API Performance A journey down the rabbit hole
summary
A real investigation finds the bottlenecks: a table grows from ~20 to 58 columns with only a few indexes so queries scan too much, RBAC policy checks run twice during a transition causing 7 checks → 14 checks per node, and many in-memory object copies run inside loops. Simple instrumentation—print timestamps and a tiny timing script (benchmark folder in tools) plus a non‑voting CI job that logs results—measures each stage: 1.9s → 11s down to 1.96s for 1,000 nodes, yielding about 96 nodes/s rising to 540 nodes/s after adding indexes, front‑loading/caching policy decisions, and selecting fields through the stack.
AI-written summary · about
chapters (9)
- 0:00Introduction to api performance
- 1:04Identifying performance issues
- 2:41Analyzing object compatibility
- 4:24Impact of policy checks
- 5:57Optimizing queries and results
- 9:25Measuring and benchmarking
- 10:50Best practices for operators
- 13:30Developer guidelines
- 15:34Q&A and concluding remarks
related talks
24mAPI Optimization Tale: Monitor, Fix and Deploy
29mYour API is Too Slow!
30mOn API Optimization
17mHands-On Experience in API Performance Testing | What is API Testing
29mAccelerate API Innovation: Secrets to Staying Ahead
25mOptimizing GraphQL Performance Tips and Techniques
51mGraphQL: advantages, challenges, and how to overcome them
38mAPI Modernization with GraphQL