API Optimization Tale: Monitor, Fix and Deploy
summary
Green build goes to production under a feature flag percent rollout while a fallback to the old direct call keeps users working. The team adds instrument every request logging (method, params, stack) and discovers hotspots: background jobs turn thousands of REST calls into a few heavy ones, then a batched calls cut thousands to ~10 optimization and MySQL preloading fixes most load. GraphQL reduces payloads and client-side filtering. A buggy sanitization default causes DB scans and minute-long queries; fixing sanitization and adding jitter, proper rate limiting and a single-query DB fix instead of Kafka read models stabilizes production.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:25Why deploy before the weekend
- 2:20Incremental approach
- 6:00Billing
- 7:55Kafka
- 11:33The hottest day of 2020
- 15:27What I learned
- 16:25Final Castle conquered
- 20:11Conclusion
related talks
28mAPI Performance A journey down the rabbit hole
25mDesign, Build, and Run Sustainable APIs and Softwares
30mOn API Optimization
25mOptimizing GraphQL Performance Tips and Techniques
55mGraphQL - APIs the New Way
29mAccelerate API Innovation: Secrets to Staying Ahead
38mAPI Modernization with GraphQL
29mYour API is Too Slow!