UCT HPC cluster updates

We have noticed an increasing number of people installing software\libraries directly via the head node.  Please do not do this.  Many of these installs require a large amount of RAM or induce heavy load on the processors.  The head node is a small, virtual machine that is only intended to run the HPC scheduler.  Please use an interactive job to install software or libraries.  If you are uncertain how to do this please contact us and we will gladly assist.

To assist researchers in identifying if they are currently running an interactive job we have inverted the colours of the worker node server name in the prompt to black on white text.

If you no longer require your UCT HPC account, or if you have recently submitted a thesis or paper acknowledging usage of the cluster, please let us know.

During our migration project the new HPC head node was called hex. It will retain this name but now also uses the standard name, hpc.  You can use either.  If you normally login to hex you can just keep doing that. Logging into hpc is the same as logging into hex, just using the official name.

Last week we added another five 40 core ada:200 series nodes.  We are currently adding another twenty 48 core ada:100 series nodes and an additional GPU server with four L40S cards.

We will be increasing the default allocations for postgrad researchers to 200 simultaneous cores and 250 hours walltime. Honours and undergrad accounts will also have their limits increased to 120 cores / 170 hours and 80 cores / 100 hours respectively.

When requesting resources please do not oversubscribe your jobs.  Requesting 40 cores for a job when your code can only address 1 or 2 cores prohibits other researchers from using these valuable resources. The exception being jobs that require large amounts of RAM.

We have updated our Scratch Storage Policy. Please also note that the /scratch file system is not backed up at all, nor is it meant for long term storage.

We will be running an Advanced HPC training course on the 17th of July. This course will cover job submission methods, resource selection, environments and modules, library installation, containers and virtual sessions. A degree of familiarity with Linux is required. If you would like to attend this course please let us know.

We are looking at running several additional introductory training courses in the next few months. If you are interested in attending these introductory courses please let us know.

Software development course

For those researchers who are developing their own software RSSE Africa is hosting a two day Workshop on Sustainable Research Software Engineering at UWC.