Betting on Performance: A note on Hypothesis-driven Performance Testing
summary
Project replaces a legacy publishing platform, uses Siege and Vegeta for quick load probes, Graphite and OpenTracing for telemetry, and Ansible for infra-as-code to make fast changes. Initial observation shows 35 seconds initial page load against targets of 0.8s first byte and 1.5s full load. Small bets—add CPUs, move compute to US (where S3 data lives), optimize XSLT transforms, then insert Varnish—cut latency stepwise and reach 0.2s after caching for smaller files. The approach treats performance tests as hypothesis-driven bets.
AI-written summary · about
chapters (5)
- 0:00Start
- 16:28Betting on Performance
- 26:49a note on hypothesis-driven performance testing
- 32:00What do we mean when we say monitoring?
- 39:08Lightweight performance tests
related talks
41mShift Left your Performance Testing
48mThink Fast, First: Empowering a Performance Culture
50mCreating Performance Impact in a Changing Tech Landscape
40mAutomated Performance Testing In Preproduction
29mWeb Performance Kinetosis
29mProgressive Testing to Meet the Performance Imperative
57mSite-Speed That Sticks
01h 02mPerformance