- A cluster computer is a supercomputer assembled from cheap, replaceable nodes linked by a fast network.
- Work to be performed on a cluster--a "job"--is organized as a script, typically in bash or in Python.
- Jobs are scheduled on the head node of a cluster, while work is performed on the compute nodes.
- Slurm commands are used to submit and monitor jobs on a cluster.
squeueshows the status of jobssbatchsubmits a job to the queuesinteractivecreates an interactive job on a compute node
The table below defines some terms described in this lesson.
| Term | Definition |
|---|---|
| compute node | A single computer in a cluster where computations are performed |
| head node | The entry point in a cluster computing system, where the job scheduler resides |
| job | A unit of work encapsulated in a script on a cluster computer |
| job queue | An ordered list of job assembled by the scheduling software on a cluster computer |
| node | An individual computer in a cluster |
| scheduler | Software that controls when and where a job is run on a cluster |