AWS Batch quota management lets teams lend and borrow idle SageMaker GPU capacity
AWS ML Blog · rss · 2026-10-30
AWS launched quota management for AWS Batch scheduling of SageMaker Training jobs, addressing the pain of coordinating shared GPU capacity across teams.
How it works:
- Admins define quota shares per team/project, each with capacity limits and resource-sharing strategies
- Three strategies: RESERVE (hold idle capacity), LEND (lend idle capacity out), LENDANDBORROW (lend when idle and borrow when work arrives)
- Cross-share preemption automatically reclaims lent capacity when jobs arrive back in the lending share
- In-share preemption optionally preempts lower-priority jobs so urgent work starts sooner
Setup involves a Scheduling Policy with quotaSharePolicy, a Service Environment with capacity limits, and a connected Job Queue; supported via SageMaker Python SDK v3.
The post walks through a three-project example — background jobs with LENDANDBORROW + in-share preemption, medium-priority with LEND, critical workloads with RESERVE — plus a sample notebook.
More from Infra
- Google's Ed Chi asks: will future phones be thin clients or run more on-device intelligence? — edchi · 2026-10-30
- UkisAI cuts thinking tokens 61.4% on GLM-5.3-Flash with under 0.5% accuracy loss — Secure_Recording_472 · 2026-10-30
- tiny-llm: an open-source course where you build a vLLM-style inference engine for Qwen3 on a Mac — techNmak · 2026-10-30
- WanGP one-click launcher hits Pinokio: AI video generation on 6GB+ VRAM GPUs — cocktailpeanut · 2026-10-30
- DeepSpeed fixes ZeRO-2 fp32 precision bug, users urged to upgrade to 0.19.8+ — StasBekman · 2026-10-30
- AMD ships ROCm HRX Loom v0.1 targeting its CPUs, GPUs and NPUs — Fcking_Chuck · 2026-10-30