Compute Cluster Architecture

A compute cluster is a top-level organizational unit. Each cluster consists of a set of nodes, a queue, applications, and jobs.

The following diagram illustrates the compute cluster architecture.

Compute cluster architecture

A node is a single computer with one or more processors. The system administrator adds and removes nodes. Applications can enumerate, approve, pause, and resume nodes through the ICluster interface. Applications can query node properties through the INode interface.

All nodes in a cluster are part of the same domain. Data sources such as a database or file are accessible from each node. License servers are also accessible from each node to verify that each node that will run an application has a license for the application.

The head node manages the cluster and its resources. Jobs are submitted to and scheduled by the head node.

The cluster contains a single queue, which is the organizational unit that contains queued, running, and finished jobs. Finished jobs are periodically removed from the queue. For more information, see Job Scheduler Architecture.

Related topics

About CCP

 

 

Show: