Load Testing & Benchmarking

Load / stress testing (APIs and apps)

• ★ Grafana k6 — JS-scripted load tests; runs as a single binary; cloud option for distributed runs. The default in 2026.
• ★ Artillery — JS-config load tests; great for HTTP, WebSocket, Socket.io. Hosted reporting available.
• Autocannon — fastest "give me a quick number" Node HTTP benchmarker. CLI + library.
• Vegeta — Go-based, simple, great for sustained constant load.
• wrk / wrk2 — venerable C HTTP benchmarkers.
• Locust — Python-based; nice for complex scenarios.
• JMeter — Java; massive feature surface; old-school.
• Gatling — Scala; great for big-load realistic-scenario tests.

Browser / real-user-simulating

• Playwright — repurpose your E2E tests for soak runs.
• Sitespeed.io — synthetic browser monitoring with real Chromium.
• WebPageTest — single-page deep performance reports; free public instance.

Function-level micro-benchmarks

• ★ Mitata — fastest TS micro-benchmarker; produces reliable numbers.
• ★ Tinybench — Vitest-friendly benchmark library.
• benchmark.js — older, ubiquitous.
• hyperfine (CLI) — for benchmarking shell commands.
• isolated-vm — Node module for sandboxed comparison runs.

Database load testing

• pgbench — Postgres built-in.
• HammerDB — multi-DB stress.
• sysbench — Linux benchmark suite.
• YCSB — distributed DB benchmarks.

Profiling production performance

• Clinic.js (Doctor / BubbleProf / Flame) — Node profiling toolkit.
• 0x — flame graphs for Node.
• --inspect with Chrome DevTools — built-in profiling.
• Speedscope — beautiful flame graph viewer.
• See also Observability for production traces.

CI integration

• k6 Action for GitHub Actions — run k6 in CI on every PR.
• Artillery Cloud — store results, diff over time.
• Lighthouse CI — for browser performance budgets in CI.
• Foundry — newer, GitHub-native perf budgets.

Patterns to know

• Spike, soak, stress, capacity — different test shapes; k6 / Artillery support all.
• Think time — between requests; matters for realism.
• Steady-state baseline — always run the same baseline; compare changes against it.
• Don't load-test prod without a heads-up — coordinate with infra and your CDN provider.
• Bench in the same env multiple times — variance kills small claims.

Pick this if…

• Default API load test: k6.
• WebSocket-heavy: Artillery.
• One-off "is this fast?": Autocannon.
• Function micro-bench: Mitata or Tinybench.
• Browser perf budgets in CI: Lighthouse CI.
• Profile a slow Node service: Clinic.js Flame.