Qa Performance

New project without SLAs defined: propose standard thresholds (p50<100ms, p95<500ms, p99<1000ms, error rate<1%) and let the first run establish the baseline.

CI integration: run a smoke test (1 VU, 1 iteration) as a pre-merge gate, then full load tests nightly or pre-release.

Production load testing: always use staging or a dedicated load environment. Never target production without rate limits in place.

Single-region bias: note when recommending distributed runners for global traffic simulation.

Examples

Example 1

Input: Load test the products API — 50 concurrent users for 2 minutes with SLA thresholds

Output:

// tests/perf/products-load.js
import http from 'k6/http';
import { check, sleep } from 'k6';

export const options = {
  stages: [
    { duration: '30s', target: 50 },
    { duration: '2m',  target: 50 },
    { duration: '15s', target: 0  },
  ],
  thresholds: {
    http_req_duration: ['p(50)<100', 'p(95)<500', 'p(99)<1000'],
    http_req_failed:   ['rate<0.01'],
  },
};

export default function () {
  const res = http.get(`${__ENV.BASE_URL}/products`, {
    headers: { Authorization: `Bearer ${__ENV.API_TOKEN}` },
  });
  check(res, { 'status 200': (r) => r.status === 200 });
  sleep(1);
}

k6 run \
  -e BASE_URL=https://staging.api.example.com \
  -e API_TOKEN=$TOKEN \
  tests/perf/products-load.js

Example 2

Input: Quick latency check on the health endpoint

Output:

oha -z 30s -c 10 \
  -H "Authorization: Bearer $TOKEN" \
  https://api.example.com/health

# Parse key percentiles
oha --json -n 200 -c 20 https://api.example.com/health | jaq '{
  rps: .rps,
  p50: .latencyPercentiles."50",
  p95: .latencyPercentiles."95",
  p99: .latencyPercentiles."99",
  errors: .errorCount
}'

Quick Reference

See references/performance-reference.md for complete k6 scripts (spike, scenario-based), oha and vegeta command reference, Lighthouse CLI usage, SLA validation script, and Flutter performance profiling.

Install:

brew install k6
cargo install oha
go install github.com/tsenart/vegeta@latest
npm install -g lighthouse
cargo install hyperfine

Tool selection:

Key anti-patterns:

• Never load test production without rate limits
• Always include ramp-up stages — instant VU spikes are unrealistic
• Always define thresholds — measurements without thresholds are not tests
• Test with production-scale data volumes, not toy datasets
• Include p99 in thresholds — outliers matter for user experience