Minutes of the ITC-Research Computing Standing Committee

Meeting on February 2, 2004

(This is the January meeting delayed one week by snow)

Members: Alice, Bill,, Dawn, Ed, Hamp, Kathy G., Katherine H., Jim, Mark S., Martha, Michael, Robin, Steve, Terry, Tim S., Joe Simard, Tom S.

Attending: Alice, Ed, Hamp, Kathy G., Katherine H., Mark S., Martha, Robin, Steve, Tim S., Joe S., Tom S.

Chair: Tim T. Recorder: Alice

Click here to see the agenda from this meeting.

I. Corrections to minutes from the last meeting on November 24, 2003 (no December meeting)

None.

II. Ongoing Discussion Topics:

1. SP retirement on 12/15/03 – went off successfully – Hamp blocked logins in mid-afternoon – no complaints or last minute requests. Going to try to arrange a sale-in-place for the remains.

2. Magnetic tape retirement -- in final phase. No 9-track tapes now.

3. VPN and IP filtering for FlexLM license daemons. Several vendors are now requiring it, not just ESRI.

· Unix Systems set up a separate server (aix.license.virginia.edu); testing and documentation proceeding; Mathematica done; working on other packages 1-by-1 and notifying affected users first.

4. UnixLab machines – Console logging turned on? Not yet – may need to turn on process accounting. Notes:

· We need usage stats to make/justify choices about future uses for the room – so we want to see who is logging in and out of the console.

· The machines are aging – do we replace with workstations? Or PCs running Linux? Or nodes? In any event we need to start publicizing that the SGIs will be going away as they fail – so need to generate a list of who is using what on the SGIs.

· Decided to put the clustering of the Unix Lab machines on a back burner – and instead work on making the orange and teal cluster ready as a compute cluster.

5. PBS Pro upgrade on Birch to get in version sync with Aspen

· Hamp will check with Bill. – in theory it sounds do-able -- Bill should pick a date and we’ll announce it.

6. Update Orange and Teal clusters to latest PBS Pro version – timeline?

· Proposal to change configuration based on previous discussions: create PBS cluster with front-end for each cluster – put 1 SGIs behind ACHS firewall -- then make a 4 CPU system. Can swap 0200-3, rename it Crick, and then ResComp gets current Crick. -- Hamp will put together a schedule and send it out.

7. Update on where Enterprise Linux rollout, licensing, and pricing stands:

· Hamp can produce (or collaborate in producing) a distribution based on Enterprise Linux and can dedicate a computer to compiles.

· Systems is just starting to build it and will be able to support it. Hamp thinks he can do this without it taking a person full-time to do this. Do-able within a couple of months.

· RedHat is saying that Enterprise will not be updated every few months and this is appealing. Departmental research clusters do not have to be general purpose – they usually want to do one thing as fast as possible and do not want to update frequently. Hamp will be issuing RPMs for the Enterprise distribution to keep up with updates.

· We should stick with the version of Linux that we are using on the clusters. We should encourage departmental clusters to use NPACI Rocks as is.

8. ITC’s research computing “white paper” draft will be out this week for review.

III. New Topics:

8. Software Installations: Intel 8.0 and Scalapack – planning for the next release of IMSL which will be built with Intel compiler rather than PGI and may require older version of Intel compiler and MPI:

· Intel upgrade likely to be a “moderately big deal”—will have to recompile every library that Fortran touches and users will have to rebuild – this is not urgent – Hamp will work on a schedule.

9. DNS allocations in the 10.x.x.x range – issues for Linux clusters, both within ITC and around UVA:

· Discussion of routing issues and where duplicates could potentially cause problems.

· Currently we have no rules – now need some policy on how to choose non-routable addresses for units at UVA (including ITC).

· Robin’s group is working on some network database tools and want to move forward in some managed way.

· Hamp suggested reserving a block of addresses that departments could choose to use if they do not route them AT ALL.

· We do need to have more documentation for guidance about how to choose addresses – and we need to be clearer about what addresses we are going to use.

10. Remove qmgr permissions from staff user-ids and provide a safer alternative? Agreed to use user-id "rcsc" – so that would force one to switch to another account and alert one to be careful about what you are doing.

Next meeting is Monday, February 23, at 10:30 AM.

Go to: ITC Research Computing Standing Committee Home Page