CODEWITHSUNDEEP
multithreadingapachenode.jsiiscomputation
user265445
user265445

Reputation: 341

Is Node.js suited for compute intensive web services

I am going to create a set of web services that include very compute intensive code. The calculations can run from 1/2 a second to 20 seconds. The computations exist in C code.

I would probably not have more than 20 simultaneous web service requests of which only a couple would be the longer running calculations.

My understanding is that node.js runs in a single thread so I would have to write a node.js addon which would interface to my C code. The C code would then have to execute in its own thread. I would use the thread pools provided by node.js to run these computations asynchronously. I would run this on a multi-core machine to maximize performance.

Does the architecture that I described sound correct and is node.js suited to do this? Would Apache or IIS be better at an application like this?

Upvotes: 1

Views: 553

Answers (1)

Plato
Plato

Reputation: 11052

You can make a native C program, and have node execute the binary, and callback upon completion. I think there are a couple methods, here is an example of child_process.exec(command, [options], callback) copied from nodejitsu:

var childProcess = require('child_process'),
     ls;

 ls = childProcess.exec('ls -l', function (error, stdout, stderr) {
   if (error) {
     console.log(error.stack);
     console.log('Error code: '+error.code);
     console.log('Signal received: '+error.signal);
   }
   console.log('Child Process STDOUT: '+stdout);
   console.log('Child Process STDERR: '+stderr);
 });

 ls.on('exit', function (code) {
   console.log('Child process exited with exit code '+code);
 });

edit
If you need to pass input to your process after starting it, use child_process.spawn instead.

Node can handle apache's tasks well; I use express to host files and dynamic pages; my application logic is javascript code that is called by Express when particular HTTP requests are made to particular routes.

You might bog your machine down unless you cap simultaneous computations. I have no knowledge or comment on thread pools. caolan/async can help manage concurrent computations (and other control flow tasks.)

Upvotes: 1

Related Questions