CSF3 Service Improvements 2025

Upgrade work – Jan and Feb 2025

Further to the installation of the AMD nodes in September 2024, there will be additional upgrades made to the CSF3 during January and February 2025. A summary is provided below. More information will be sent out to the service announcements email list as and when appropriate.

What will be changing?

  • New login nodes
    • New physical login nodes to replace aging hardware.
  • A new scratch filesystem
    • Approximately 1.9PB – an increase of 500TB.
    • Faster than the current one.
  • A new batch system named “Slurm”.
    • A popular queuing system on many academic HPC systems.
    • It has better support for modern hardware.
    • It will enable us to manage the system resources more effectively.
  • Deployment of a further 15 AMD compute nodes, providing an additional 2,520 cores (purchased recently with funds from several groups/academics).
  • Deployment of 6 additional 2TB “high-memory” compute nodes, bringing the total to 16.

What this means for you?

  • Currently there are no plans to interrupt the existing service to carry out the above work. However, there may be short periods where compute capacity is reduced.
  • You will need to modify workflows and jobscripts to run jobs in the new Slurm batch system, but there will be a limited period of overlap where both the old and new batch systems are available to you.
  • You will be required to copy the files you still need from old to new scratch.
    • Old scratch will be switched off in due course.
  • We may put a pause on the central install of new applications.

Must I change how I work? Do I need act now?

  • We appreciate that some of these changes will take time to get used to, but we will document them and provide as much guidance as possible on how to adapt.
  • You do not need to take any steps right now (mid-Jan). We will send further emails with detailed information as the work progresses and update this page appropriately.
  • Please be assured that we are doing our best to minimise disruption, and all upgrades are necessary to ensure the system continues to serve the University’s research community for many years to come.

Enquiries

  • Please note that at the same time as delivering the above, we continue to field your day-to-day enquiries for all our services, and have various other essential service tasks we must perform.
  • We appreciate your ongoing patience during this particularly busy period for the team. If you need any help using the CSF we would appreciate it if you could log requests via the new IT portal, “Connect”, where we have an HPC/CSF Help Form
    • Your enquiry will come directly to us and be handled by the team member on duty (Mon-Fri, business hours).

Last modified on January 28, 2025 at 10:11 am by George Leaver