Minutes for the
ITC-Research Computing Standing Committee Meeting
(in
Members: Alice, Bill, Dawn, Ed, Hamp, Kathy G., Katherine H., Jim, Mark S., Martha, Robin,
Sue Ellen, Steve, Terry, Tim S., Joe S., Tom S.
Attending:
Alice, Bill, Ed, Hamp, Kathy G., Katherine H., Jim,
Mark S., Martha, Terry, Tim S., Joe S.
Chair: Tim T. Recorder: Alice Howard
II. Ongoing Discussion Topics:
1. Cluster
projects:
· “Condo” Cluster: Will operate as 32-bit cluster for at least
first year.
o
The hardware is all here and almost all
cabled – will do the 1st install on the front-end and then check for
DOAs and cabling – so making progress.
Hamp thinks the cluster will be available for users by June
1 at the latest. Bill thinks that
the beta ROCKS he is putting on will be the final ROCKS 4 by the time the
cluster is ready for users.
o
Work on setting up the four
node test cluster (for 64-bit
operation) will come after the big cluster.
Right now work on the big cluster has the high priority.
o
Have four researchers participating: Leo Zhigilei (lz2n) – 8 nodes w/ ETF $;
Jeff Shabanowitz/Dina Bai (Chemistry)-
8 nodes; Phil Parrish/Rich Gregory – 4 nodes; Matt Neurock
– 4 nodes.
o
Teak Cluster:
o
Oak Cluster:
o
Aspen & Birch cluster: Timeline for OS upgrade and upgrade to 2.6 kernel.
§
Over the summer. The new cluster will be the template for this
upgrade.
2. /longtmp additional disk space from Library available yet?
· Disks are not configured but are available – need to move and
cable – hope to get to it pretty quickly.
· Script to automate checking usage and reminding users.
3. Linux
license server, cluster of three lm1., lm2.,
lm3.license.
· Propose moving all research software licenses (or as many as
possible) off of aix.license onto linux.license
server -- yes,
seems reasonable.
4. blue.unix upgrade of nodes to new 64-bit nodes and new
version of OS. ITC-Transitions CDP is tracking as
well. Is holmes converted over yet?
· Machines are here – but work on upgrading holmes from 5.1 to 5.3 will wait until after the big
new cluster is installed and configured.
5. Discuss/thoughts
on Andrew Grimshaw’s “grid clustering” presentation:
· Would be good/convenient if the data grid software handled
enrolling, cleaning up, managing resources, etc. for the users’ desktops – but
there are problems with Windows machines becoming a server in the grid
environment.
III. New Business:
6. Upgrades
to PGI and Intel compilers on linux clusters. PGI to Version 6.0 and Intel to Version 8.1
· Scalapak does work on
· Will do the compiler upgrades over the summer – mostly just
announce and do – will aim for May 25.
7.
Need /jobtmp symbolic link on all clusters
(aspen, birch, cedar, oak, teak, etc.)
·
OK
8.
Jim Oberhauser’s complaints about aspen.itc queue management and practice:
·
Users allowed to get
18 nodes (36 CPUs) for parallel jobs and 24 nodes serial.
·
Fair Share scheduling makes it possible for
longer jobs to run before shorter jobs based on amount user’s job have using
cluster recently.
·
Jim wants his 10 minute job to run soon –
we have made some adjustments.
·
Discussed the pros and cons of putting an
express queue on the new cluster – could do it, but we have traditionally had
problems with running more than one queue on a system.
9. VORTEX: Impact on most researchers, how to
advertise/describe to researchers?
·
Jim did a presentation on VORTEX for the LSPs. He will send
Tim a paragraph that could be used as a short article in the May ResComp Newsletter.
Basically we don’t really know what impact this will have on most
researchers.
Next Scheduled Meeting of this group:
Monday, June 27th,
Next Scheduled Meeting of ITC ResComp Management team: Monday, May 23rd,
Go to: ITC Research Computing Standing Committee Home Page