Kito - benchmarks

Performance benchmarks comparing Kito against popular Node.js web frameworks.

---

📊 Latest Results

Note: Benchmarks are automatically updated on each release.

---

🎯 Benchmarked Frameworks

• Kito - High-performance TypeScript framework written in Rust
• Fastify - Fast and low overhead web framework
• Hono - Fast web framework for the edge
• Express - Fast, unopinionated web framework
• Koa - Expressive middleware framework
• Restify - REST API framework
• TinyHTTP - Lightweight Express-like framework
• Hapi - Rich framework for building applications

---

🔧 Benchmark Configuration

{
  connections: 100,
  pipelining: 10,
  duration: 30, // seconds
  workers: undefined // auto
}

All frameworks are tested with a simple "Hello World" endpoint to measure raw performance.

---

🚀 Running Benchmarks

Prerequisites

pnpm install

Run All Benchmarks

pnpm bench:run basic

This will:

1. Start each framework on sequential ports (3000, 3001, ...)
2. Run wrk load tests
3. Generate comparison charts
4. Save detailed results to results/data/

Mixing Node and Bun runtimes

Some framework adapters only work on Node.js while others can run under Bun. You can pin the runtime per framework in config.ts:

frameworks: [
  { name: "kito", runtime: "bun" },
  { name: "elysia", runtime: "bun" },
  { name: "restify", runtime: "node" },
  "koa", // uses whichever runtime you launch the runner with
];

• Default (string) entries execute inside the same runtime that starts runBench.ts.
• runtime: "bun" executes via pnpm dlx bun run ..., so make sure pnpm is available (it will download Bun on demand).
• runtime: "node" executes via pnpm dlx tsx ..., so no local CLI install is required beyond pnpm.

---

📁 Project Structure

bench/
├── cases/
│   └── basic/           # Basic "Hello World" benchmarks
│       ├── kito.ts
│       ├── fastify.ts
│       ├── express.ts
│       └── ...
├── results/
│   ├── charts/          # Generated benchmark charts
│   └── data/            # Raw JSON results
├── utils/
│   ├── http.ts          # Wrk runner
│   └── chart.ts         # Chart generation
├── config.ts            # Benchmark configuration
└── runBench.ts          # Main benchmark runner

---

📈 Metrics Measured

• Requests/sec - Average requests per second
• Latency (ms) - Average response time in milliseconds
• Throughput (bytes/sec) - Average data throughput

---

🎨 Adding New Benchmarks

1. Create a new folder in cases/ (e.g., cases/routing/)
2. Add framework implementations following the pattern:

export function start(port: number): { stop: () => void } {
  const app = createServer();
  
  app.get("/", handler);
  app.listen(port);
  
  return {
    stop: async () => app.close(),
  };
}

3. Run your benchmark:

pnpm bench:run routing

---

📊 Understanding Results

Requests/sec (Higher is better)

Number of requests the server can handle per second.

Latency (Lower is better)

Time taken to receive a response. Includes network and processing time.

Throughput (Higher is better)

Amount of data transferred per second.

---

⚙️ Configuration

Edit config.ts to customize:

export default {
  frameworks: ["kito", "fastify", "hono", ...],
  connections: 100,      // Concurrent connections
  pipelining: 10,        // Requests per connection
  duration: 30,          // Test duration in seconds
  chart: {
    enabled: true,
    output: "results/charts/result.png"
  }
};

---

🔬 Methodology

• All tests run on the same machine with identical conditions
• Each framework uses its recommended setup and defaults
• Tests measure only the framework overhead, not business logic
• Results are averaged over the test duration

---

📝 Notes

• Benchmarks are indicative and may vary based on hardware
• Real-world performance depends on application complexity
• These tests measure raw throughput, not production scenarios
• All frameworks are excellent choices for different use cases

---

🤝 Contributing

Found an issue with the benchmarks? Want to add a new framework?

1. Fork the repository
2. Add your changes
3. Run the benchmarks locally
4. Submit a pull request

---

📄 License

Licensed under the MIT License.

---