The Diabolical Developer's Guide to Perfomance Tuning
summary
Performance diagnosis uses a strict decision tree (PDM): measure continuously, ask what the CPU is doing, then narrow to memory vs threads vs JIT vs I/O. Common fixes are resizing heaps or fixing object churn—surprisingly 60% of performance complaints are GC-related—and you must only fix one thing at a time. Practical tools: vmstat/netstat/wireshark, GC logs + GCViewer, VisualVM, JMH and JITWatch; example problem shows a 43-second stop-the-world pause that freezes the JVM.
AI-written summary · about
chapters (8)
- 0:00Start
- 17:44take the proper measurements
- 18:23measuring time in your systems
- 18:52running performance tests
- 19:30taking timings at each entry and exit point of your architecture
- 26:19correlate this data with an actual performance complaint
- 38:24start fixing the problem a word of warning with memory profilers
- 47:41theoretical background of how the threading model works
related talks
43mGC Tuning Confessions Of A Performance Engineer
40mDebugging JVM performance issues
54mYour Performance Tuning Check List
43mPerformance Tuning, Getting the Biggest Bang for Your Buck
57mYour Performance Tuning Check List
38mFinding a needle in the haystack - Debugging performance issues
01h 00m7 Tips to Performance Tuning, Optimization, and Everything
12mUsing Dynatrace and Kibana to Identify Bad Queries and Optimise Them