An HTTP layer that isn't your cold start

Start-up, measured

Cold start is the time from spawning the process to the first successful answer on its port. In campaign 4 every server ran on the same DigitalOcean c-4 droplet as simoom, in the same round.

Campaign 4, 2026-11-12, a selection of the 21 servers; the median of the 20 others started 6.4x slower than simoom's default build. Program size is shown only for self-contained executables. With HTTP/3 on, simoom starts every event loop at once instead of one, part of why the full server starts later than the default build. The platform's own sandbox start comes on top for every runtime. All servers: start-up, memory at rest. On a workstation, against the server it replaced, see Cameleer's old server.

Why it starts fast

The start-up work behind these numbers is described in runtime. Build options: features and size.

A function's HTTP server

Container platforms such as Cloud Run send requests to a port; listen on it and answer. A graceful stop tells clients to finish when the platform retires the instance.

use simoom::{Backend, Reply, Request, Server};
use std::time::Duration;

let server = Server::builder()
    .listen("0.0.0.0:8080")
    .threads(2)               // 0 = one loop per core (the default)
    .backend(Backend::IoUring) // Linux 6.1+, falls back to epoll
    .service(|_: &mut Request<'_>, r: Reply<'_>| r.body_static(b"ok"))
    .start()
    .unwrap();
// ...
server.stop(Duration::from_secs(5)); // tells clients to finish, then waits up to 5 s
server.join();

simoom serves HTTP on a socket. On AWS Lambda that means a container image or an HTTP adapter in front of the port: simoom does not speak the Lambda Runtime API itself. Cameleer, which uses simoom, is its own Lambda runtime: see Cameleer on AWS Lambda.

Measure it yourself

Build the hello example in release mode and time the first answer on your platform.

cargo add simoom
cargo run --release -p simoom --example hello -- 127.0.0.1:8080   # from a checkout of the repository