Load Test Scenario Builder

Load testing simulates multiple concurrent users hitting your application to measure performance under expected (and peak) load. Run load tests before major releases, after infrastructure changes, before anticipated traffic spikes (product launches, marketing campaigns), and regularly in CI to catch performance regressions.

k6 uses JavaScript for test scripts and is optimized for high concurrency from a single machine with minimal CPU overhead. Locust uses Python and has a web UI for real-time monitoring. k6 is better for CI/CD integration and DevOps workflows; Locust is better for teams that prefer Python and want a visual dashboard.

Key scenarios: (1) Ramp-up test — gradually increase load to find the breaking point. (2) Spike test — sudden traffic surge then return to normal. (3) Soak test — sustained moderate load for hours to find memory leaks. (4) Stress test — load above expected peak to find failure modes. Start with ramp-up, then soak.

Common thresholds: http_req_duration p(95) < 500ms (95th percentile response under 500ms), http_req_failed rate < 0.01 (under 1% error rate). For strict SLAs, use p(99) < 1000ms. Set thresholds that match your user experience SLA, not arbitrary numbers.

Base your VU count on real traffic data. If your peak concurrent users is 500, test with 500 VUs (steady state) and spike to 1,500 VUs (3x peak). If you don't have data, start with 50 VUs, observe server metrics, then scale up. 1 VU ≈ 1 concurrent active user.