← All Articles
May 2, 202613 min readCloud

Cost-Aware Batch Scheduling on Spot Nodes

Reduce compute spend safely with interruption-aware scheduling, checkpointing, and queue-level SLA classes for batch workloads.

Spot capacity is powerful only when jobs are built to survive interruption.

Queue Classes

  • SLA-critical: on-demand only.
  • Flexible: spot preferred with fallback.
  • Experimental: spot only.
nodeSelector:
  capacity-type: spot
tolerations:
  - key: interruption
    operator: Exists

Cost optimization without interruption policy is just deferred incident creation.

Next ArticleFeature Flag Control-Plane Consistency ModelsPlatform · 12 min read