.svg)
SageMaker HyperPod now supports idle resource sharing for dynamic cluster utilization
Amazon SageMaker HyperPod Resource Sharing Update
Amazon SageMaker HyperPod now supports dynamic resource sharing, enabling teams to utilize unallocated compute capacity in HyperPod clusters beyond their guaranteed quotas. This feature helps address underutilization challenges in shared compute clusters for generative AI workloads.
Key Features
- Dynamic Resource Sharing: Automatically identifies and makes unallocated cluster capacity available for teams to borrow.
- Borrow Limits: Administrators can configure absolute and percentage-based borrow limits for specific resource types.
- Automatic Recalculation: HyperPod task governance monitors cluster state and recalculates borrowable resources when instances and compute quota policies change.
- Eligible Instances: Includes instances with partitioned GPU configurations in the borrowable pool.
Regions Available
- US East (N. Virginia)
- US East (Ohio)
- US West (N. California)
- US West (Oregon)
- Asia Pacific (Mumbai)
- Asia Pacific (Singapore)
- Asia Pacific (Sydney)
- Asia Pacific (Tokyo)
- Asia Pacific (Jakarta)
- Europe (Frankfurt)
- Europe (Ireland)
- Europe (London)
- Europe (Stockholm)
- Europe (Spain)
- South America (São Paulo)
What to do
- Review the SageMaker HyperPod webpage for more information.
- Read the HyperPod task governance documentation.
Source: AWS release notes
If you need further guidance on AWS, our experts are available at AWS@westloop.io. You may also reach us by submitting the Contact Us form.
