AWS sagemaker documentation change
Summary
Added documentation for new model deployment option using local NVMe storage including benefits, features, and a new deployment guide section.
Security assessment
The change introduces a new deployment method leveraging local NVMe storage for performance benefits (reduced cold-start, network independence). There's no mention of security vulnerabilities, access controls, encryption, or security features. The focus is solely on performance optimization.
Diff
diff --git a/sagemaker/latest/dg/sagemaker-hyperpod-model-deployment-deploy.md b/sagemaker/latest/dg/sagemaker-hyperpod-model-deployment-deploy.md index 0abf34730..6cc857128 100644 --- a//sagemaker/latest/dg/sagemaker-hyperpod-model-deployment-deploy.md +++ b//sagemaker/latest/dg/sagemaker-hyperpod-model-deployment-deploy.md @@ -11,3 +11,3 @@ Whether you're deploying pre-trained foundation open-weights or gated models fro -| Deploy open-weights and gated foundation models from JumpStart | Deploy custom and fine-tuned models from Amazon S3 and Amazon FSx ----|---|--- -**Description** | Deploy from a comprehensive catalog of pre-trained foundation models with automatic optimization and scaling policies tailored to each model family. | Bring your own custom and fine-tuned models and leverage SageMaker HyperPod's enterprise infrastructure for production-scale inference. Choose between cost-effective storage with Amazon S3 or a high-performance file system with Amazon FSx. +| Deploy open-weights and gated foundation models from JumpStart | Deploy custom and fine-tuned models from Amazon S3 and Amazon FSx | Deploy models from local NVMe storage +---|---|---|--- +**Description** | Deploy from a comprehensive catalog of pre-trained foundation models with automatic optimization and scaling policies tailored to each model family. | Bring your own custom and fine-tuned models and leverage SageMaker HyperPod's enterprise infrastructure for production-scale inference. Choose between cost-effective storage with Amazon S3 or a high-performance file system with Amazon FSx. | Load model weights from a node's local NVMe storage to eliminate network latency during pod startup. Useful for autoscaling events, scale-from-zero workloads, and latency-sensitive failovers. @@ -26,0 +27,7 @@ Whether you're deploying pre-trained foundation open-weights or gated models fro +| + + * Reduced cold-start time by reading weights locally + * No network dependency for model loading + * Optional fallback to Amazon S3 when NVMe cache is missing + * Custom Kubernetes volumes and initContainers + @@ -40,0 +48,6 @@ Whether you're deploying pre-trained foundation open-weights or gated models fro +| + + * kubectl for Kubernetes-native operations + * Python SDK for programmatic integration + * HyperPod CLI for command-line automation + @@ -43 +56 @@ Whether you're deploying pre-trained foundation open-weights or gated models fro -The following sections step you through deploying models from Amazon SageMaker JumpStart and from Amazon S3 and Amazon FSx. +The following sections step you through deploying models from Amazon SageMaker JumpStart, from Amazon S3 and Amazon FSx, and from local NVMe storage. @@ -51 +64,3 @@ The following sections step you through deploying models from Amazon SageMaker J - * [Deploy custom fine-tuned models from Amazon S3 and Amazon FSx using kubectl](./sagemaker-hyperpod-model-deployment-deploy-ftm.html) + * [Deploy models from Amazon S3, Amazon FSx, or Hugging Face Hub using kubectl](./sagemaker-hyperpod-model-deployment-deploy-ftm.html) + + * [Deploy models from local NVMe storage using kubectl](./sagemaker-hyperpod-model-deployment-deploy-nvme.html)