AWS adds quota management to AWS Batch for SageMaker Training jobs

Summary

AWS Batch now includes quota management for Amazon SageMaker Training jobs, allowing administrators to divide compute capacity into quota shares across teams. The feature supports lending, borrowing, and reserving GPU capacity, along with cross-share and in-share preemption for priority workloads.

What changed

• Administrators can divide compute capacity into quota shares organized by team, project, or workload.
• Configurable resource-sharing strategies allow quota shares to reserve, lend, or borrow idle compute capacity.
• AWS Batch performs cross-share preemption to reclaim lent capacity when a lending quota share receives new jobs.
• Priority-based in-share preemption preempts lower priority jobs within the same quota share so higher priority jobs run sooner.

Availability

“With the launch of AWS Batch support for Amazon SageMaker Training jobs, customers gained a new fully managed way to queue, schedule, and retry machine learning (ML) training workloads without building their own infrastructure.”

Source