Hi folks,
I thought I'd give a brief introduction, and see if this list is a good fit for
my questions that I have about my HPC-"ish" infrastructure...
I am a ~30yr sysadmin ("jack-of-all-trades" type), completely self-taught (B.A.
is in English, that's why I'm a sysadmin :-P) and have ended up working at an
industrial research lab for a large multi-national IT company
(http://www.nec-labs.com). In our lab we have many research groups (as detailed
on the aforementioned website) and a few of them are now using "HPC"
technologies like Slurm, and I've become the lead admin for these groups.
Having no prior background in this realm, I'm learning as fast as I can go :)
Our "clusters" are collections of 5-30 servers, all collections bought over
years and therefore heterogeneous hardware, all with locally-installed OS (i.e.
not trad head-node with PXE-booted diskless minions) which is as carefully
controlled as I can make it via standard OS install via Cobbler templates, and
then further configured via config management (we use Ansible.) Networking is
basic 10GbE between nodes (we do have Infiniband availability on one cluster,
but it's fell into disuse now since the project that has required it has
ended.) Storage is one or more traditional NFS servers (some use ZFS, some
not.) We have within the past few years adopted Slurm WLM for a job-scheduling
system on top of these collections, and now are up to three different Slurm
clusters, with I believe a fourth on the way.
My first question for this list is basically "do I belong here?" I feel there's
a lot of HPC concepts it would be good for me to learn, so as I can improve the
various research group's computing environments, but not sure if this list is
for much larger "true HPC" environments, or would be a good fit for a "HPC
n00b" like me...
Thanks for reading, and let me know your opinions :)
Best,
Will
_______________________________________________
Beowulf mailing list, [email protected] sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
