Hi All,
I promised both the Physics Computing Committee and the FRCC that I
would send around and email for people to circulate to aquire
information about HX2 and other future needs. I started and then thought
that it might be more fruitful to circulate a draft first and then
refine it slightly. So below is a (fairly rough) draft for comment.
Comments are very welcome.
Best,
david
Sub: Input into future research computing needs
Dear All,
As you may know the College's HPC platform has been upgraded with HX1
coming online earlier this year. HX1 is a cluster of over 18k cores
(Intel Xeon Platinum 8358) and 60 GPUs (Nvidia A100s).
This year the focus is on the update of the data storage with a move
away from the current RDS to a more structured Research Data Facility
(RDF). The RDF will be coming online later this year.
We are now looking for input as to what will be needed both next year
and further into the future on a 3 to 5 year time scale. Next year the
plan is to decommission the College's main High Throughput Cluster (HTC)
CX3 and replace with a new system. The budget available for this
purchase will be approximately £5M. Because this will be our first fully
watercolled system, we are negotiating space in a new data centre with
this capability and so this resource will not come online until the
summer of 2025. We are seeking your input as to what we should purchase.
As this is replacing an HTC system it is assumed that the majority of
the money will be spent on another HTC system, however, if your research
has need for other shapes of resources then please let us know. Also
even if what you want is an HTC resources what parameters are important
to you (e.g. memory/core, access to GPUs, throughput from disk, etc
etc). *** Different for different departments - this is the physics
example *** Please give us feedback via your representatives on the
Departmental Computing Committee.
We are also trying to establish what the more medium term computing
needs will be. So please can you also discuss these with your
representatives on the Departmental Computing Committee. These may be
expected needs from from specific instruments/experiments/projects of
general growth in computing needs. We will then try to amalgamate these
into a coherent input for future purchses.
Best,
david