Add Koa: async/await Node.js framework by the Express team (~36k ⭐)

[BennyFranciscus]

Koa

Language: JavaScript (Node.js)
Stars: ~35,747
Repo: https://github.com/koajs/koa

Koa is the spiritual successor to Express, designed by the same team (TJ Holowaychick) with a cleaner, more modern foundation built entirely on async/await. With ~36k stars, it's one of the most popular Node.js frameworks.

Why Koa in HttpArena?

The Node.js framework lineup now covers the full evolution:

• bare node — raw http module, no framework
• Express — the OG, callback-based middleware
• Koa — Express team's rewrite with async/await
• Fastify — schema-based, performance-focused
• Hono — Web Standards (Request/Response API)
• ultimate-express — Express API on µWebSockets

The key comparison everyone wants: Koa vs Express vs Fastify. Same ecosystem, different middleware philosophies. Express chains callbacks, Koa uses async/await with cascading middleware, Fastify uses schema-based validation for speed.

Implementation

• Koa 2.x with koa-router
• Cluster mode (one worker per CPU core)
• Pre-computed JSON + gzip buffers (/json, /compression)
• Raw body stream reading for POST endpoints (zero-overhead, no body-parser)
• better-sqlite3 with mmap for /db
• HTTP/2 via native http2 module on port 8443
• Static files served from in-memory cache

All HttpArena endpoints implemented

• /pipeline — simple text response
• /baseline11 — GET/POST with query param summing
• /baseline2 — GET with query param summing
• /json — JSON serialization benchmark
• /compression — gzip compressed JSON
• /db — SQLite query benchmark
• /upload — POST body size reporting
• /static/{filename} — H2 static file serving

cc @dead-claudia @jonathanong @juliangruber — would be cool to see how Koa stacks up against Express and Fastify in HttpArena!

[BennyFranciscus]

CI is consistently hitting Bind for 0.0.0.0:8080 failed: port is already allocated on the self-hosted runner. Same issue affecting PR #75 (Rocket). The Docker image builds successfully — it's a stale container on the runner holding port 8080.

A quick docker ps + docker stop on the runner should clear it. Build and implementation are ready to go! 🚀