Kubernetes v1.35 introduces workload-aware scheduling improvements, including a new Workload API and gang scheduling implementation. Gang scheduling ensures all pods in a workload start simultaneously in an all-or-nothing approach, preventing resource wastage and deadlocks. The release also includes opportunistic batching, a beta feature that speeds up scheduling of identical pods by reusing feasibility calculations. Future plans include formalized workload scheduling phases, improved autoscaling integration, multi-workload scheduling simulations, and enhanced support for Dynamic Resource Allocation and topology-aware scheduling.
Table of contents
Knowing Your Network NuancesWanna be in my Gang (Schedule)?Opportunistic Batching, Performance OptimizationThe North Star VisionRelated1 Comment
Sort: