Performance

Benchmarks

Tested with wrk -t 4 -c 100 -d 30s:

Mode	Requests/sec	Latency (avg)
Spawn + Keep-Alive	45,000+	~2.2ms
Spawn (no Keep-Alive)	5,000-8,000	~3.5ms
WorkerPool	7,500-8,500	~12ms

Keep-Alive

Enabled by default. Reuses TCP connections for multiple requests.

let config = Config.default()
    .with_keep_alive(true)           // Default: true
    .with_keep_alive_timeout(30000)  // 30s idle timeout
    .with_max_requests_per_connection(100);

Keep-Alive provides ~8x throughput improvement.

Connection Limiting

Fast 503 rejection when at capacity:

let config = Config.default()
    .with_max_connections(10000);

Rejected clients receive:

HTTP/1.1 503 Service Unavailable
Retry-After: 1
Connection: close

Optimization Tips

Use Spawn for I/O

Default spawn mode is faster for I/O-bound work:

// Good: I/O-bound in spawn
app.get("/users", |ctx| {
    let users = db.query("...");  // Yields during I/O
    try? ctx.json(users);
});

Use WorkerPool Sparingly

Only for genuinely CPU-intensive work:

// Good: CPU-bound in WorkerPool
app.get("/hash", |ctx| {
    let hash = bcrypt(data);  // Actually CPU-intensive
    ctx.string(hash);
}).worker();

Disable WorkerPool If Unused

If no routes need WorkerPool:

let config = Config.default()
    .with_worker_pool(false);  // Simpler architecture

Resource Usage

Mode	Memory/Connection	CPU	Threads
Spawn	~1KB	Single core	1
WorkerPool	~2KB	Multi-core	1 + workers

Scaling

For maximum throughput:

• Enable Keep-Alive (default)
• Use spawn mode for I/O-bound routes
• Reserve WorkerPool for CPU-intensive routes
• Tune max_connections for your deployment