9.8 KiB

Raw Blame History

title
Planner Examples

Planner Examples: Throughput-Based Scaling

Practical examples for deploying the SLA Planner with throughput-based scaling. All examples below use the DGDR workflow with pre-deployment profiling. For deployment concepts, see the Planner Guide. For a quick overview, see the Planner README.

Basic Examples

Minimal DGDR with AIC (Fastest)

The simplest way to deploy with the SLA planner. Uses AI Configurator for offline profiling (20-30 seconds instead of hours):

apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeploymentRequest
metadata:
  name: sla-aic
spec:
  model: Qwen/Qwen3-32B
  backend: vllm

  profilingConfig:
    profilerImage: "nvcr.io/nvidia/ai-dynamo/vllm-runtime:0.6.1"
    config:
      sla:
        isl: 3000
        osl: 150
        ttft: 200
        itl: 20
      sweep:
        useAiConfigurator: true
        aicSystem: h200_sxm
        aicHfId: Qwen/Qwen3-32B
        aicBackendVersion: "0.20.0"

  deploymentOverrides:
    workersImage: "nvcr.io/nvidia/ai-dynamo/vllm-runtime:0.6.1"

  autoApply: true

Deploy:

export NAMESPACE=your-namespace
kubectl apply -f components/src/dynamo/profiler/deploy/profile_sla_aic_dgdr.yaml -n $NAMESPACE

Online Profiling (Real Measurements)

Standard online profiling runs real GPU measurements for more accurate results. Takes 2-4 hours:

apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeploymentRequest
metadata:
  name: sla-online
spec:
  model: meta-llama/Llama-3.3-70B-Instruct
  backend: vllm

  profilingConfig:
    profilerImage: "nvcr.io/nvidia/ai-dynamo/vllm-runtime:0.6.1"
    config:
      sla:
        isl: 3000
        osl: 150
        ttft: 200
        itl: 20
      sweep:
        useAiConfigurator: false
        prefillInterpolationGranularity: 16
        decodeInterpolationGranularity: 6

  deploymentOverrides:
    workersImage: "nvcr.io/nvidia/ai-dynamo/vllm-runtime:0.6.1"

  autoApply: true

Deploy:

kubectl apply -f components/src/dynamo/profiler/deploy/profile_sla_dgdr.yaml -n $NAMESPACE

Available sample DGDRs in components/src/dynamo/profiler/deploy/:

profile_sla_dgdr.yaml: Standard online profiling for dense models
profile_sla_aic_dgdr.yaml: Fast offline profiling using AI Configurator
profile_sla_moe_dgdr.yaml: Online profiling for MoE models (SGLang)

Profiling Config Cases: Prior to 0.8.1, fields under profilingConfig.config use snake_case. Starting 0.8.1, fields use camelCase. There is backwards compatibility to snake_case, but example DGDRs use camelCase.

Kubernetes Examples

MoE Models (SGLang)

For Mixture-of-Experts models like DeepSeek-R1, use SGLang backend:

apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeploymentRequest
metadata:
  name: sla-moe
spec:
  model: deepseek-ai/DeepSeek-R1
  backend: sglang

  profilingConfig:
    profilerImage: "nvcr.io/nvidia/ai-dynamo/sglang-runtime:0.6.1"
    config:
      sla:
        isl: 4000
        osl: 500
        ttft: 300
        itl: 10
      sweep:
        useAiConfigurator: false

  deploymentOverrides:
    workersImage: "nvcr.io/nvidia/ai-dynamo/sglang-runtime:0.6.1"

  autoApply: true

Deploy:

kubectl apply -f components/src/dynamo/profiler/deploy/profile_sla_moe_dgdr.yaml -n $NAMESPACE

Using Existing DGD Configs (Custom Setups)

Reference an existing DynamoGraphDeployment config via ConfigMap:

Step 1: Create ConfigMap from your DGD config:

kubectl create configmap deepseek-r1-config \
  --from-file=disagg.yaml=/path/to/your/disagg.yaml \
  --namespace $NAMESPACE \
  --dry-run=client -o yaml | kubectl apply -f -

Step 2: Reference it in your DGDR:

apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeploymentRequest
metadata:
  name: deepseek-r1
spec:
  model: deepseek-ai/DeepSeek-R1
  backend: sglang

  profilingConfig:
    profilerImage: "nvcr.io/nvidia/ai-dynamo/sglang-runtime:0.6.1"
    configMapRef:
      name: deepseek-r1-config
      key: disagg.yaml  # Must match the key used in --from-file
    config:
      sla:
        isl: 4000
        osl: 500
        ttft: 300
        itl: 10
      sweep:
        useAiConfigurator: true
        aicSystem: h200_sxm
        aicHfId: deepseek-ai/DeepSeek-V3
        aicBackendVersion: "0.20.0"

  deploymentOverrides:
    workersImage: "nvcr.io/nvidia/ai-dynamo/sglang-runtime:0.6.1"

  autoApply: true

The profiler uses the DGD config from the ConfigMap as a base template, then optimizes it based on your SLA targets. The controller automatically injects spec.model and spec.backend into the final configuration.

Inline Configuration (Simple Use Cases)

For simple use cases without a custom DGD config, provide profiler configuration directly. The profiler auto-generates a basic DGD configuration:

profilingConfig:
  config:
    sla:
      isl: 8000
      osl: 200
      ttft: 200.0
      itl: 10.0

    hardware:
      minNumGpusPerEngine: 2
      maxNumGpusPerEngine: 8
      gpuType: h200_sxm

    sweep:
      prefillInterpolationGranularity: 16
      decodeInterpolationGranularity: 6

Mocker Deployment (Testing)

Deploy a mocker backend that simulates GPU timing behavior without real GPUs. Useful for:

Large-scale experiments without GPU resources
Testing planner behavior and infrastructure
Validating deployment configurations

spec:
  model: <model-name>
  backend: trtllm  # Real backend for profiling
  useMocker: true  # Deploy mocker instead of real backend

  profilingConfig:
    profilerImage: "nvcr.io/nvidia/dynamo/trtllm-runtime:<image-tag>"
    config:
      sla:
        isl: 3000
        osl: 150
        ttft: 200
        itl: 20
      sweep:
        useAiConfigurator: true
        aicSystem: h100_sxm
  autoApply: true

Profiling runs against the real backend (via GPUs or AIC). The mocker deployment then uses profiling data to simulate realistic timing.

Model Cache PVC (0.8.1+)

For large models, use a pre-populated PVC instead of downloading from HuggingFace:

See SLA-Driven Profiling for configuration details.

Advanced Examples

Custom Load Predictors

Warm-starting with Trace Data

Pre-load predictors with historical request patterns before live traffic:

# In planner arguments
args:
  - --load-predictor arima
  - --load-predictor-warmup-trace /data/trace.jsonl
  - --load-predictor-log1p

The trace file should be in mooncake-style JSONL format with request-count, ISL, and OSL samples.

Kalman Filter Tuning

For workloads with rapid changes, tune the Kalman filter:

args:
  - --load-predictor kalman
  - --kalman-q-level 2.0      # Higher = more responsive to level changes
  - --kalman-q-trend 0.5      # Higher = trend changes faster
  - --kalman-r 5.0            # Lower = trusts new measurements more
  - --kalman-min-points 3     # Fewer points before forecasting starts
  - --load-predictor-log1p    # Often helps with request-rate series

Prophet for Seasonal Workloads

For workloads with daily/weekly patterns:

args:
  - --load-predictor prophet
  - --prophet-window-size 100   # Larger window for seasonal detection
  - --load-predictor-log1p

Virtual Connector

For non-Kubernetes environments, use the VirtualConnector to communicate scaling decisions:

from dynamo._core import DistributedRuntime, VirtualConnectorClient

# Initialize client
client = VirtualConnectorClient(distributed_runtime, namespace)

# Main loop: watch for planner decisions and execute them
while True:
    # Block until the planner makes a new scaling decision
    await client.wait()

    # Read the decision
    decision = await client.get()
    print(f"Scale to: prefill={decision.num_prefill_workers}, "
          f"decode={decision.num_decode_workers}, "
          f"id={decision.decision_id}")

    # Execute scaling in your environment
    scale_prefill_workers(decision.num_prefill_workers)
    scale_decode_workers(decision.num_decode_workers)

    # Report completion
    await client.complete(decision)

See components/planner/test/test_virtual_connector.py for a full working example.

Planner Configuration Passthrough

Pass planner-specific settings through the DGDR:

profilingConfig:
  config:
    planner:
      plannerMinEndpoint: 2

Review Before Deploy (autoApply: false)

Disable auto-deployment to inspect the generated DGD:

spec:
  autoApply: false

After profiling completes:

# Extract and review generated DGD
kubectl get dgdr sla-aic -n $NAMESPACE \
  -o jsonpath='{.status.generatedDeployment}' > my-dgd.yaml

# Review and modify as needed
vi my-dgd.yaml

# Deploy manually
kubectl apply -f my-dgd.yaml -n $NAMESPACE

Profiling Artifacts with PVC

Save detailed profiling artifacts (plots, logs, raw data) to a PVC:

spec:
  profilingConfig:
    outputPVC: "dynamo-pvc"
    config:
      sla:
        isl: 3000
        osl: 150
        ttft: 200
        itl: 20

Setup:

export NAMESPACE=your-namespace
deploy/utils/setup_benchmarking_resources.sh

Access results:

kubectl apply -f deploy/utils/manifests/pvc-access-pod.yaml -n $NAMESPACE
kubectl wait --for=condition=Ready pod/pvc-access-pod -n $NAMESPACE --timeout=60s
kubectl cp $NAMESPACE/pvc-access-pod:/data ./profiling-results
kubectl delete pod pvc-access-pod -n $NAMESPACE

Planner README -- Overview and quick start
Planner Guide -- Deployment, configuration, integration
Planner Design -- Architecture deep-dive
DGDR Configuration Reference
SLA-Driven Profiling

9.8 KiB Raw Blame History

Planner Examples: Throughput-Based Scaling

Basic Examples

Minimal DGDR with AIC (Fastest)

Online Profiling (Real Measurements)

Kubernetes Examples

MoE Models (SGLang)

Using Existing DGD Configs (Custom Setups)

Inline Configuration (Simple Use Cases)

Mocker Deployment (Testing)

Model Cache PVC (0.8.1+)

Advanced Examples

Custom Load Predictors

Warm-starting with Trace Data

Kalman Filter Tuning

Prophet for Seasonal Workloads

Virtual Connector

Planner Configuration Passthrough

Review Before Deploy (autoApply: false)

Profiling Artifacts with PVC

Related Documentation

9.8 KiB

Raw Blame History