app.js

  var
      express = require( 'express' ),
	  connect = require('connect'),
      app     = express();
	  count   = 0;

  app.configure( function() {
	app.use(connect.urlencoded());
	app.use(connect.json());
  });

  function log( what ) { console.log( what ); }

  app.get( '/', function( req, res ) {
	  count++;
      res.end(count+' bottles of beer on the wall!');
  });

  app.listen( 8080 );

cluster.js

var cluster = require('cluster')

if ( cluster.isMaster ) {
  for ( var i=0; i<8; ++i )
    cluster.fork();
} else {
  var
      express = require( 'express' ),
      connect = require( 'connect' ),
      app     = express(),
      count   = 0;

  app.configure( function() {
	app.use(connect.urlencoded());
	app.use(connect.json());
  });

  function log( what ) { console.log( what ); }

  app.get( '/', function( req, res ) {
     count++;
     res.end(count+' bottles of beer on the wall!');
  });

  app.listen( 8080 );
}

Try using strong-supervisor, you can control worker count from command line (or strongops): https://github.com/strongloop/strong-supervisor

Not to tout our modules too much... but better than changing that 8 into other numbers.

Definitely true. Was just testing the scaling limits of node cluster out of the box on version 0.10, and the performance gains weren't useful because of the native process scheduling. Varnish, nginx, HA proxy, etc are better solutions for this while we wait and see what kind of gains we can get from 0.12