Fantastic Performance and Where to Find It
summary
Uses continuous profiling (Opsian) and flame graphs to find real bottlenecks in production-like runs: a Java service shows median latency drop from ~836 ms to 67 ms then to 22 ms after eliminating repeated disk reads and simplifying string comparisons (about 40x faster); a separate legacy-API demo drops a 27.9 s merge to 1.4 s by replacing a global synchronized lock with per-customer locks. Shows vmstat, ApacheBench, and wall-clock vs CPU sampling.
AI-written summary · about
chapters (7)
- 0:00Introduction to performance
- 1:05What is fantastic performance
- 3:55Profiling basics and methodology
- 8:30CPU bottleneck demo
- 22:22Blocking bottleneck demo
- 33:21Real-world profiling challenges
- 43:56Conclusion and takeaways
