Finding out what's really going on, with DTrace!
summary
DTrace instruments real events and produces concrete answers: tracing GC shows at most 160 ms pauses, CPU sampling reveals mostly kernel_task idle time, socket latency scripts report local DB connections fast while an external host 72.52.119.80 appears, and DNS lookups for someplace.com take 30 seconds to resolve. DTrace lets one trace socket connect latency and dynamically insert probes into user and kernel code, so the root cause surfaces and fixes (change DNS, mock the external call, or avoid external lookups) follow.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 0:53What is DTrace
- 1:48DTrace language
- 2:24DTrace probes
- 2:39DTrace instrumentation
- 3:21DTrace aggregate
- 3:55The test suite
- 4:29The next step
- 5:00Results
- 6:38Socket Connects
- 9:00Review
- 9:39Resources
- 9:49Book
- 10:08General learning
- 10:42Conclusion
related talks
37mDiving into the Details with DTrace
54mUnderstanding Performance w/DTrace (While The Customer Yells)...
30mDTrace at 21: Reflections on Fully-grown Software
05mPerformance analysis using Systrace
42mHow Much Is Tracing? Measuring the Overhead Caused by the Tracing Infrastructure
39mLinux Server Deep Dives
52mTracking Down What Happened: A Comparison of Tracing Solutions
55mRecent Advances in Linux Tracing