Measure, do not guess
Load testing drives synthetic traffic at a system to find where it breaks. Estimates point you toward a number, but only a test confirms the real ceiling.
The kinds of test
- Load test: ramp to expected peak and confirm latency holds.
- Stress test: push past the peak to find the breaking point.
- Soak test: hold load for hours to expose leaks and slow degradation.
- Spike test: jump load suddenly to test elasticity.
Doing it well
- Use realistic traffic mixes, not a single endpoint hammered alone.
- Watch tail latency and error rate, not just average throughput.
- Test in an environment that mirrors production, including data volume.
A test that hammers one cached endpoint reports a fantasy number, since production traffic mixes cold and hot paths in ways a single endpoint never does.
Key idea
Load testing finds true capacity through realistic ramped traffic, watching tail latency and errors rather than average throughput alone.