Minutes of the ITC-Research Computing Standing Committee

Meeting on January 27, 2003

Members: Alice, Bill, Brian, Dawn, Ed, Hamp, Kathy, Jim, Mark S., Martha, Michael, Robin, Sue Ellen, Steve, Terry, Tim S., Tom S.

Attending: Alice, Bill, Ed, Hamp, Kathy, Jim, Mark S., Martha, Robin, Steve, Terry, Tim S., Tim T.

Chair: Tim T.

Recorder: Alice

I. Discussion Topics:

1. Update on getting the new 32 Xeon-Myrinet cluster online: there was discussion about the possible need for another KVM plus some additional cables (subsequently Bill and Chip got everything on one KVM – so all that’s needed are some more cables). Before we put the new cluster online, Hamp would like to install the latest version of Rocks and a newer version of PBSPro on it. All the compute nodes came up and are working – which inspires confidence. We are aiming to announce the new Xeon cluster by the end of February (after the new version of Rocks is installed) and have it ready for testing from some friendly users. What to call the clusters? Aspen and Aspen2? Tim thinks we ought to call them the Aspen AMD Linux Cluster (AALC) and the Aspen Xeon Linux Cluster (AXLC) in our documentation.

2. Update on moving the AIX SMP box to PBSPro: there was considerable discussion about whether and when to make this change. On the one hand, since we will shortly be urging users to move off the SP to the SMP or the Aspen cluster, they would then be shifting to systems that both use PBSPro (instead of LoadLeveler). On the other hand it’s some work to make this change, and if we leave LoadLeveler on the SMP it would provide an easy migration path off the SP. Ultimately it seemed like a timing question – and it was decided to wait for AIX 5.2 – and make the change to PBSPro at the same time. Since AIX 5.2 will not run on all of our existing systems, it could be that we will go with AIX 5.1. In any case, it will not be a quick or trivial switch. Decided that we would make the switch and have PBSPro on it and ready for users by July 1^st, 2003.

3. Update on SEAS J40.J50 Migration: SEAS is contacting their users – and in a few weeks Mitch will be able to give us a list of who is left. One sticking point for moving these users is that they are accustomed to having a large amount of disk space on these systems – so they are having storage issues.

4. Timeline for SP Migration: The Research group is planning to send a letter to SP users during February to remind/warn them that the SP is not under a maintenance contract -- that they should plan to run code on some other system (SMP, Aspen cluster) soon – and that they should plan to migrate off the SP by July. The letter will also include some incentives for migrating off the SP (e.g. enhanced performance on the other, newer systems ).

(Note: We do not yet have a pull-the-plug date for the SP – but we’ll need to set a date at some point – hopefully before something catastrophic happens to the SP.)

5. Misc: Hamp is ready to replace the UnixLab usage page with the ganglia page – just do it. There was some discussion about which compilers to put on the Xeons – Intel? PGI? Both for now? Ultimately we would prefer a single compiler.

Next regular meeting: Monday, February 24, 2003, at 10:30 AM in 117 Astronomy (later cancelled)

Go to: ITC Research Computing Standing Committee Home Page