# Pre-Deployment Profiling ## Profiling Script To ensure Dynamo deployments comply with the SLA, we provide a pre-deployment script to profile the model performance with different parallelization mappings and recommend the parallelization mapping for prefill and decode workers and planner configurations. To use this script, the user needs to provide the target ISL, OSL, TTFT SLA, and ITL SLA. Support matrix: | Backends | Model Types | Supported | | --- | --- | --- | | vLLM | Dense | ✅ | | vLLM | MoE | 🚧 | | SGLang | Dense | ✅ | | SGLang | MoE | 🚧 | | TensorRT-LLM | Dense | 🚧 | | TensorRT-LLM | MoE | 🚧 | > [!NOTE] > The script considers a fixed ISL/OSL without KV cache reuse. If the real ISL/OSL has a large variance or a significant amount of KV cache can be reused, the result might be inaccurate. We assume there is no piggy-backed prefill requests in the decode engine. Even if there are some short piggy-backed prefill requests in the decode engine, it should not affect the ITL too much in most conditions. However, if the piggy-backed prefill requests are too much, the ITL might be inaccurate. The script will first detect the number of available GPUs on the current nodes (multi-node engine not supported yet). Then, it will profile the prefill and decode performance with different TP sizes. For prefill, since there is no in-flight batching (assume isl is long enough to saturate the GPU), the script directly measures the TTFT for a request with given isl without kv-reusing. For decode, since the ITL (or iteration time) is relevant with how many requests are in-flight, the script will measure the ITL under different number of in-flight requests. The range of the number of in-flight requests is from 1 to the maximum number of requests that the kv cache of the engine can hold. To measure the ITL without being affected by piggy-backed prefill requests, the script will enable kv-reuse and warm up the engine by issuing the same prompts before measuring the ITL. Since the kv cache is sufficient for all the requests, it can hold the kv cache of the pre-computed prompts and skip the prefill phase when measuring the ITL. After the profiling finishes, two plots will be generated in the `output-dir`. For example, here are the profiling results for `examples/llm/configs/disagg.yaml`: ![Prefill Performance](../images/h100_prefill_performance.png) ![Decode Performance](../images/h100_decode_performance.png) For the prefill performance, the script will plot the TTFT for different TP sizes and select the best TP size that meet the target TTFT SLA and delivers the best throughput per GPU. Based on how close the TTFT of the selected TP size is to the SLA, the script will also recommend the upper and lower bounds of the prefill queue size to be used in planner. For the decode performance, the script will plot the ITL for different TP sizes and different in-flight requests. Similarly, it will select the best point that satisfies the ITL SLA and delivers the best throughput per GPU and recommend the upper and lower bounds of the kv cache utilization rate to be used in planner. The script will recommend the best TP size for prefill and decode, as well as the upper and lower bounds of the prefill queue size and decode kv cache utilization rate if using load-based planner. The following information will be printed out in the terminal: ``` 2025-05-16 15:20:24 - __main__ - INFO - Analyzing results and generate recommendations... 2025-05-16 15:20:24 - __main__ - INFO - Suggested prefill TP:4 (TTFT 48.37 ms, throughput 15505.23 tokens/s/GPU) 2025-05-16 15:20:24 - __main__ - INFO - Suggested planner upper/lower bound for prefill queue size: 0.24/0.10 2025-05-16 15:20:24 - __main__ - INFO - Suggested decode TP:4 (ITL 4.83 ms, throughput 51.22 tokens/s/GPU) 2025-05-16 15:20:24 - __main__ - INFO - Suggested planner upper/lower bound for decode kv cache utilization: 0.20/0.10 ``` After finding the best TP size for prefill and decode, the script will then interpolate the TTFT with ISL and ITL with active KV cache and decode context length. This is to provide a more accurate estimation of the performance when ISL and OSL changes and will be used in the sla-planner. The results will be saved to `/_tp_interpolation`. Please change the prefill and decode TP size in the config file to match the best TP sizes obtained from the profiling script. ### Prefill Interpolation Data In prefill engine, prefills are usually done with batch size=1 and only the ISL (excluding prefix cache hit) affects the iteration time. The script profiles the selected prefill TP configuration across different ISLs and record the TTFT and prefill throughput per GPU under those ISLs. ### Decode Interpolation Data In decode engine, decode requests are added inflight and iteration time (or ITL) depends on both the context length and the real-time load of the engine. We capture the real-time load of the engine with active kv usage and average context length. The active kv usage determines the complexity of the memory-bounded attention kernel while the active kv usage divided the average context length determines the complexity of the computation bound MLP kernel. For example, the below figure shows the ITL of DS-Distilled Llama 8b model on H100 TP4. The ITL grows near-linearly with active kv usage under a fixed context length. And the slope increases as the context length decreases. ![images](../images/itl_interpolation.png) The script profiles the selected decode TP configuration across different active kv blocks and average context length. ### Output Format of Interpolation Data After suggesting the optimal TP configuration, two `.npz` files that describe the performance characteristics of the prefill and decode engines in their suggested parallel configurations will be generated. The two `.npz` files are: * `${benchmark_result_dir}/selected_prefill_interpolation/raw_data.npz}` * `prefill_isl`: a 1D Numpy array to store the ISLs used to profile the prefill engine. * `prefill_ttft`: a 1D Numpy array to store the TTFTs under the corresponding ISLs when the prefill engine is exclusively running each prefill request (i.e., with batch size of 1). The unit is in milliseconds. * `prefill_thpt_per_gpu`: a 1D Numpy array to store the prefill throughput per GPU under the corresponding ISLs. The unit is in tokens per second per GPU. * `${benchmark_result_dir}/selected_decode_interpolation/raw_data.npz` * `max_kv_tokens`: a 1D Numpy array with only one element to store the total number of KV tokens in the decode engine. * `x_kv_usage`: a 1D Numpy array to store the percentage of the active KV blocks (in the range of [0, 1]) used to profile the decode engine. The active KV blocks can be controlled by varying `(ISL + OSL / 2) * concurrency`. * `y_context_length`: a 1D Numpy array to store the average context length (ISL + OSL / 2) used to profile the decode engine. * `z_itl`: a 1D Numpy array to store the ITLs under the corresponding active KV usage and context length. To skip the prefill stage while maintaining the context length, benchmark can be done by turn on kv reuse and warmup the engine with the prompts first before running the actual profiling. The unit is in milliseconds. * `z_thpt_per_gpu`: a 1D Numpy array to store the decode throughput per GPU under the corresponding active KV usage and context length. The unit is in tokens per second per GPU. SLA planner can work with any interpolation data that follows the above format. For best results, use fine-grained and high coverage interpolation data for the prefill and decode engines. ## Running the Profiling Script in Kubernetes Set your environment variables: ```bash export NAMESPACE=your-namespace ``` **Optional Step 0: add a kubernetes secret** ```bash kubectl create secret docker-registry nvcr-imagepullsecret \ --docker-server=nvcr.io \ --docker-username='$oauthtoken' \ --docker-password= \ -n $NAMESPACE ``` **Step 1: Configure container image** You have two options for configuring your profiling setup: **Option A: Use pre-built image with custom config injection (recommended)** Use the default pre-built image and inject custom configurations via PVC: 1. **Set the container image:** ```bash export DOCKER_IMAGE=nvcr.io/nvidia/ai-dynamo/vllm-runtime:0.4.0 # or any existing image tag ``` 2. **Inject your custom disagg configuration:** ```bash # Use default disagg.yaml config python3 benchmarks/profiler/inject_disagg_config.py --namespace $NAMESPACE # Or use a custom disagg config file python3 benchmarks/profiler/inject_disagg_config.py --namespace $NAMESPACE --disagg-config my-custom-disagg.yaml # Or specify a custom target path in the PVC python3 benchmarks/profiler/inject_disagg_config.py --namespace $NAMESPACE --target-path /profiling_results/my-disagg.yaml ``` 3. **Set the config path for the profiling job:** ```bash export DGD_CONFIG_FILE=/profiling_results/disagg.yaml # or your custom path ``` This approach allows you to: - Customize DGD configurations without rebuilding container images - Test different model configurations easily - Version control your DGD configs alongside your code > **Important**: For profiling, disagg configs should be run with Grove disabled by adding the annotation `nvidia.com/enable-grove: "false"` to avoid alpha Grove status issues. > **Note**: The default location in the PVC is `/profiling_results/disagg.yaml`. If you don't inject a config, the profiler will fall back to the built-in config at `/workspace/components/backends/vllm/deploy/disagg.yaml`. **Option B: Build custom image (only if you need code changes)** Only needed if you require custom code modifications beyond configuration changes: ```bash # in the project's root folder ./container/build.sh --framework # Tag and push to your container registry export DOCKER_IMAGE= export DGD_CONFIG_FILE= # path to your disagg.yaml file within the DOCKER_IMAGE ``` **Step 2: Set SLA target** Edit `$DYNAMO_HOME/benchmarks/profiler/deploy/profile_sla_job.yaml` to set the target ISL, OSL, TTFT, and ITL. Also, set the backend type to `vllm` or `sglang`. The backend type must match the dynamo deployment in the `DGD_CONFIG_FILE`. ```yaml spec: template: spec: containers: - name: profile-sla args: - --isl - "3000" # average ISL is 3000 tokens - --osl - "150" # average OSL is 150 tokens - --ttft - "200" # target TTFT is 200ms - --itl - "20" # target ITL is 20ms - --backend - ``` **Step 3: Run profiling (required)** ```bash cd $DYNAMO_HOME/benchmarks/profiler/deploy envsubst < profiling_pvc.yaml | kubectl apply -f - envsubst < profile_sla_sa.yaml | kubectl apply -f - envsubst < profile_sla_rbac.yaml | kubectl apply -f - envsubst < profile_sla_binding.yaml | kubectl apply -f - envsubst < profile_sla_job.yaml | kubectl apply -f - ``` **Step 4: Wait for profiling to complete** ```bash kubectl get jobs -n $NAMESPACE kubectl logs job/profile-sla -n $NAMESPACE ``` ### RBAC Configuration The SLA profiling job requires specific Kubernetes permissions to manage DynamoGraphDeployment resources and access namespace information. The RBAC setup consists of: - **`profile_sla_sa.yaml`** - Service account with image pull secret for NVIDIA Container Registry access - **`profile_sla_rbac.yaml`** - Role defining required permissions for managing deployments and accessing namespace resources - **`profile_sla_binding.yaml`** - RoleBinding that associates the Role with the service account All three files are necessary: 1. The service account provides identity and image pull credentials 2. The Role defines what operations are allowed 3. The RoleBinding connects the permissions to the service account ### Viewing Profiling Results After the profiling job completes successfully, the results are stored in the persistent volume claim (PVC) created during Step 2. Here's how to access and view your profiling results: #### Accessing the Profiling Results PVC The profiling results are stored in a PVC named `profiling-pvc`. To access the results: 1. **Deploy the PVC access pod (if not already running):** ```bash kubectl apply -f benchmarks/profiler/deploy/pvc-access-pod.yaml -n $NAMESPACE ``` 2. **Access the PVC through the pod:** ```bash kubectl exec -it pvc-access-pod -n $NAMESPACE -- /bin/bash cd /profiling_results ls -la ``` > **Note**: The same `pvc-access-pod` is used for both injecting disagg configs and accessing results. If you used the `inject_disagg_config.py` script earlier, the pod may already be running. The pod auto-deletes after 5 minutes of activity. #### File Structure The profiling results directory contains the following structure: ``` /workspace/profiling_results/ ├── prefill_performance.png # Main prefill performance plot ├── decode_performance.png # Main decode performance plot ├── prefill_tp1/ # Individual TP profiling directories ... ├── decode_tp1/ ... ├── selected_prefill_interpolation/ │ ├── raw_data.npz # Prefill interpolation data │ ├── prefill_ttft_interpolation.png # TTFT vs ISL plot │ └── prefill_throughput_interpolation.png # Throughput vs ISL plot └── selected_decode_interpolation/ ├── raw_data.npz # Decode interpolation data └── decode_tp{best_tp}.png # 3D ITL surface plot ``` #### Downloading Results Locally You can download the profiling results using the automated download script or manually: **Option 1: Automated Download (Recommended)** Use the provided download script to automatically fetch all relevant files: ```bash # Download to ./results directory python3 benchmarks/profiler/download_pvc_results.py --namespace $NAMESPACE --output-dir ./results # Download to specific directory python3 benchmarks/profiler/download_pvc_results.py --namespace $NAMESPACE --output-dir /path/to/my/results # Download without any of the auto-created config.yaml files used in profiling python3 benchmarks/profiler/download_pvc_results.py --namespace $NAMESPACE --output-dir ./results --no-config ``` The script will: - Deploy a temporary access pod (auto-deletes after 5 minutes) - Scan for relevant files (*.png, *.npz, *.yaml) - Download all files maintaining directory structure - Generate a README.md with file descriptions - Clean up automatically **Option 2: Manual Download** To download the profiling results manually: 1. **Download performance plots and data files:** ```bash # Create a local directory for results mkdir -p ./profiling_results # Copy main performance plots kubectl cp pvc-access-pod:/profiling_results/prefill_performance.png ./profiling_results/ -n $NAMESPACE kubectl cp pvc-access-pod:/profiling_results/decode_performance.png ./profiling_results/ -n $NAMESPACE # Copy interpolation directories (includes additional plots and data) kubectl cp pvc-access-pod:/profiling_results/selected_prefill_interpolation/ ./profiling_results/ -n $NAMESPACE -r kubectl cp pvc-access-pod:/profiling_results/selected_decode_interpolation/ ./profiling_results/ -n $NAMESPACE -r ``` 2. **Alternative: Tar and download entire results directory:** ```bash # Inside the access pod, create a tar archive tar -czf /profiling_results/profiling_results.tar.gz -C /profiling_results . # Download the archive to your local machine kubectl cp pvc-access-pod:/profiling_results/profiling_results.tar.gz ./profiling_results.tar.gz -n $NAMESPACE # Extract locally tar -xzf profiling_results.tar.gz -C ./profiling_results/ ``` #### Viewing Performance Plots The profiling generates several performance visualization files: **Main Performance Plots:** - **`prefill_performance.png`**: Shows TTFT (Time To First Token) performance across different tensor parallelism (TP) sizes - **`decode_performance.png`**: Shows ITL (Inter-Token Latency) performance across different TP sizes and in-flight request counts **Interpolation Plots:** - **`selected_prefill_interpolation/prefill_ttft_interpolation.png`**: TTFT vs Input Sequence Length with quadratic fit - **`selected_prefill_interpolation/prefill_throughput_interpolation.png`**: Prefill throughput vs Input Sequence Length - **`selected_decode_interpolation/decode_tp{best_tp}.png`**: 3D surface plot showing ITL vs KV usage and context length #### Understanding the Data Files The `.npz` files contain raw profiling data that can be loaded and analyzed using Python: ```python import numpy as np # Load prefill data prefill_data = np.load('selected_prefill_interpolation/raw_data.npz') print("Prefill data keys:", list(prefill_data.keys())) # Load decode data decode_data = np.load('selected_decode_interpolation/raw_data.npz') print("Decode data keys:", list(decode_data.keys())) ``` #### Cleaning Up The access pod automatically deletes after 5 minutes of activity, but you can also clean it up manually: ```bash # Exit the access pod (if still inside) exit # Delete the access pod immediately (optional - it will auto-delete) kubectl delete pod pvc-access-pod -n $NAMESPACE ``` > **Note**: The access pod has `activeDeadlineSeconds: 300` and will auto-delete after 5 minutes to prevent resource waste. ### Troubleshooting #### Image Pull Authentication Errors If you see `ErrImagePull` or `ImagePullBackOff` errors with 401 unauthorized messages: 1. Ensure the `nvcr-imagepullsecret` exists in your namespace: ```bash kubectl get secret nvcr-imagepullsecret -n $NAMESPACE ``` 2. Verify the service account was created with the image pull secret: ```bash kubectl get serviceaccount profile-sla-sa -n $NAMESPACE -o yaml ``` 3. The service account should show `imagePullSecrets` containing `nvcr-imagepullsecret`.