JavaScript Jabber Node.js Performance, Kubernetes, and Why “Fast” Isn’t Always Fast - JSJ 702
16 snips
Feb 10, 2026 Matteo Collina, Node.js project chair and founder of Platformatic, talks Node performance, Kubernetes scaling, and server-side rendering at high traffic. He covers surprising benchmark results, event-loop blocking and restart strategies, WattPM’s approach to better P95/P99, and why newer runtimes are not automatically faster. Short, technical, and provocative.
AI Snips
Chapters
Transcript
Episode notes
SSR Creates Hidden Runtime Cost
- Server-side rendering (SSR) with React can create severe latency and GC pressure under load, making many SSR backends fail at scale.
- Matteo shows smarter scheduling and multi-threaded app servers can reduce P95/P99 latencies dramatically in Kubernetes.
Restart On Event-Loop Block
- Monitor the event loop and restart processes when the loop is blocked to recover responsiveness quickly.
- Matteo recommends killing stuck Node processes rather than letting them degrade user latency during heavy GC or CPU work.
Scale Pods Internally With Workers
- Use a process manager that leverages SO_REUSEPORT and multiple workers to spread load across cores efficiently.
- Matteo suggests running heterogeneous multi-worker servers (like Watt) inside pods to absorb spikes and reduce Kubernetes scheduling pressure.