Cutting latency in half: What actually worked—and what didn’t
summary
Argues that proactive profiling (not just reactive fixes) finds surprising wins: average server response falls from 395 ms to 218 ms by profiling P95, excluding admin pages, fixing a missing index, replacing counts with previous/next, and batching cache lookups. Uses APMs (Datadog, Sentry), middleware timing, Django tools and local vs production comparisons, plus ORM tactics like select_related vs prefetch_related and lazy context processors. Shows low‑code changes often outperform heavy ORM surgery.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:56Reactive profiling failures
- 7:46What is proactive profiling
- 12:21Proactive profiling benefits
- 18:40Bottlenecks deep dive
- 21:40ORM optimization basics
- 30:51Django ORM debugging tools
- 32:38Low-code performance gains
- 39:00Django hot path optimization
- 42:18What worked for our site
- 43:22Q&A
related talks
37mPerformant Django
28mThe life of a web request - techniques for measuring and improving Django application performance
32mRails Performance First Aid
36mMake It Fast
45mRails Performance Monitoring 101: A Primer for Junior Developers
42mA Nice Problem to Have: Django Under Heavy Load
41mProfiling to make your Rails app faster
01h 02mPerformance