Compute Nodes

The bulk of the nodes on an HPC cluster are compute nodes. These are the heavy-processing nodes where jobs submitted to the cluster's scheduler are actually run. Compute nodes frequently have better interconnects than the head nodes, and in some cases (such as bluegene) will have a substantially different CPU architecture.

Compute nodes are seldom directly accessible via SSH. Users who are running interactive jobs can SSH to the compute nodes on which the job is running, but the interactive job itself must be submitted via script on the head node.