Developing and comparing my fast XML parser with other libraries, I came across a new project for using the worker process very easy and effective.

Piscina is created by some developers of node.js. And it has absolutely surprised me.

Also: recently I had an argument with a colleague, that some task was so much faster after implementing it in golang and not in javascript. I told him that we also could use workers and inter-process communication (IPC) to be much faster using node.js. But implementing was too complex and out of scope for the time being.

Now, with piscina implementing functionalities in a worker process became a piece of a cake and the process now runs just as fast directly in JS. Because the real limitation lies not on the CPU but on networking.

Using Piscina

First, we need a worker.js file. It exports a single function. That you want to execute in a separate thread.

const sleep = ms => new Promise(resolve => setTimeout(resolve, ms));

module.exports = async ({ a, b } => {
await sleep(100);
return a + b;

Of course, you will not sleep, but do some processing work and return that result. Piscina is made for bettter usage of CPU. In tasks that a single node.js process can process concurrently, such as db queries and API calls, you don’t need this module.

The CPU processing task could include image processing, encryption, and decryption, or parsing data. The process can be implemented with sync or async function.
The result can be returned back to the main process or let’s say uploaded into the cloud. Whatever is needed.

In the main process, to use the worker module, you do the following:

const Piscina = require('piscina');

const workerPool = new Piscina({
filename: __dirname + '/worker.js'

(async function() {
const result = await workerPool.runTask({ a: 4, b: 6 });
console.log(result); // Prints 10

This is basically it. The worker pool can be renamed, .runTask can be called let’s say in an API handler of express or graphql. The Argument has to be a single object. But it can have any number of props and dept.

For configuration, you can pass more options into the Piscina constructor. And the options did not net me down. They let you pick the number of threats, behavior for pooling to save some memory, limit the worker memory, and processing time. Really everything I could think of, to be done differently by the library, had a reasonable configuration available.


I think this will open up many options to make processes and performance in node.js applications better.

For the txml xmp parser however, I decided not to integrate the module, because when used by the application developer, even more CPU heavy processing of the data can be moved from the main thread into the worker.

What do you think of piscina? Do you have an Idea what you can use it for? You can leave a comment on the article at

  1. 1. Using Piscina
  2. 2. Future