performance-testing
Design and execute performance tests — load, stress, soak, spike — to validate latency, throughput, and resource consumption under realistic conditions.
Goals
- Validate performance SLIs (latency, throughput, error rate)
- Identify bottlenecks before users encounter them
- Set performance baselines and regression thresholds
- Report actionable findings with evidence
Contract
Input
A system to test: endpoints, expected load, performance targets.
Output
A performance test report with:
- Baseline metrics (latency, throughput, resource usage)
- Bottleneck analysis and recommendations
- Regression thresholds for CI
Test Types
| Type | Goal | Duration |
|---|---|---|
| Load test | Validate SLA at expected load | 15–60 min |
| Stress test | Find breaking point | Until failure |
| Soak test | Detect memory leaks | 2–8 hours |
| Spike test | Measure recovery | Minutes |
| Smoke test | Quick sanity check | 2–5 min |
Metrics
- Latency: p50, p95, p99 response time
- Throughput: requests/second
- Error rate: % of failed requests
- Resource: CPU, memory, I/O, network
Steps
- Define SLIs and targets — what is "fast enough"
- Create test scenarios — representative user journeys
- Set up test environment — mirror production configuration
- Run baseline tests — measure under normal conditions
- Stress incrementally — increase load until breaking point
- Analyze results — find bottlenecks, graph correlations
- Set regression thresholds — gate CI on performance
References
../qa-automation/SKILL.md— test infrastructure../../platform/monitoring-alerting/SKILL.md— metrics collection../../delivery/webapp-testing/SKILL.md— e2e testing