Loughborough University
Leicestershire, UK
LE11 3TU
+44 (0)1509 222222
Loughborough University

IT Services : High Performance Computing

Job Submission Concepts


Introduction

The primary way work is run on hydra is through the SLURM resource management system, which is a batch job submission system. What this means is that you have to create a shell script which runs your work, with hints to SLURM on how to run it, and then submit this to the batch job system. This then runs it on one or more of the cluster nodes on your behalf when there is time to do so.

Even if the cluster is currently full of work you do not need for it to be empty to submit jobs. They will be held in a queue and run as soon as resources become available for your job(s).

It is reasonable to do light compilation work and other things on the command line, but any serious computation should be done via SLURM.

A prerequisite for running work via SLURM is having an account, which you should have at least one of, and you should send work to the system to be debited against the appropriate account. You should keep a track of your account balance as if this falls to zero then you will not be able to submit jobs.

Example scripts

For example scripts for packages see the Software A-Z which contains scripts for each package.

For generic MPI scripts see MPI .

For generic OpenMP scripts see OpenMP .

For serial jobs, jobs that require less than one whole node's worth or cores or high throughput, see Small Jobs .

Important Commands