Premature Optimization vs Hot Path

#60 - Software Tradeoffs and How to Make Good Programming Decisions - Tomasz Lelek

Topic 6 of 10

Published Oct 18, 2021 · Clip 20:09–25:06 (4:57)

Tomasz proposes a framework for just-in-time optimization: use request-per-second (throughput) data and latency measurements to compute the overall latency distribution and detect the hot path—the code executed for most requests. Systems often follow Pareto-like distributions, so optimizing a small part of the code yields large gains. He explains how to get data via SLAs, benchmarks (Gatling, WRK), and expected traffic growth from product input.