You Can't Optimize What You Can't Measure
summary
Claims that production metrics expose both code and business-level failures and let teams prove or refute assertions with data; shows tools actually used: statsd feeding graphite (or hosted DataDog), plus Lobster for logs; concrete wins include Tesla using per-second telemetry to rebut a New York Times drive log, a 250-node spider cluster hitting thousands of requests/sec, and a measured 10x performance improvement with alerts that text on anomalous drops so issues get found and fixed.
AI-written summary · about
chapters (14)
- 0:00Start
- 2:47Data
- 5:22Debugging production
- 7:47What happened?
- 10:20Logs suck
- 11:26I want to sleep
- 13:05Business problems
- 14:08Counting and timing
- 14:23The solution
- 17:19Graphite
- 18:42Lobster
- 19:19Logster
- 23:11Web spiders
- 28:03Wrapping up
related talks
23mYou Get What You Measure—Why Metrics Are Important
25mWhen Metrics Lie
36mIt's not in Production Unless it's Monitored
34mMetrics driven development
33mWhat Developers Should Know About Observability
36mMetrics - What's your code actually doing?
19mOops my app performance sucks... what if observability could be a game changer?
01h 01mFinding Metrics that matter and using them safely