Configuring Worker Threads

For enable multi-threading you can use fileSystemRouter({ workers: true }) which creates a thread pool up to the number of available logical processors or use specify the number of workers using fileSystemRouter({ workerCount: 4 }).

src/main.ts

import { fileSystemRouter } from "keiro/node";
import http from "node:http";

http
  .createServer(
    fileSystemRouter({
      workers: { workerCount: 4 },
    }),
  )
  .listen(3000);

Worker pool

The workers are spawn inside a WorkerPool, you can specify the worker using the pool option:

const enum WorkerPoolType {
  Dynamic = 1,
  Fixed = 2,
}

By default it uses the WorkerPoolType.Fixed pool, you can also pass function that returns a WorkerPool:

(workerCount: number, filename: string, options?: WorkerOptions) => WorkerPool

How it works?

Internally we use node Worker API, and create a pool of workers. When a request comes we take an available worker and serialize the request and send it to the worker and then deserialize the response of the worker and send it as a response.

Each worker have its own router which is used to handle the request.

Workers are isolated!

Workers cannot directly share data between each other, you cannot share global state, so each request should be handle independently.

Is not multi-threading always faster?

No.

We recommend profiling your app before enabling workers, unless you are handling a big number of request you may not see any difference when enabling workers, it may be even slighly inneficient to using workers for a few number of request per second.

Checkout the Worker threads example.