To access software that is installed on the cluster, read the subsection on Environment Modules:


There is a distiction between software provided by the Linux distribution’s package management system, e.g. YUM for .rpm on Red Hat based systems and APT for .deb on Debian based systems, and the software stack on the cluster which is accessed via the module command:

Distro Package

  • Maintained by open source community
  • Compatible with all components of distro
  • Need to be root user to install

Software Stack

  • Maintained by sysadmins at Sussex
  • Built according to requirements of HPC users
  • Can be custom installed by the user
  • If the software libraries or packages you require aren’t currently installed on the cluster, then you can download them and install them locally into your home directory or research area on the cluster.

    You can also send in a ticket to the support team for them to add the software to the cluster’s modules to make reusing the software easier for colleagues - however the time this takes can vary based on demand, so it is always a good idea to attempt to install the software yourself locally first, and then you can help the support team with guidance to speed up adding the software to the cluster.

  • If you want to run a compile, then again look at interactive. Start an interactive session and run your compile there. Do NOT run compiles on the login nodes. If you want to do a parallel make, make -j 4 say, then request 4 slots with your interactive job as described in Running multi-core or high-memory workloads interactively.

  • You are strongly encouraged to use some sort of revision control system when developing your software. We encourage the use of git, and have a server you can use for this.

Running your

After you have more of an idea of what you are doing and have programs you wish to run unattended or at greater scale then it is time to read up on batch.

Also read queues, which outlines what queues are available on the cluster, and also on the policies that are attached to the queues. These policies will set limits on how long your jobs can run for, or how many slots you can utilise at one time.