CODEWITHSUNDEEP
node.js
John
John

Reputation: 51

Node.js - HTTP Keep-Alive and thread pool

I have a few questions that I hope you can help with.

Node.js use single threaded event loop and there is a thread pool that consist of 4 threads by default.

Let say i'm using Node.js as a web server and Keep-alive is enabled and is set to be idle up to 2 seconds before closing the connection.

if there is 8 unique clients trying to visit my website page, will only 4 be able to see the page until one of clients in the threads close the connection?

If not can you please explain why?

Upvotes: 1

Views: 436

Answers (1)

m1ch4ls
m1ch4ls

Reputation: 3435

Answer is no - one thread can serve many clients.

TLDR; It's because Node.js is using asynchronous I/O

Under the hood Node.js uses two technologies - libuv and v8

  • v8 is Javascript engine used by Google Chrome and other browsers (not interesting for us now).
  • libuv is a library that provides the event loop and asynchronous I/O capabilities

What is the event loop?

Generally speaking it's a loop that's polling or querying some (usually external) source of events and a callback for each event that is executed when event is received. Something like this:

while (running) {
  event = get_event();
  callback = get_callback(event)
  callback();
}

If we do fs.read(), socket.write(), setTimeout()...

Under the hood this is going to 'schedule' events with callbacks that are about to be completed (using the callbacks) in future - but the code is not going to wait for each individual one, instead only the first event will be waited for (no matter which one completes first).

So get_event will return first event that completes.

What is asynchronous I/O?

For example reading from a file is usually so called blocking operation - meaning we ask x bytes from a file and we are going to block execution until OS to gives us the bytes. This is synchronous I/O.

Using asynchronous I/O it's possible to ask OS to give us the bytes and not block so we can do something else. When the bytes are ready we will get notified by event loop and process them - so called non-blocking operation.

Node is doing exactly that - it's using mostly non-blocking operations to achieve concurrency on a single thread. Manipulating files and using network can be done using OS's asynchronous I/O.

It's extremely useful because I/O operations usually make majority of execution time so our code instead of waiting is doing something useful - like handling connection from 8th or 100th client.

Upvotes: 1

Related Questions