Members: David, Dawn, Dee, Ed, Hamp, Jim J., Mark S., Mark McKeown, Michael, Robin, Stan, Sue Ellen, Tim S., Tim T., Tom S.,
Convener: Tim T.
In Attendance: David, Ed, Hamp, Jim, Mark S., Mark McKeown, Mark Manley, Tim S., Tim T.
Recorder: Tim T.
Next Meeting: April 24, 2000 at 10:30 AM in Astronomy 117.

Click here to see the agenda from this meeting.

To Dos & Old Business:

Research Computing platforms and SP

Hamp: Allow interactive use on SGI(Teal) and run PBS to balance batch load. On Orange cluster, run PBS and not allow interactive login.
For now on Teal cluster, can use "round-robin" interactive login balancing. He's still looking for a SGI or Sun load balancing tool. He needs to merge UnixLab password database with research password database. He's got alittle more account clean up to do and then he can merge.

Specifically SP Ed: has been submitting parallel computing jobs using 2, 4, 6, and 8 nodes, about 30 jobs per category to monitor performance and turn around time.
Discussion on job times. Debugging requires quick turn around time.
Hamp: one issue is the "load" setting of loadleveler. Right now it's set at 1.
Tim S.: could we set it higher and allow only one serial job, so that parallel job could get that node even if a serial jobs running on it?
ToDo(Hamp):Will check with Bill P. about whether this can be setup?
Ed: What about creating a half-hour queue to quick turnaround?
Tim T: Could have these on both the SP and Orange/Teal clusters. We could create this queue and when advertise it, we could say something about having additional dollars available for more nodes if this change creates bottlenecks in other queues.

HSM
Discussion on Earl Mark's inquiry about large disk space area for students to use with courses. Hamp has previously talked with Earl Mark about this, we have such a space on the Home Directory, Jeff Hollier is overseeing it.
ToDo(Tim T.): Let Earl know about this; contact Jeff to let know.
Re-iterated that there's no need for a /longtmp. The HSM, /bigtmp and the courseware directory on the Home Directory replace /longtmp.
Hamp: What about shutdown of Maggie? Should we mothball or re-format the drives? When he was cleaning up he found some files that UniTree hadn't included in its reports, so he didn't know they were missing. He'll check on how widespread this is, checking through file system and matching to user list to what's stranded.
As for decommissioning, could leave old 2GB disk cache around for 6 months. He'll see what's possible.
Tim S:What is current usage of HSM?
Hamp: Olaf still working on creating real quotas that will reflect what's on disk and what's on tape. No one has more than David Seaman's 15 GB.
He just added second set of tapes to use for backup of the home directory and test server. So still not at or appoaching capacity.

Software
Hamp: Need to get SAS installed on Orange cluster.
Hamp: gcc 2.9.5 is available on Suns and SGIs; but can't get it to compile on AIX. Has been a real struggle trying to get a working AIX version.
The AIX 4.3. upgrade on the blue cluster is on hold, IBM broke the "yppassword" functions in 4.3 and we're waiting on a fix from IBM before proceeding. Nodes 7 and 8 are "yellow" cluster, that is already upgraded to 4.3

NPACI/Legion
Tim S: Anyone using Grimshaw's Centurions or Intel clusters?
Ed: No, he's been working with Eduardo Bringa. The original discussion was to get PBS working on our cluster and then use that knowledge to bring it up and maintain it on his.
He did did test PBS on Alpha Centurions
Mark McKeown will be working on Legion and these clusters for us, which will help us make progress.

Research Committee
Tim T: ACAC meeting monthly. Working on filling out implementation grid of UCIT's strategic plan.

New Business
None

Meeting adjourned at 11:40, next meeting April 24, 2000, 10:30 AM

Go to: ITC Research Computing Standing Committee Home Page

by Tim F.J. Tolson