February 26, 2007 Minutes of the ITC-Research Computing Management Meeting

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

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

I. Connection: In the last month, share one thing you did related to research computing support at UVa?

III. Old Business:

Proposed research computing hardware expenditures. The group reviewed the feedback from the forum. There was no consensus from the forum, but some opinions included :

Do not need another small cluster – i.e. not buy new nodes to make "new birch" cluster; more interest in enhancing dogwood: e.g. add Infiniband to some nodes of dogwood – or just add more nodes to dogwood. There was some discussion on who needs Infiniband.
There was interest in storage – so we could purchase additional disk space and make it a priority to add lots more in the future.
There was some dissatisfaction with existing queuing policies – so we could/should review and modify queuing policies that discourage rapid turnaround over long running jobs.
Notes from the forum are on this web site "ResComp-Forum-Notes2007Jan30.txt" Special thanks to Joe Simard for his organized note taking.
One attendee noted that “something for everyone” is not a strategy for success – so what could we offer that people cannot get somewhere else (i.e. a unique resource)?

Big memory; a short-run cluster and a long-run cluster.
Some discussion about adequacy of our existing temp storage.
We do have test queues for code/application development – but we need some stats on the test queues to see how they are being used.
More discussion on how easy (or not) is it for researchers to go back and forth – move data in advance? Stage in their data? Get results back and moved off? And we do not have a good handle on how much data people are moving – or where the bottlenecks might be.
If we knew where their data is now, it might turn out to be faster to leave it and access it where it is.
We can get more information/stats out of process accounting – and could snapshot the disks for usage.
So: we could add more storage – add more nodes to dogwood – add a small cluster with big memory (e.g. 4-8 16GB nodes) – look at queuing no matter what – and do forums more often.
Decided to get estimates for the large memory/small-number-of-nodes option – and then see what might be left. (And look at the costs of Infiniband just so we’ll know.)
We will again have the “buy-in” offer – with the option to have exclusive use of 80%.
We will also ask users about queuing practices and policies – what choices do they have? Want? Need?
We could also see if there is anything we could do to make it easier for folks here to use resources outside UVa.

Two Proposals for "buy-in" to the cluster, both new one and existing dogwood. One gives priority to running jobs anytime you want by dedicating X purchased nodes to exclusive use of the researcher. The other gives priority queue access to the entire cluster. Draft proposal of these two is detailed at: http://holmes.itc.virginia.edu/ResNotes/cluster-purchase.html

Jim & Hamp will come up with rate for buying "priority" time on dogwood cluster.
Should we put cost of "disposal" into cost of buying into nodes? Could be as much as $30/node to surplus.
Tom S.: Genomics might want to buy in to cluster – but infiniband type cluster not appropriate for them. They need large memory computers – check with them.

Linux Clusters: Birch, Cedar, Dogwood - any issues, concerns or upcoming updates/changes?

Any update on conversion to authenticate to Eservices via Kerberos for Unix logins/connections for Blue.unix? (AKA Phase 2 of Netbadge CDP and "unified password" web site).

This transition will involve a major user education issue – especially for blue.unix users.

Infrastructure Supporting Research Task Force (https://www1.seas.virginia.edu/itrtf/). (Maillist is itrtf@virginia.edu)

UVa's participation in ORNL's response to NSF RFP for next generation/iteration of Supercomputing Center. The full proposal was due on 2/2/07 - Mike to check with Mitch on status/update tomorrow (2/27).

SURA survey request- TimT. will contact Phil about pulling together group, possibly by email. Deadline for this has long past, so letting it drop.

IV. New Business:

V. Adjourn by 12:00 and next meeting

Next Scheduled Meeting of ITC ResComp Standing Committee: Monday, March 26th, 10:30 AM in ITC-2015 Ivy Road, Room #220.

Please send suggestions, additions, corrections to: Tim Tolson or Alice Howard

Go to: ITC Research Computing Committee Home Page