Cluster terminology¶

Glossary of terms¶

There is some terminology with which you should familiarize yourself to get the most out of this documentation.

batch Most often used in the phrase 'run in batch', which means that the instructions to run are entered into a file and are then run without further interaction with the user.

cluster A collection of individual computers to which jobs are sent by some sort of scheduling and job management software.

Cluster Schematic

High Performance Computing Cluster Architecture

core A processing unit that is part of a physical chip inside a computer.

CPU A physical chip inside a node that performs computations.

job A set of instructions in the form of text commands that will execute commands. These are typically in a file, which is submitted to the cluster scheduler to be run in batch.

job manager Software that starts and stops jobs on a cluster. Typically it works in conunction with the scheduler. The job manager insures that jobs start when the scheduler indicates they should, and it insures that jobs that attempt to use too much memory, disk, or running time are ended.

node A node is a physical machine. The three types of nodes you will see references to are login nodes, compute nodes, and (data) transfer nodes.

processors or procs This term is loosely defined and often used to mean either core or CPU (see above). It is usually clear from context whether it refers to the full CPU or to only a subset of it. This will always mean "something that can process instructions", and it is often not important exactly to what it refers.

scheduler A software program that attempts to efficiently run many jobs on a cluster.