MeetingNotes20081117
(permalink)
last edited by Brian Davison on Monday, 11/17/2008 5:09 PM
Attendees - Peter Bryan, Brian Davison (minutes), Terry Delph, Bruce Dodson, Ben Felzer, Gale Fritsche, Liangjie Hong, Bob Kendi, Kamil Klier, Steve Lidie, Imre Polik, Slava Rotkin
Excused - Ian Laurenzi (conference), Brandon Leeds (conference), David Myers (dept. candidate)
Agenda
- Standing sub-committees -- Brian
In the future, subcommittee chairs will determine agendas for their areas.
- Issues considered by sub-committees:
- Strategic Planning (Brian Davison, chair)
1- Linux teaching lab -- KamilKamil presented the need for a Linux lab for Chem 341 (serving ~50 students annually) and Chem 443 (typically 7-9 students biannually). Requested 50 seats, with a Spartan license tied to each machine. Also to use Maple, Matlab and Gaussian (open version). Spartan jobs may run for a week.Slava reminded us to contact Arnold Kritz as another potential user. The subcommittee will prepare a request to send to Bruce Taggart.2- Application software licensing -- Kamil, BrianBrian reminded the group of its relationship to the LTS Software Committee which is meeting early in December to decide software licenses and purchases. Kamil described needs for the Linux lab and Peter described current licensing needs at Atlss for Matlab. Gale Fritsche is the HPC liason and will bring our concerns for Spartan (at least 25 licenses, according to Kamil), Matlab (might be ready for a site license, according to Peter), Origin (used in Chemistry), and Gaussian. Gale said that Kamil and/or Peter could address the software committee. Peter will provide documentation of need to Gale.Slava brought up concerns regarding support needs for HPC applications.
- Policies and Procedures (Peter Bryan, chair)
1- Machine room colocation policy -- GaleGale described the arrangement that Ben has with LTS, and discussion brought up issues to be resolved, such as funding for repair when equipment is not under service, when the arrangement is complete, a need for formal policy to advertise/explain possible arrangements for future resource sharing. Ben's system is still not part of the Condor-scheduled resources. The subcommittee will need to prepare such a policy, and Gale will provide a draft from existing arrangements.2- HPC project-specific GA support requests -- BenBen described the need for GA support for installing and optimizing large distributed models. We discussed the possibility of a cost-sharing arrangement to help support additional GAs. The committee decided that at present the existing HPC GA could be supplied to assist in such situations, and encouraged Ben to contact him. If his time becomes in high demand we may revisit this issue, and/or use the demand to justify additional resource requests.3- Dedicated use of HPC resources -- GaleGale briefly described the situation with S. Roy in which LTS management decided to temporarily dedicate 12 nodes (64 cores) from the Inferno cluster to Dr. Roy's research needs as a result of difficulties with running his jobs via Condor on Altair. The allocation was for two weeks, in which one week remains. Slava noted that such arrangements can be beneficial to all (better use of resources, publicity, correcting problems), if they do not adversely impact others. The subcommittee will prepare a policy under which future requests can be made.
- Outreach and Education (Ben Felzer, chair)
- Future Systems (Slava Rotkin, chair)
1- Egenera Goldman-Sachs donations update -- GaleA purchase order for $10K was signed last week to Egenera for the packaging, movement and installation of two Egeneras. Expected delivery date is first half of December.2- New SMP selection -- BrianBrian described an existing budget of approximately $30K that we had planned to use for a second SMP machine with capabilities similar (perhaps better) than Altair. Bruce suggested that we consider two 16 core machines instead. The subcommittee will begin to examine possible choices for eventual recommendation to the full committee.
- Proposal Development (Bob Kendi, chair)
- Operations
1- Altair performance update -- Steve L.This question came up while discussing dedicated use of HPC resources. The original problem has not been corrected.2- Condor Job priority settings -- Kamil
No time to visit this topic.
3- Cluster node problem repair workflow -- Steve L.
No time to visit this topic.
- Next Meeting -- Brian
There will be no HPC SC meeting in December. Subcommittees can use this time to make progress in advance of our January meeting. Brandon will be in touch via email to set up our meeting schedule for the spring semester.