Skip to content
Ahmad Ammar

Getting started on an HPC cluster

This is the guide I wanted when I first logged into a supercomputer as a PhD researcher. I trained in geo-informatics and remote sensing, which combine geography, geospatial science, programming and mathematics, yet a shared supercomputer was still new to me. It assumes you have never used a command line. By the end, you will have logged in securely, moved files, installed software, built confidence with seven small practice jobs on real data, and run a parameter sweep of many jobs. A final chapter carries these skills to other clusters, the cloud and everyday Linux administration.

The examples use Kelvin2, the cluster run by the Northern Ireland High Performance Computing service (NI-HPC). Most of what you learn carries over to any Linux cluster that uses the Slurm scheduler. Kelvin2 only marks details specific to that system; check your own cluster's documentation for the equivalents.

How to use it

Prefer one long page? Read the whole tutorial on one page.

Read in order if you are new, or jump in if you are not. Blocks of code are labelled with where they run, because typing a command on the wrong computer is the most common beginner's mistake:

  • laptop means a terminal on your own computer.

  • cluster means after you have logged in.

Lines beginning with # are comments, which the computer ignores. In this guide, they explain the line beneath them. Anything in <angle brackets> is a placeholder to replace, without keeping the brackets.

Practice files

The starter kit holds small programs and open data for the exercises. Start there if you learn best by doing.

Other good guides

This guide is my own, written from setting up on Kelvin2, and it is not an official NI-HPC document. Always treat the official documentation as the authority on your cluster's rules.

Corrections are welcome: contact details are at the foot of the page.