Intro + Deep Dive: SIG Scalability
summary
SIG Scalability defines a multi‑dimensional scalability envelope and six SLOs, measures them with cluster-loader2 and simulates large clusters with kubemark, running daily 5,000-node tests and a 100-node pre-submit CI guard. Recent work finds regressions (e.g., pod startup latency rose when priority and fairness doubled goroutines), drives fixes (immutable secrets, efficient watch resumption), and notes GKE reaches 15,000 nodes while open‑source focuses on control‑plane throughput and targeted component improvements.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:28What is Scalability
- 5:38Scalability in Kubernetes
- 13:17How to measure scalability
- 13:47Cluster Loader 2
- 15:49Cluster Node
- 16:32Whole Nodes
- 18:02PerfDash
- 18:59Graphana
- 19:25Profiling
- 19:58Test Grid
- 22:04Protecting Scalability
- 22:38Breaking Scalability
- 23:29Pot Startup Latency
- 25:37Getting involved
- 26:25QA
- 28:24Rewriting components
- 33:24Documentation
- 34:54Pod Latency
- 39:44Challenges
