1. Request HPC access or support

1.1. Requesting access to HPC

IT Services maintains multiple High Performance Computing (HPC) clusters at the University of Surrey. Which cluster you get access to depends on which Research Group / Department you belong to.

For access, please first consult with your Supervisor or Line Manager for information regarding which specific cluster/s you can have access to. If you are unable to get this information, submit a request explaining your needs and provide any relevant information, and we will advise you how best to proceed.

You can submit a HPC cluster access request via the support portal or e-mail. Within this request, we need written approval for access from your Supervisor/Line Manager.

Note

Access to the Eureka HPC Cluster is open to all University research staff and students, however Supervisor/Line-management approval maybe required in most cases.

1.1.1. Via the IT support portal

https://support.surrey.ac.uk

  • Full Name

  • Surrey Username

  • Name of Cluster you are requesting access to

  • Your Department and specific Research group (If unsure of exact group, see: https://www.surrey.ac.uk/academic-departments)

  • Name of Supervisor/Line manage

    • Written approval from your Supervisor / Line-manager to use the HPC cluster.

1.1.2. Via e-mail to the IT service desk

E-mail: itservicedesk@surrey.ac.uk

Please CC in your Supervisor / Line Manager and provide the following information*:

  • Full Name

  • Surrey Username

  • Name of Cluster you are requesting access to

  • Your Department

  • Your Research group

  • Supervisor/Line manager

  • Faculty Name

Mailing list

When your account is created on the cluster, you will also be added to the relevant cluster’s mailing list, so we can keep you updated on news related to the cluster

1.2. Get HPC support

1.2.1. Open a support ticket

Research Computing operate from 9am to 5pm Monday through Friday, except for Holidays.

You can submit a support request any time via:

Hint

Put “HPC” in the e-mail subject to reduce the time it takes for the ticket to get assigned to our team

You can help us and yourself by following these best practices when submitting tickets:

  • 📖 Do your research before submitting a support request. What does the documentation or user guide say? Have you searched the internet for key phrases from your jobs log files / error messages? What have you changed since the last time your job succeeded? These are all the things that the support consultant responding to your ticket will do or ask you.

  • 🎯 Be precise in your description of your problem. What did you do? What happened? Verbatim error messages and other meaningful output are very helpful. Where relevant provide details of storage directories your job might be using, jobs scripts, Job ids, Modules you were using, hostnames of servers your jobs are running on etc.

  • 🤔 Manage your expectations. Consultants can address system issues and answer questions about the clusters, job schedulers etc. But they can’t teach parallel programming in a support request and might not know anything about the package you are using. They may offer advice that might help you to build, debug or optimise your code but you shouldn’t expect these things to be done for you.

  • ⏱ Be patient it may take a day for a consultant to respond to your support request, especially if your issue is complex. it may take a few messages before you and the consultant are on the same page. If your jobs are paused/held by the admins its not punitive. When the system is being impacted or in danger of crashing they may not have time to notify you before acting.

  • 📰 Subscribe to Research Computing Community on Microsoft Teams

1.2.2. Request new software

If you have a request for a new software package to be installed on one of the HPC clusters please Open a support ticket as above and provide the following information:

  • Name of the software

  • Summary/background information about software - what you want to use it for.

  • Which cluster you would like the software installed on.

  • A link to the website to download the software.

  • Version number of the software required.

  • What type of version/additional libraries or specific compilation options you require.

Note

Please be prepared to test the software for us once installed to ensure its been compiled properly.

1.2.3. Research Computing Community on Microsoft Teams

Another great way to get support is from your peers and colleagues here at Surrey. There is an open community on Microsoft teams where you will find plenty of knowledgeable and experienced HPC users already discussing all things HPC. This is also a great way to speak directly to us in the IT Services Research Computing team and get updates on scheduled maintenance and system outages.

Get involved…

Join our Research Computing Community on Microsoft Teams

1.2.4. Research Computing virtual appointments

Book a “HPC Appointment” using our Research Computing Bookings if you prefer a one-to-one chat with us regarding anything HPC related.

1.3. Purchasing HPC servers

1.3.1. Eureka2 HPC cluster

Any member of staff is eligible to buy into the Eureka2 HPC cluster. Contributing resources to a cluster will increase your SLURM fairshare on the cluster.

See: Job priority and “Fairshare” on Eureka2 for more information on how fairshare works.

Below outlines the specification for a standard Eureka2 HPC compute node:

  • Node: Dell PowerEdge R6525

  • CPU: AMD EPYC 7452 32-Core Processor, 3344 MHz

  • CPUs per node: 2

  • CPU cores per node: 64

  • GPUs per node: 0

  • RAM per node: 512 GB

  • High-performance Interconnect (Speed): Infiniband (100 Gb/s)

If you have any further questions relating to buying a similar node but perhaps with slightly higher specification (eg. a higher memory node) or you require a quotation with details and costs, please raise a support ticket by emailing itservicedesk@surrey.ac.uk. In the email subject, include “Eureka2 HPC node purchase”.