Node.js如何处理更多并发请求？

Question

I have been reading the following argument between apache and nodejs , which I do not understand.

Apache is thread and process based ie each request is handled by a separate thread or process (depending upon configuration), which means if the process is waiting for the I/O, whole thread is blocked.

• What did I loose here? I will also wait when running nodejs, even if it means waiting asynchronously. Isn't it?

Every nodejs instance runs in a single thread and due to its asynchronous nature, it can handle far more number of concurrent requests as compared to apache

• Could anyone explain me how?

I have been trying to understand how nodejs scores and its main strong point, but I just do not get the essence of it.

Answer 1

Apache is totally thread based as you described.WHile nodejs is event based.It uses event loop. Now lets take a example. Suppose you have thousands of requests.In apache it will spawn thread for each and every request(Which is obviously have a bottleneck,You cant scale too much vertically to support this.) While in node js 1 instance will have single thread.It doesnt blocks for an i/o.In mean time while i/o is there it can accept and start processing(cpu related) on the new request. Concurrency is way higher in nodejs.As it allows many requests to be handled at the same time.However i donot have any benchmarking on whether request processing is bit higher in nodejs or not.But yes in terms of concurrency nodejs will defeat any of its competitors.

Answer 2

For better understanding let's start from this picture:

And there are following steps:

1. Client sends HTTP requests to NodeJS server.
2. Requests come to Event Loop (single thread), which is woken up by operation system. It passes every request/response as Javascript closure to relevant worker functions with callbacks.
3. After that you do some operations (long-running jobs etc.) inside non-blocking workers.
4. As soon as it's done, response is sent back to main thread via callback.
5. In the end Event Loop returns result to client.

This way NodeJS main thread doesn't perform any operations by itself and only delegates management process to non-blocking workers where you can do your magic stuff concurrently.

Note: NodeJS can only receive or respond to a single request at a time. However, during the same time it can process multiple requests.

---

For example you have 100000 requests at a time. It really makes sense blocking or non-blocking operations you do in your requests handlers.

Blocking operations will block javascript processing since it's synchronous action takes place. This behavior can lead to server's halt.

Non-blocking operations will happen immediately, so everything will depend on vertical scaling . It means it will receive/respond as fast as it's possible for current hardware configuration.

To increase your server performance:

• Horizontally , you might want to play with NodeJS clusters , load-balancer configurations etc.
• Vertically , you might want to play with V8 engine. Look here to find more V8 flags. Or you can upgrade your hardware of course.

For example, max_old_space_size :

–max-old-space-size=8192

Increases the limit for each V8 node process to use max 8Gb of heap memory instead of the 1,4Gb default on 64-bit machines(512Mb on a 32-bit machine).