compute-cluster

Distributed Computation for NodeJS

How can you build a responsive and robust nodejs server that does some heavy computational lifting? Some node libraries (like the awesome node-bcrypt) do their own threading internally and combine that with an async API. This allows libraries to internally thread their calls and use multiple cores.

While this is pretty awesome, it is significant work for library implementors, and as this pattern becomes rampant, the application author loses fine grained control over the resource usage of their server as well as the relative priority of compute tasks.

If you just naively run computation on the main evaluation thread, you're blocking node.js from doing anything else and making your whole server unresponsive.

The solution?

node-compute-cluster is a tiny abstraction around a group of processes and the built-in IPC introduced in NodeJS 0.6.x. It provides a simple API by which you can allocate and run work on a cluster of computation processes. This allows you to perform multiprocessing at a more granular level, and produce a responsive yet efficient computation server.

Installation

$ npm install compute-cluster

Usage

First you write your main program:

const computecluster = require('compute-cluster');

// allocate a compute cluster
var cc = new computecluster({
  module: './worker.js'
});

var toRun = 10

// then you can perform work in parallel
for (var i = 0; i < toRun; i++) {
  cc.enqueue({}, function(err, r) {
    if (err) console.log("an error occured:", err);
    else console.log("it's nice:", r);
    if (--toRun === 0) cc.exit();
  });
};

Next you write your worker.js program:

process.on('message', function(m) {
  for (var i = 0; i < 100000000; i++);
  process.send('complete');
});

All done! Now you're distributing your computational load across multiple processes.

API

Constructor - new require('compute-cluster')(<options>);

Allocates a computation cluster. Options include:

• module - required the path to the module to load
• max_processes - the maximum number of processes to spawn (default is ciel(#cpus * 1.25))
• max_backlog - the maximum length of the backlog, -1 indicates no limit (default is 10 * max_processes) an error will be returned when max backlog is hit.
• max_request_time - the maximum amount of time a request should take, in seconds. An error will be returned when we expect a request will take longer.

Example:

var cc = new require('compute-cluster')({
  module: './foo.js',
  max_backlog: -1
});

Event: 'error'

An error event will be emited in exceptional circumstances. Like if a child crashes. Catch error events like this:

cc.on('error', function(e) { console.log('OMG!', e); });

Default behavior is to exit on error if you don't catch.

Events: 'debug' or 'info'

Events raise that hold an english, developer readable string describing the state of the implementation.

cc.enqueue(, [cb])

enqueue a job to be run on the next available compute process, spawning one if required (and max_processes isn't hit).

args will be passed into the process (available via process.on('message', ...)).

cb is optional, and will be invoked with two params, err and response. err indicates hard errors, response indicates successful roundtrip to the compute process and is whatever the decided to process.send() in response.

cc.exit([cb])

Kill all child processes, invoking callback (with err param) when complete.

LICENSE

Copyright (c) 2011, Lloyd Hilaiel [e-mail]

Permission to use, copy, modify, and/or distribute this software for any purpose with or without fee is hereby granted, provided that the above copyright notice and this permission notice appear in all copies.

THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.