MeetingNotes20080915 (permalink)

Agenda

       Attendees -Bruce Taggart, Brian Davison, Kamil Klier, Liangjie Hong,  Meghanad Wagh, Slava Rotkin, Steve Roseman, Gale Fritsche, Bruce Dodson, Peter Bryan, Bob Kendi, David Myers, Tim Foley, Ben Felzer, Imre Polik, Brandon Leeds (minutes)

•  Welcome and charge -- Bruce Taggart

Bruce gave background on how the Committee started as a grassroots effort amongst faculty such as Terry Delph. HPC resources on campus are now used for research and instruction. Committee will have to address issues in the coming months about how support of scientific computing on campus is to be accomplished, such as GIS in EES. Because currently HPC SC oversees and provides guidance for HPC hardware such as compute server and clusters on campus. Another topic the Committee will most likely give input in over the coming months is about Data Center capacity and expansion study.

• New leadership and new members introductions

 Going around the table, everyone introduced themselves.

• Standing Committees -- Brian

Brian presented his ideas for the overall structuring of the SC as a whole body with separate subcommittees taking on specific responsibilities, namely, Operations, Policies and Procedures, Outreach and Education,  Future Systems, Strategic Planning, and Proposal Development.   Sign-up sheets were passed around for all Committee members to sign-up for two (2) subcommittees or one (1) subcommittee and alternating minutes duties. Subcommittees would then meet and give monthly reports back to the whole committee and alleviate burdens of in depth discussions on topics that really require more time and focus than the 50 minutes the HPCSC meetings should be limited to.

• Operations

1- Altair performance update -- Steve

Nothing new to report. Steve needs to wait until Steve Lidie is back from disability leave to work on this problem. It is thought that the I/O subsystem is improperly configured (uses software raid) and may need to be replaced with an alternative approach entirely. Possible alternatives are limited by the fact that there are few if any available expansion slots in the F1200 chassis - need to check on this. However there are many 1GBE ports that might be used for iSCSI based disk subsystem.

• Policies and Procedures

1- Machine room equipment policy -- Gale

A new set of usage/co-location policies needs to be drawn up in light of our success with co-locating the EES global climate modeling cluster in the data center. Left for a subcommittee.

• Outreach and Education

1- Migration from old Coral Lab TRAC software based wiki to new Blackboard based wiki -- Brandon

In operation and being used.

2- Updated HPC web pages -- Brandon

Asked committee members to take a look at and report back on their findings and on ways to improve our updated information. Re-write of many pages were required to reflect the major changes that had taken place over the summer. Up to date information on software application ,  plus new information has been added in several areas. Work still needs to be done on Condor on campus documentation. Need input from Operations on how Condor sub-groups is configured. Need to wait until Steve Lidie returns.

3- IBM Cell BE training status -- Brandon

Good registration numbers - up to 20 people so far.

4- Condor documentation for blaze/inferno -- Steve/Brandon/Liangjie

More in depth information is needed on how to use Condor. This will be an objective in the coming month to augment documentation provided on line to HPC users. Input from Operations staff is required to understand the current Condor groups and how to submit to different groups.

5- Committee mailing list archives -- Gale

 Gale will check on how to access archives that have been enabled for the list so that we can fix the broken link in our HPC web pages.

• Future Systems

1- Egenera Goldman-Sachs donations update -- Gale

Gale needs to gets quotes for boxing and shipping ( Egenera needs information on measurements/weight/insurance ) before we can get delivery. Once received, 2 racks of more of the same as our most current leaf nodes which can handle the latest Linux kernel- which one of the current Egenera racks we have cannot, we can then upgrade that rack, add a third rack- expanding our current leaf login farm node count - and get rid of the older rack with a spine that cannot be upgraded.

• Strategic Planning

1- Linux teaching lab -- Kamil

No time for discussion.

• Proposal Development

No time for discussion.