Intel® Cluster Ready Head Node Used in Sub-clusters

Intel® Cluster Ready head node used as a compute node in sub-clusters



The Intel® Cluster Ready specification defines a sub-cluster as a group of similar nodes within a cluster that allow partitioning of specific uses of the cluster, while providing a single system for ease of operation, management, and/or operation. A cluster can have many sub-clusters, and it is possible for the head node to also be a compute node within one or more sub-clusters.

A head node that is also a compute node is sometimes referred to as a “mixed-head node”. The mixed-head node in a sub-cluster must contain a super-set of the software stack for a compute node in that sub-cluster. That means every system file on a compute node must have an identical copy on the head node, except for those files that are used for local configuration, are node-specific, or are normally altered after provisioning. The head node is allowed to have additional files installed that don’t exist on compute nodes. If a cluster has multiple sub-clusters, the head node can be a compute node in as many different sub-clusters where this requirement is met.

It is also possible for a head node to function as a mixed-head node in one or more sub-clusters and function only as a dedicated head node in the remaining sub-clusters.
For more complete information about compiler optimizations, see our Optimization Notice.