My apologies - I'm on holiday next week.
But here's a item from EXSS / people I work with:-
####
The RCS service changed the queues for the college HPC quite considerably at the end of March 2022. It also seems that our jobs are now defaulting to run on an AMD EPYC platform, which doesn't offer as much 'bang for your buck' per CPU core, yet a lot of our scientific codes do not scale with the number of cores beyond a 'sweet spot'.
Since then, PhD students and PDRAs have reported very long wait times for their jobs to run, and quite unsatisfactory performance. The ramifications of these changes haven't been particularly well communicated (and generally, I'd state that the running of the queues is quite opaque from the point of view of a user), but it seems that as well as updating things, there has been a general shift in resources from the batch running queues (that power computational projects, and PhDs) to interactive queues such as the Jupyterhub (which are mainly used by undergraduates).
Could we get some feedback on whether these extended wait times / job backlog have been noticed, and request that more information / transparency is sent out to the RCS users?
Over the next few years, what is the intended investment in new hardware for the RCS?
Should we just accept that these queues are slow from now on & seek national HPC resources? Or will the situation change?
####
Best wishes,
Jarv