Notes for the ITC-Research Computing Management Meeting
November 30, 2007 at 10:30 AM
2015 Ivy Road, 1st Floor Conference Room

Members: Alice, Hamp, Jim, Joe S., Mark, Mike, Robin, Terry, Tim S., Tom, Tim T.

Attending: Alice, Hamp,Jim, Joe S., Mark, Mike, Robin, Tim S., Tom, Tim T.

Chair: Tim T.
Recorder: Alice Howard

I. Connection Question: How participative do you plan to be at this meeting
(On a scale where 0 = Low to 7 = High) - and what's the story you're telling yourself about that.

II. Corrections to minutes from last meeting on October 29, 2007?
(Click to review the agenda from this meeting)

III. Old Business:

• Update from Mike on the CSAC lunches and Computational Science plans:
• Mike gave an update on the current state of the documents prepared by the Commission on the Future of the University: IT remains ‘at the table’ through various revisions and distillations. The Provost is emphasizing that the VPs are going to move on these recommendations – with James and Karin given responsibility for the Academic Infrastructure piece (that includes much of the IT recommendations).
• At a recent President’s cabinet meeting, James stressed that the ambition to make a big play in computational science will require strategic investment – note: the response to August’s HPC Bootcamp was a great indicator of the level of interest at UVa.
• So – given these developments, we might want to do a bigger cluster – something significant to help get some buzz going.
• We accept the ITRTF report logic, so the big issues are creating a culture for computational science, so like will put more dollars into people, graduate fellowships, etc. than hardware resources
• In addition, James and Mike are talking about “declaring” a Center for HPC and giving it some resources to increase visibility – and then let it grow. Has potential to lift every discipline’s work. [Jim had to leave for another meeting]
• Mike also talked about the next phase in the NSF Petascale Initiative – and that we are talking with some other mid-Atlantic institutions about a possible partnership that could put together a regional proposal.
• There was some discussion about the new data center building: we are close to having an engineering firm – and may hire someone to help oversee the whole project and do a 3^rd-party review.

• Time for our next "condo cluster" planning - what, when, policies about procurement.
  Current web site: http://itc.virginia.edu/research/itc-clusters/
• Agreed last meeting that would offer dedicated nodes (with 80% rule) and consider pre-emptive queue access to these nodes for other users.
• So – where to go on our next condo cluster purchase? Cannot replace Cedar unless we have lots of $$. Mike suggested that we start thinking about how we might spend $0.5 million on a cluster.
• And where to plug it in? We are so tight on machine room capacity (until the new data center building is finished) that we need to start investigating where we can acquire temporary machine room space. [@11:40 Mike & Alice had to leave for another meeting]

• Public forum on queuing policies - in late January or early February? Whom to invite?
• Ed & Katherine doing review of other institutions queuing and cluster purchase programs in preparation.
• It would be interesting and helpful to find out how these other places educate their users about the queues and how successful they think it all is.
• Katherine has gathered some information while at BioMed HPC Conference in Boston.
• Katherine's 2006 proposal PBS Service Units Proposal
• Linux Clusters: Cedar, Dogwood Elder- any issues, concerns or upcoming updates/changes?
• Cluster upgrades on hold waiting for next version of ROCKS release. Hamp: ROCKS projects it’s next release for late spring, so will aim for summer upgrade.
• Hamp discussed current power situation with Cedar, Dogwood & Elder. Had a problem with the Carruthers machine room UPS before Thanksgiving and had to turn off 2 frames of Dogwood to keep from overloading UPS/generator. One rack of Dogwood is 3 to 4% of the capacity of the UPS/generator. So all three are currently off the main UPS/generator. If power goes out, have only 5-10 minutes of ride through to shutdown the cluster. Will need to schedule and announce some downtime for the clusters to move them to their own individual UPSes.
• HPC Bootcamp January 7-11, start with UVa only, Intro to Unix workshop before it, like Jan 3-4. Planning and logistics proceeding.

• Update on Computational Science Speaker Series None

• CSAC request to Tom for report on current cluster use vis-a-vis PBS queue use/analysis.
• Data from Tom on cluster usage of UVa HPC bootcamp attendees.
• Looking at what applications 8 possible TeraGrid candidates using
• Given our early discussions, should we consider moving our next cluster to VT? Or even a joint cluster with us and VT, hosted at VT. Then the second year, we would host the next joint cluster purchase.
• Think about making fewer but bigger cluster purchases, with an 18 or 24 month gap between cluster purchases. We don’t know if VT is making periodic cluster purchases or not.
• We should also think about purchasing blades next time - better power, more cooling, less space, more cost per unit.
• As part of this and queuing review – we need to consider dropping PBS-Pro.
• Tom: ACHS already has – using Maui/Torq because we get no response when need licenses for PBS-Pro. So as part of queue policies, need to look at what other schedulers can do – pros/cons. Plan to ask some carefully targeted questions of our users.

• New chair of this committee for 2008 Deferred until next meeting – as several had to leave early.
• Discussion of purpose of group:is it a Big-picture & Policy group? A Nitty-Gritty operational group?

• Establish our Operating Agreements: Some possiblities: (Deferred until next meeting – as several had to leave early.)
• notification of attending or not?
• Start and end on time? Build in travel time to meeting schedule?
• Meeting day and time?
• End meetings with clear requests and promises for action.
• Don't leave meeting with unspoken issues or concerns.
• Others?

IV. New Business:

• Any news, information from E-School? SURA? CSAC brownbags? UCIT? Other organizations?
• Meeting next week with A-School about possible HPC rendering farm.
• Any other Items?
• Tom: giving away a Pro-Curve 48 port gigabit switch if someone can make use it. (Hamp claimed).
• Dave Luebke talk in CS this afternoon on using GPUs for computations.
• Completion and review promises and requests from this meeting.

V. Adjourned 12:10 PM

• Next Scheduled Meeting of ITC ResComp Standing Committee: Monday, December 17th, 10:30 AM in ITC-2015 Ivy Road, Room #102 (First Floor Conference Room)!
  
  Please send suggestions, additions, corrections to: Tim Tolson or Alice Howard
  

Go to: ITC Research Computing Committee Home Page