Go HTTP Performance Server

High-throughput HTTP server for load testing. It responds with a tiny JSON body and is optimized for maximum requests per second (RPS) under heavy pipelining.

Response format

• HTTP/1.1 200 OK
• Body: {"r":<0..9999>} with a random value per request

This server uses a minimal HTTP parser to count \r\n\r\n boundaries, which is intentionally fast and sufficient for benchmarking.

Features

• Extremely fast response path using precomputed HTTP responses
• Keeps connections alive and supports aggressive pipelining
• Lightweight per-second RPS logging
• Simple, dependency-free Go codebase

Requirements

• Go 1.25+
• macOS or Linux

Build and Run

go build -o server .
./server

The server listens on :8007.

Logging

The server prints a startup line and then logs RPS once per second:

2026/02/03 13:02:10 rps=7571783 total=76440064

If running in the background, a common pattern is:

./server >/tmp/fast-metrics.log 2>&1 &
tail -f /tmp/fast-metrics.log

Load Testing (wrk)

pipeline.lua is included and uses an aggressive pipeline size.

wrk --latency -t12 -c 1000 -d 10s -s pipeline.lua http://127.0.0.1:8007/metrics

To push higher RPS, run multiple wrk processes in parallel:

for i in 1 2 3 4; do
  wrk -t12 -c 1000 -d 10s -s pipeline.lua http://127.0.0.1:8007/metrics &
done
wait

Tests

If your system restricts access to the default Go build cache, run:

GOCACHE=/tmp/go-build go test ./...

Notes

• This server is designed for performance testing, not full HTTP compliance.
• The response payload is intentionally minimal to maximize throughput.