cd ../ High Performance Computing - Software

High Performance Computing

Documentation

Access and use of the Nottingham HPC service

The HPC service is available to any member of academic staff or research student, from any School or Faculty, who has the need for compute resource substantially greater than a standard PC. Those who wish to use the HPC should:

  • Obtain a University Unix account, if they don't have one (usually from School's IT rep).
  • Mail hpc@nottingham.ac.uk, with username, and brief details of what they want to use the system for, and approx resource requirements (No. of CPUs needed, length of jobs, storage requirements)
  • Students should specify their supervisor's name, and School.
  • All users should specify the 'IP' address of the workstation from which they wish to access the HPC. If you are not sure how to find out your IP address, please contact your local IT support.
Note that some features of the HPC mean it is not suitable for all applications:
  • The HPC system runs a variety of Linux as its operating system, so software must be compatible with this.
  • Software which only runs on Microsoft Windows is not suitable for use on the HPC.
  • All compute jobs are submitted to a 'queue' and automatically run by a 'scheduler'. This means the software used must be run in a non-interactive mode, by a script file.
Once an account has been obtained, the system can be directly accessed using 'secure shell' (ssh). Files can be transferred to the system using 'secure copy' (scp).

The 'client' software required for this is available on most University desktops, or downloadable (e.g. see links for Putty and WinSCP). Once you have located/installed the ssh/scp client, run it, and specify one of the HPC 'login nodes' as the Host Name. These are lin01.hpc, lin02.hpc up to lin04.hpc. Enter your Unix username, and when prompted your password.

If you are on a network connection remote to the University, you will have to provide a 'fully qualified' host name, e.g. lin02.hpc.nottingham.ac.uk

Interactive sessions on the login nodes can be used, for example, to edit files, compile programs, and run graphical applications (using X-windows), but should not be used to do large calculations. It is the 'compute nodes', accessible only via the scheduling system, that should be used for major computing tasks.

Please Note: currently only the newer hardware in 'tower E' is available for use. This is accessed via the module 'jupitere' - see The Facility for more information about the system specification, and Environments for information on how to set your working environment to use the various resources on the HPC.

The facility is a substantial compute resource, but is shared by over 300 users, with typically several dozen running jobs simultaneously. If you have requirements for running on hundreds of cores for many days or weeks, you may need to consider applying for use of the National HPC facility known as HECToR. Note that access to HECToR is via a peer review mechanism, or other technical assessment of projects, and it also only runs parallel jobs, so that jobs consisting of task arrays of single core jobs are not considered suitable for HECToR - see the HECToR website for more information.

Training and further assistance

If you are unfamiliar with Unix or Linux, then you are encouraged to attend one of the regular 'Introduction to Unix' sessions, run by Information Services. See the Central Short Courses web page for more details. Some complementary Unix training resources are also available.

A selection of introductory and more advanced training courses for HPC users is now available.

There is more information in these pages (see menus at left), and if you have further questions, the HPC support team is available for advice.