Shift Left your Performance Testing
summary
Claims that performance hypotheses can be validated inside sprints on a laptop using a Dockerized setup and that API tests can double as perf tests. Demonstrates Perfume converting Karate→Gatling tests, feeding metrics to Grafana and Influx/Prometheus, and running local experiments that falsify trends (1, 10, 100 RPS) instead of full-scale load. Shows a cache case where expectations of 10K→100K RPS fail initially due to 100% miss rate, fixes on the laptop change behavior, and higher‑env verification remains necessary.
AI-written summary · about
chapters (14)
- 0:00Shift Left Performance Testing intro
- 1:17Why late performance issues
- 5:46Challenges of shift left
- 7:53Local testing scaling down
- 8:55Hypothesis invalidation
- 13:09Caching experiments comparison
- 14:56Hardening sprint anti-pattern
- 18:06Reducing testing complexity
- 20:35Containerizing test setups
- 23:03Perfume tool demo
- 31:49Perfume for shift left
- 33:55Testing to engineering mindset
- 34:56Experiment evolution template
- 40:16Summary
related talks
38mCreating & Scaling Meaningful Performance Tests in CI
01h 00mPerformance In Modern Times: An Evolutionary Journey
29mProgressive Testing to Meet the Performance Imperative
56mBetting on Performance: A note on Hypothesis-driven Performance Testing
29mPerfecting Perf: Rapid, Rigorous and Reproducible Performance Experiments
40mPerformance Testing Done Right
06mPerformance Testing at
42mAdding Performance Assertion to Standard Functional Testing