Performance Testing

When to use

• Designing a load, stress, soak, or spike test scenario for an API or service
• Selecting a performance testing tool (k6, JMeter, Gatling, Artillery)
• Establishing performance baselines before a major release
• Investigating a latency regression — bottleneck analysis workflow
• Adding performance gates to a CI/CD pipeline
• Setting up Grafana/InfluxDB dashboards for load test result visualization

Core principles

1. Baselines before gates — you cannot fail a threshold you never measured; establish baselines on a stable environment first
2. P95 and P99, never averages — average latency hides tail pain; users experience the tail, not the mean
3. Test types serve different questions — load tests validate capacity, stress tests find limits, soak tests expose leaks, spike tests prove autoscaling
4. Bottleneck order matters — check application first, then database, then network, then infra; jumping to infra first wastes everyone's time
5. Store results with the code — a test run without the commit SHA and environment spec is an anecdote, not a baseline

Reference Files

• references/tools-and-test-types.md — test type selection table, k6 script with stages and thresholds, custom metrics with Trend/Counter, JMeter CLI, Gatling DSL, Artillery overview
• references/baselines-analysis-ci.md — baseline establishment and storage, bottleneck investigation order, APM correlation, CI gate configuration, result archiving and Grafana trend analysis