Autocannon

Overview

Autocannon is a fast HTTP benchmarking tool written in Node.js. It generates high request volumes to test API throughput, latency percentiles, and error rates. Lighter than k6 or Artillery for quick benchmarks.

Instructions

Step 1: CLI Usage

npm install -g autocannon

# Basic benchmark: 10 connections, 10 seconds
autocannon http://localhost:3000/api/projects

# Custom parameters
autocannon -c 100 -d 30 -p 10 http://localhost:3000/api/projects
# -c 100: 100 concurrent connections
# -d 30:  30 seconds duration
# -p 10:  10 pipelined requests per connection

# POST with body
autocannon -c 50 -d 20 -m POST \
  -H "Content-Type=application/json" \
  -H "Authorization=Bearer token123" \
  -b '{"name":"test","email":"test@example.com"}' \
  http://localhost:3000/api/users

Step 2: Programmatic

// benchmark.ts — Scripted benchmark with comparison
import autocannon from 'autocannon'

async function benchmark(title: string, url: string, opts = {}) {
  const result = await autocannon({
    url,
    connections: 100,
    duration: 10,
    ...opts,
  })

  console.log(`\n--- ${title} ---`)
  console.log(`Requests/sec: ${result.requests.average}`)
  console.log(`Latency avg:  ${result.latency.average}ms`)
  console.log(`Latency p99:  ${result.latency.p99}ms`)
  console.log(`Throughput:   ${(result.throughput.average / 1024 / 1024).toFixed(1)} MB/s`)
  console.log(`Errors:       ${result.errors}`)
  console.log(`Timeouts:     ${result.timeouts}`)

  return result
}

// Compare endpoints
await benchmark('GET /projects (cached)', 'http://localhost:3000/api/projects')
await benchmark('GET /projects (no cache)', 'http://localhost:3000/api/projects?nocache=1')
await benchmark('POST /tasks', 'http://localhost:3000/api/tasks', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ title: 'Benchmark task', projectId: 'proj_1' }),
})

Step 3: CI Integration

# .github/workflows/perf.yml — Performance regression detection
- name: Start server
  run: npm start &
  env: { NODE_ENV: production, DATABASE_URL: ${{ secrets.TEST_DB_URL }} }

- name: Wait for server
  run: npx wait-on http://localhost:3000/health

- name: Benchmark
  run: |
    npx autocannon -c 50 -d 10 -j http://localhost:3000/api/projects > bench.json
    RPS=$(node -e "console.log(JSON.parse(require('fs').readFileSync('bench.json')).requests.average)")
    echo "Requests/sec: $RPS"
    if [ "$RPS" -lt 1000 ]; then echo "Performance regression!" && exit 1; fi

Guidelines

• Run benchmarks on isolated environments — other processes affect results.
• Focus on p99 latency, not average — outliers matter more for user experience.
• Start with low concurrency and increase — find where performance degrades.
• Benchmark after code changes to catch regressions early.
• Autocannon is for HTTP benchmarks. For load testing with scenarios, use k6.