Figure 3

Communication between processes on the master server. Communication between processes on a master server (configured strictly as a master server, not also as a worker). At the top, the down pointing arrow indicates a client request for the IP of an available worker computer (geval). The Apache web server (large oval) accepts the request via a PHP script, which gets the IP address of the currently best available worker from the worker workload information file (rectangle) and returns it to the client. The worker workload file is updated periodically by a crontab job which polls the necessary information form the workers. Each new worker (outside the Master rectangle on the right) willing to participate in mGrid, executes a volunteer command, which calls a PHP script on the master node (going through Apache) to update the workers workload file. Note that if a given worker is already registered with this master, the request will simply be ignored. If a worker is new, then before adding it to the list the master will request the workload information from the potential worker and only upon successful receipt of the workload information, which verifies that the worker responds to the HTTP requests, will this volunteering computer be added to the pool of workers.