Sched.net: A Network-Aware Kubernetes Scheduler [I]
summary
A custom scheduler module reads per-node network health agent annotations from the API server produced by ovn-kubernetes and uses OVN topology information to make placement decisions. In a 3‑node demo four pods send different packet sizes (3 KB, 30 KB, 55 KB) and the scheduler computes a score as 1 / log throughput ×100 to prefer less loaded nodes. Result: the high‑bandwidth pod lands on a different minion and overall network load becomes more uniform, showing the approach works end‑to‑end.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:15What is Scheduler
- 1:48How Scheduler works
- 4:50Predicate functions
- 6:45Priority functions
- 8:47Drawbacks
- 10:35Prioritization
- 14:11Experimental Test Bed
- 19:12Network Agent
- 20:09How it works
- 20:58Demo
- 27:15Other approaches
- 28:16Why care
related talks
36mNetwork-aware Scheduling in Kubernetes
36mNetwork Bandwidth-Aware Kubernetes Cluster
36mNetwork Bandwidth-Aware Kubernetes Cluster
20mHow Scheduling in Kubernetes Works
47mNHD - A Topology-Aware Scheduler for K8s for Low-Latency & HPC Applications
18mLet There Be Topology-Awareness in Kube-Scheduler!
34mSIG-Scheduling Intro and Deep Dive
30mEnhancing the Kubernetes Scheduler for Diverse Workloads in Large Clusters