Resource allocation method and system

A distributed computing system for executing a plurality of tasks of an application, comprising:

• a server having a grid resource manager (GRM);
• a remote user of the distributed computing system configured to submit a request for execution of the application to the GRM;
• and
• a cluster of nodes forming a part of or constituting the distributed computing system, the cluster having a cluster resource manager (CRM);
• wherein the GRM is operable to request that the CRM authorize the remote user to access any idle nodes of the cluster and, if the CRM authorizes the remote user to access the idle nodes, to submit the tasks and a plurality of processes to the CRM, and
• wherein the CRM is operable to:
• load as many of the processes as there are idle nodes without allocating any to task to any particular idle node,
• sequentially schedule the tasks to the processes,
• execute the processes until the node in which the process is loaded is preempted by a local allocation request, and
• reschedule any preempted tasks to another process.

A non-transitory computer readable storage medium provided with program data that, when executed on a distributed computing system, implements a method comprising:

• in a server having a grid resource manager (GRM), receiving a request for execution of a plurality of tasks of an application from a remote user of the distributed computing system having a cluster resource manager (CRM) and a cluster of nodes forming a part of or constituting the distributed computing system;
• requesting, by the GRM, that the CRM authorize the remote user to access any idle nodes of the cluster;
• if the CRM authorizes the remote user to access the idle nodes of the cluster, submitting, by the GRM, the tasks and a plurality of processes to the CRM;
• the CRM then:
• loading as many of the processes as there are idle nodes without allocating any task to any particular idle node;
• sequentially scheduling the tasks to the processes;
• executing the tasks by the processes until the node in which a particular process is loaded is preempted by a local allocation request, and
• rescheduling any preempted tasks to another process.