$devvkit learn --librarie wrk-&-hey-guide

wrk & hey Guide

[benchmark][http][cli][latency]

Performance & Profiling

Install

# wrk
brew install wrk
sudo apt install wrk

# hey (Go)
brew install hey
# or: go install github.com/rakyll/hey@latest

wrk is a multithreaded HTTP benchmarking tool that uses a LuaJIT scripting engine for custom request generation. It's perfect for measuring raw throughput and latency distribution (latency, req/s, transfer rate) on a single endpoint.

hey (formerly boom) is built in Go and is simpler than wrk: it supports POST with body, custom headers, and HTTP/2 out of the box. Use it when you need to benchmark API endpoints with JSON payloads.

Both tools report latency percentiles (p50, p75, p90, p99, max), request rate, and error counts. For distributed load generation across multiple machines, use vegeta or k6 instead. The wrk2 fork adds constant throughput mode.

wrk Basics

wrk: basic· Simple throughput test.

wrk -t12 -c400 -d30s http://localhost:8080
# -t: threads, -c: connections, -d: duration

wrk: keep-alive· Disable keep-alive.

wrk -t4 -c100 -d30s --latency -H "Connection: close" http://localhost:8080
# --latency prints detailed percentile distribution

wrk: custom timeout· Set request timeout.

wrk -t2 -c50 -d10s --timeout 5s http://localhost:8080/api

wrk Scripting

wrk: POST via Lua· POST request with body.

-- post.lua
wrk.method = "POST"
wrk.body = '{"name": "test"}'
wrk.headers["Content-Type"] = "application/json"

wrk -t4 -c100 -d30s -s post.lua http://localhost:8080/users

wrk: auth token· Dynamic header per request.

-- auth.lua
token = os.getenv("TOKEN")
request = function()
  wrk.headers["Authorization"] = "Bearer " .. token
  return wrk.format("GET", "/api/users")
end

TOKEN=abc123 wrk -t2 -c10 -d10s -s auth.lua http://localhost:8080

hey Basics

hey: basic· Simple GET benchmark.

hey -n 10000 -c 100 https://api.example.com/users
# -n: total requests, -c: concurrent workers
hey -z 30s -c 50 https://api.example.com/users

hey: POST JSON· Benchmark POST endpoint.

hey -n 5000 -c 50 -m POST \
  -H "Content-Type: application/json" \
  -d '{"email":"test@test.com","password":"123456"}' \
  https://api.example.com/login

hey: custom headers· Multiple headers.

hey -n 1000 -c 10 \
  -H "Authorization: Bearer token123" \
  -H "X-Request-Id: ${RANDOM}" \
  https://api.example.com/protected

hey: HTTP/2· Force HTTP/2.

hey -n 10000 -c 100 -h2 https://example.com/grpc-endpoint

Analysis

Parse results· Extract key metrics.

# Extract p99 latency from wrk output
wrk -t2 -c50 -d10s --latency http://localhost:8080 | grep "99%"

# hey JSON output for programmatic use
hey -n 1000 -c 50 -o json https://example.com | jq '.latencyPercentiles