Notes for the ITC-Research Computing Standing Committee Meeting
March 9, 2009 at 10:45 AM
2015 Ivy Road, 1st Floor Conference Room

Members: Alice, Andrew, Hamp, Jim, Joe S., Mark, Mike, Robin, Terry, Tim S., Tom and members of Systems, Research Computing and ACHS as desired

Attending: Alice, Andrew, Hamp, Jim, Joe S., Mark, Mike, Robin, Tom, Ed Hall, Katherine Holcomb, Kathy Gerber

Chair: Tom
Recorder: Alice

I. Corrections to minutes from last meeting on February 9, 2009? (Click to review the agenda from this meeting)

II. Old Business:

• Update on UVACSE APB and related.  Queuing policy / documentation
• There was some discussion about whether (or not) the new queuing policy had been implemented – or documented.
• Also discussion about the possibility of having a single queue with resource routing – who might work on this?
• Andrew wants the new policy in place (announced, documented, implemented) by the time the new equipment comes. Could this get done by the next Town Hall on 03/20/09? Jim and Andrew will work this out with Hamp and Bill P.
• The new policy could be announced on the new UVACSE web site – and we could redirect the high-performance-computing information on ITCWeb so that information would become re-branded as UVACSE resources.
• So what would it take to get the new queuing policy implemented? What are the potential impacts on users – and how much help might they need? Katherine will head a sub-group (includes Bill P.) and they will come up with a configuration proposal, a list of challenges that would arise – and then do goals and a migration plan – in the next week or so.
• Storage needs assessment and strategy (e.g., Lustre?)
• The CS department is using Lustre – currently have 40 TB storage and adding more – it has been effective for CS so far. There is some concern over impact on the network so they are keeping network stats for awhile. If this experiment seems successful it could be tried more widely at UVA.
• Andrew recommends: (1) waiting and seeing how Lustre works for the CS department (and Michael Shirts); (2) then trying an internal experiment. (Note: Lustre drives need to be network-addressable and running under Linux – they are not backed-up.)
• Meanwhile we could add an extra drive to each new node – Andrew, Jim, and Hamp to look into this.
• Cluster node purchase (Update from Hamp)
• Hamp reported that 26 nodes have been received (8 for Joe Mychaleckyj, 16 for Michael Shirts); Infiniband came last week. Will put current version of Rocks on it and try it – and go to the next higher version of Rocks if needed.
• PBS Renewal (pricing?)
• Katherine will check on pricing.
• There was some discussion about the pros/cons of staying with PBS (costly on an annual basis) vs. going to Torque (may have problems with throughput and robustness).
• Leave this on the agenda and we will reconsider it; some cost-sharing for PBS might be an option.
• VPN Solution for 64-bit Windows (esp. remote Matlab use)
• We are slowly moving towards a solution – meanwhile an ssh tunnel does work – Mark is updating documentation.
• Compiler usage (Hamp)
• Save for next meeting’s agenda. Looking for PGI compiler usage so we can evaluate whether or not we need to keep the PGI compilers in addition to the Intel compilers.

III. New Business:

• Cluster usage reporting to users (Andrew)
• UVACSE’s new hire will merge and sort what Tom sent before the next Town Hall meeting.
• There was some discussion about using the PBS logs to produce some regular usage reports – Hamp already has a mysql database, created by processing the PBS logs, that can be used for this purpose. Hamp will email Andrew the schema, plus a username/password for access to the database.
• Ownership of /common directories (Katherine/Andrew)
• Katherine, Ed, and UVACSE staff will have access through a UVACSE account – and we will remove the RCSC account (but keep the USERV account which is used for various other purposes)
• Any other Items?
• Remember the UVACSE Town Meeting on March 20, 3-5pm, in MEC 205; short course schedule is coming out too.
• CS department got a Cray (4 nodes with Infiniband) – UVACSE is doing some benchmarking.

IV. Action Items:

• The sub-committee chaired by Katherine (with Hamp, Bill P., and Ed) will look into implementing the new queuing policy (including configuration, challenges, and a migration plan), and also investigate the possibility of implementing a single queue with resource routing.
• Katherine will check on PBS renewal pricing.
• As part of the Lustre experiment, Andrew, Jim, and Hamp will look into adding an extra drive to each new node.
• Hamp will look into PGI compiler usage.
• Hamp will email Andrew the schema for the mysql database of PBS logs info – and give Andrew access to this database.
• Add UVACSE account and remove RCSC account.

V. Adjourn by 11:45 and next meeting

• Next Scheduled Meeting of ITC ResComp Standing Committee: April 13, 2009 in ITC-2015 Ivy Road, Room #102 (First Floor Conference Room)
  Please send suggestions, additions, corrections to: Tom or Alice