Deep Dive: Scheduling SIG
summary
Kubernetes scheduler now reaches 100 pods/sec on 5,000-node clusters after recent performance work, with inter-pod affinity trimmed from ~1000x slower to roughly 200x speedup for inter‑pod affinity bringing it within a fewx of other predicates. Pod priority and preemption are stable in 1.14. Planned and incubating pieces include scheduling framework as plugin system, even‑pod spreading with imbalance tolerance, gang scheduling (kube-batch), and a descheduler to rebalance running pods.
AI-written summary · about
chapters (13)
- 0:00Start
- 1:19Disclaimer
- 1:46Introduction
- 2:50Notable features
- 4:49Inter-Pod Affinity/Anti-affinity
- 6:05Pod Priority and Preemption
- 6:24Even Pod Spreading
- 9:39Scheduling Framework
- 10:54Gang Scheduling (Coscheduling)
- 11:56Batch Scheduling
- 12:45Descheduler
- 14:14Questions and Comments
- 14:31Where to find us
related talks
35mSIG Scheduling Deep Dive
42mDeep Dive Into the Latest Kubernetes Scheduler Features
36mDeep Dive: SIG Scheduling
39mIntro: Scheduling SIG
30mSIG Scheduling Intro & Updates
34mKubernetes Scheduling Features or How Can I Make the System Do What I Want?
29mSIG Scheduling Deep Dive
31mSIG Scheduling Update: Transition From Pod To Workload Scheduling