictionary.Kind;
+import org.apache.cassandra.db.compression.ZstdCompressionDictionary;
+
+import static org.assertj.core.api.Assertions.assertThat;
+
+public class SystemDistributedKeyspaceCompressionDictionaryTest extends CQLTester
+{
+    private static final String TEST_KEYSPACE = "test_keyspace";
+    private static final String TEST_TABLE = "test_table";
+    private static final String OTHER_TABLE = "other_table";
+
+    private CompressionDictionary testDictionary1;
+    private CompressionDictionary testDictionary2;
+
+    @Before
+    public void setUp()
+    {
+        DictId dictId1 = new DictId(Kind.ZSTD, 100L);
+        DictId dictId2 = new DictId(Kind.ZSTD, 200L);
+
+        byte[] dictData1 = "test dictionary data 1".getBytes();
+        byte[] dictData2 = "test dictionary data 2".getBytes();
+
+        testDictionary1 = new ZstdCompressionDictionary(dictId1, dictData1);
+        testDictionary2 = new ZstdCompressionDictionary(dictId2, dictData2);
+
+        clearCompressionDictionaries();
+    }
+
+    @Test
+    public void testCompressionDictionariesTableExists()
+    {
+        Set<String> tableNames = SystemDistributedKeyspace.TABLE_NAMES;
+
+        assertThat(tableNames)
+        .as("TABLE_NAMES should contain compression_dictionaries")
+        .contains(SystemDistributedKeyspace.COMPRESSION_DICTIONARIES);
+
+        // Verify the table exists in the schema
+        KeyspaceMetadata systemDistributedKs = SystemDistributedKeyspace.metadata();
+        TableMetadata compressionDictTable = systemDistributedKs
+                                             .getTableOrViewNullable(SystemDistributedKeyspace.COMPRESSION_DICTIONARIES);
+
+        assertThat(compressionDictTable)
+        .as("compression_dictionaries table should exist in schema")
+        .isNotNull();
+    }
+
+    @Test
+    public void testStoreCompressionDictionary() throws Exception
+    {
+        // Store a dictionary
+        SystemDistributedKeyspace.storeCompressionDictionary(TEST_KEYSPACE, TEST_TABLE, testDictionary1);
+
+        // Verify it was stored
+        CompressionDictionary retrieved = SystemDistributedKeyspace.retrieveLatestCompressionDictionary(
+        TEST_KEYSPACE, TEST_TABLE);
+
+        assertThat(retrieved)
+        .as("Retrieved dictionary should not be null")
+        .isNotNull();
+
+        assertThat(retrieved.dictId())
+        .as("Retrieved dictionary ID should match stored")
+        .isEqualTo(testDictionary1.dictId());
+
+        assertThat(retrieved.kind())
+        .as("Retrieved dictionary kind should match stored")
+        .isEqualTo(testDictionary1.kind());
+
+        assertThat(retrieved.rawDictionary())
+        .as("Retrieved dictionary data should match stored")
+        .isEqualTo(testDictionary1.rawDictionary());
+
+        retrieved.close();
+    }
+
+    @Test
+    public void testStoreMultipleDictionaries() throws Exception
+    {
+        // Store multiple dictionaries for the same table
+        SystemDistributedKeyspace.storeCompressionDictionary(TEST_KEYSPACE, TEST_TABLE, testDictionary1);
+        SystemDistributedKeyspace.storeCompressionDictionary(TEST_KEYSPACE, TEST_TABLE, testDictionary2);
+
+        // Should retrieve the latest one (higher ID due to clustering order)
+        CompressionDictionary latest = SystemDistributedKeyspace.retrieveLatestCompressionDictionary(
+        TEST_KEYSPACE, TEST_TABLE);
+
+        assertThat(latest)
+        .as("Should retrieve the latest dictionary")
+        .isNotNull();
+
+        assertThat(latest.dictId())
+        .as("Should retrieve dictionary with higher ID")
+        .isEqualTo(testDictionary2.dictId());
+
+        latest.close();
+    }
+
+    @Test
+    public void testRetrieveSpecificDictionary() throws Exception
+    {
+        // Store both dictionaries
+        SystemDistributedKeyspace.storeCompressionDictionary(TEST_KEYSPACE, TEST_TABLE, testDictionary1);
+        SystemDistributedKeyspace.storeCompressionDictionary(TEST_KEYSPACE, TEST_TABLE, testDictionary2);
+
+        // Retrieve specific dictionary by ID
+        CompressionDictionary dict1 = SystemDistributedKeyspace.retrieveCompressionDictionary(
+        TEST_KEYSPACE, TEST_TABLE, new DictId(Kind.ZSTD, 100L));
+        CompressionDictionary dict2 = SystemDistributedKeyspace.retrieveCompressionDictionary(
+        TEST_KEYSPACE, TEST_TABLE, new DictId(Kind.ZSTD, 200L));
+
+        assertThat(dict1)
+        .as("Should retrieve dictionary 1")
+        .isNotNull();
+
+        assertThat(dict1.dictId())
+        .as("Should retrieve correct dictionary by ID")
+        .isEqualTo(testDictionary1.dictId());
+
+        assertThat(dict2)
+        .as("Should retrieve dictionary 2")
+        .isNotNull();
+
+        assertThat(dict2.dictId())
+        .as("Should retrieve correct dictionary by ID")
+        .isEqualTo(testDictionary2.dictId());
+
+        dict1.close();
+        dict2.close();
+    }
+
+    @Test
+    public void testRetrieveNonExistentDictionary()
+    {
+        // Try to retrieve dictionary that doesn't exist
+        CompressionDictionary nonExistent = SystemDistributedKeyspace.retrieveLatestCompressionDictionary(
+        "nonexistent_keyspace", "nonexistent_table");
+
+        assertThat(nonExistent)
+        .as("Should return null for non-existent dictionary")
+        .isNull();
+
+        // Try to retrieve specific dictionary that doesn't exist
+        CompressionDictionary nonExistentById = SystemDistributedKeyspace.retrieveCompressionDictionary(
+        TEST_KEYSPACE, TEST_TABLE, new DictId(Kind.ZSTD, 999L));
+
+        assertThat(nonExistentById)
+        .as("Should return null for non-existent dictionary ID")
+        .isNull();
+    }
+
+    @Test
+    public void testStoredDictionaryIncludesLengthAndChecksum() throws Exception
+    {
+        // Store a dictionary
+        SystemDistributedKeyspace.storeCompressionDictionary(TEST_KEYSPACE, TEST_TABLE, testDictionary1);
+
+        // Query the table directly to verify dict_length and dict_checksum are stored
+        String query = String.format("SELECT dict_length, dict_checksum FROM %s.%s WHERE keyspace_name = '%s' AND table_name = '%s' AND dict_id = %d",
+                                     SchemaConstants.DISTRIBUTED_KEYSPACE_NAME,
+                                     SystemDistributedKeyspace.COMPRESSION_DICTIONARIES,
+                                     TEST_KEYSPACE,
+                                     TEST_TABLE,
+                                     testDictionary1.dictId().id);
+
+        var resultSet = QueryProcessor.executeInternal(query);
+
+        assertThat(resultSet.isEmpty())
+        .as("Should have at least one result")
+        .isFalse();
+
+        var row = resultSet.one();
+        int storedLength = row.getInt("dict_length");
+        int storedChecksum = row.getInt("dict_checksum");
+
+        // Verify the stored length matches the actual dictionary length
+        assertThat(storedLength)
+        .as("Stored dict_length should match actual dictionary length")
+        .isEqualTo(testDictionary1.rawDictionary().length);
+
+        // Verify the stored checksum matches the calculated checksum
+        byte[] dict = testDictionary1.rawDictionary();
+        int expectedChecksum = CompressionDictionary.calculateChecksum(
+            (byte) testDictionary1.kind().ordinal(),
+            testDictionary1.dictId().id,
+            dict);
+
+        assertThat(storedChecksum)
+        .as("Stored dict_checksum should match calculated checksum")
+        .isEqualTo(expectedChecksum);
+    }
+
+    private void clearCompressionDictionaries()
+    {
+        for (String table : List.of(TEST_TABLE, OTHER_TABLE))
+        {
+            String deleteQuery = String.format("DELETE FROM %s.%s WHERE keyspace_name = '%s' AND table_name = '%s'",
+                                               SchemaConstants.DISTRIBUTED_KEYSPACE_NAME,
+                                               SystemDistributedKeyspace.COMPRESSION_DICTIONARIES,
+                                               TEST_KEYSPACE,
+                                               table);
+            QueryProcessor.executeInternal(deleteQuery);
+        }
+    }
+}

test/unit/org/apache/cassandra/streaming/compression/CompressedInputStreamTest.java

@@ -142,7 +142,7 @@ public class CompressedInputStreamTest
             writer.finish();
         }
 
-        CompressionMetadata comp = CompressionInfoComponent.load(desc);
+        CompressionMetadata comp = CompressionInfoComponent.load(desc, null);
         List<SSTableReader.PartitionPositionBounds> sections = new ArrayList<>();
         for (long l : valuesToCheck)
         {

test/unit/org/apache/cassandra/tools/ToolRunner.java

@@ -666,6 +666,13 @@ public class ToolRunner
                 return this;
             }
 
+            public AssertHelp stdoutContains(String message)
+            {
+                assertThat(message).hasSizeGreaterThan(0);
+                assertThat(stdout).isNotNull().contains(message);
+                return this;
+            }
+
             private void fail(String msg)
             {
                 StringBuilder sb = new StringBuilder();

test/unit/org/apache/cassandra/tools/nodetool/TrainCompressionDictionaryTest.java

@@ -0,0 +1,268 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.tools.nodetool;
+
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import org.apache.cassandra.cql3.CQLTester;
+import org.apache.cassandra.tools.ToolRunner;
+
+import static org.apache.cassandra.tools.ToolRunner.invokeNodetool;
+import static org.assertj.core.api.Assertions.assertThat;
+
+public class TrainCompressionDictionaryTest extends CQLTester
+{
+    @BeforeClass
+    public static void setup() throws Throwable
+    {
+        requireNetwork();
+        startJMXServer();
+    }
+
+    @Test
+    public void testTrainCommandSuccess()
+    {
+        // Create a table with dictionary compression enabled
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+
+        createSSTables(true);
+
+        // Test training command with --force since we have limited test data
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary", "train", "--force", keyspace(), table);
+        result.assertOnCleanExit();
+
+        assertThat(result.getStdout())
+        .as("Should indicate training completed")
+        .contains("Training completed successfully")
+        .contains(keyspace())
+        .contains(table);
+    }
+
+    @Test
+    public void testTrainCommandWithDataButNoSSTables()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+
+        // Add test data but don't flush - memtable should be flushed automatically
+        createSSTables(false);
+
+        // Test training, the command should run flush before sampling
+        // Use --force since we have limited test data
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary",
+                                                      "train",
+                                                      "--force",
+                                                      keyspace(),
+                                                      table);
+        result.assertOnCleanExit();
+
+        assertThat(result.getStdout())
+        .as("Should flush automatically when no SSTables available")
+        .contains("Training completed successfully");
+    }
+
+    @Test
+    public void testTrainCommandWithNoSSTables()
+    {
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary",
+                                                      "train",
+                                                      keyspace(),
+                                                      table);
+        assertThat(result.getStderr())
+        .contains("Failed to trigger training: No SSTables available for training", "after flush");
+    }
+
+    @Test
+    public void testInvalidKeyspace()
+    {
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary",
+                                                      "train",
+                                                      "nonexistent_keyspace",
+                                                      "nonexistent_table");
+        result.asserts()
+              .failure()
+              .errorContains("Failed to trigger training");
+    }
+
+    @Test
+    public void testInvalidTable()
+    {
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary",
+                                                      "train",
+                                                      keyspace(),
+                                                      "nonexistent_table");
+        result.asserts()
+              .failure()
+              .errorContains("Failed to trigger training")
+              .errorContains("does not exist or does not support dictionary compression");
+    }
+
+    @Test
+    public void testTrainingOnNonDictionaryTable()
+    {
+        // Create table without dictionary compression
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'LZ4Compressor'}");
+
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary",
+                                                      "train",
+                                                      keyspace(),
+                                                      table);
+        result.asserts()
+              .failure()
+              .errorContains("does not support dictionary compression");
+    }
+
+    @Test
+    public void testTrainingWithoutDictionaryCompressionEnabled()
+    {
+        // Create table with Zstd but without dictionary compression
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdCompressor'}");
+
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary",
+                                                      "train",
+                                                      keyspace(),
+                                                      table);
+        result.asserts()
+              .failure()
+              .errorContains("does not support dictionary compression");
+    }
+
+
+    @Test
+    public void testAlterCompressionToZstdDictionary()
+    {
+        // Create table with LZ4 compression
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'LZ4Compressor'}");
+
+        // Training should fail on LZ4 table
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary", "train", keyspace(), table);
+        result.asserts()
+              .failure()
+              .errorContains("Failed to trigger training")
+              .errorContains("does not exist or does not support dictionary compression");
+
+        // Alter table to use ZstdDictionaryCompressor
+        execute("ALTER TABLE %s WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+
+        // Training should fail with no sstables
+        result = invokeNodetool("compressiondictionary", "train", keyspace(), table);
+        assertThat(result.getStderr())
+        .contains("Failed to trigger training: No SSTables available for training", "after flush");
+
+        // Write sstables
+        createSSTables(true);
+
+        // Training should now succeed (use --force since we have limited test data)
+        result = invokeNodetool("compressiondictionary", "train", "--force", keyspace(), table);
+        result.assertOnCleanExit();
+
+        assertThat(result.getStdout())
+        .as("Should indicate training completed with new dictionary")
+        .contains("Training completed successfully")
+        .contains(keyspace())
+        .contains(table);
+    }
+
+    @Test
+    public void testHelpOutput()
+    {
+        ToolRunner.ToolResult result = invokeNodetool("help", "compressiondictionary", "train");
+        result.assertOnCleanExit();
+
+        assertThat(result.getStdout())
+        .as("Should show command help")
+        .contains("nodetool compressiondictionary train - Manually trigger compression")
+        .contains("dictionary training for a table")
+        .contains("keyspace name")
+        .contains("table name")
+        .contains("-f", "--force");
+    }
+
+    @Test
+    public void testForceOptionShortForm()
+    {
+        // Create a table with dictionary compression enabled
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+
+        createSSTables(true);
+
+        // Test training command with -f flag
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary", "train", "-f", keyspace(), table);
+        result.assertOnCleanExit();
+
+        assertThat(result.getStdout())
+        .as("Should indicate training completed with force option")
+        .contains("Training completed successfully")
+        .contains(keyspace())
+        .contains(table);
+    }
+
+    @Test
+    public void testForceOptionLongForm()
+    {
+        // Create a table with dictionary compression enabled
+        String table = createTable("CREATE TABLE %s (id int PRIMARY KEY, data text) WITH compression = {'class': 'ZstdDictionaryCompressor'}");
+
+        createSSTables(true);
+
+        // Test training command with --force flag
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary", "train", "--force", keyspace(), table);
+        result.assertOnCleanExit();
+
+        assertThat(result.getStdout())
+        .as("Should indicate training completed with force option")
+        .contains("Training completed successfully")
+        .contains(keyspace())
+        .contains(table);
+    }
+
+    @Test
+    public void testCommandLineArgumentParsing()
+    {
+        // Test missing required arguments
+        ToolRunner.ToolResult result = invokeNodetool("compressiondictionary", "train");
+        result.asserts()
+              .failure()
+              .stdoutContains("Missing required parameter");
+
+        // Test missing table argument
+        result = invokeNodetool("compressiondictionary", "train", keyspace());
+        result.asserts()
+              .failure()
+              .stdoutContains("Missing required parameter");
+    }
+
+    private void createSSTables(boolean flush)
+    {
+        for (int file = 0; file < 10; file++)
+        {
+            int batchSize = 1000;
+            for (int i = 0; i < batchSize; i++)
+            {
+                int index = i + file * batchSize;
+                execute("INSERT INTO %s (id, data) VALUES (?, ?)", index, "test data " + index);
+            }
+            if (flush)
+            {
+                flush();
+            }
+        }
+    }
+}

test/unit/org/apache/cassandra/utils/StorageCompatibilityModeTest.java

@@ -40,6 +40,7 @@ public class StorageCompatibilityModeTest
             {
                 case UPGRADING:
                 case NONE:
+                case CASSANDRA_5:
                     mode.validateSstableFormat(big);
                     mode.validateSstableFormat(trie);
                     break;