To Dos & Old Business:
Numbers in parenthesis match the numbers on the To-Do Table
we reviewed at this meeting.
(1) Plans for unix cluster
- Final announcements going out
- Faraday goes away March 25th
- Nudging POP users
- Put it on cable TV
JAJ: Funded to do E-school (WATT), doesn't mean we couldn't do Poe at same time.
ECI: If not doing "schools" for each cluster.
JAJ: One account for any cluster with single netapps/Home directory.
ECI: So can log on to whatever machine is best for their uses.
Question of Umenu on Sun Cluster. TAS has a version for Suns.
JAJ: Some issues with compiling on one (e.g., Sun) can't run on another (e.g., AIX)
DMA: Some issures of hypermenu extensions for AIX.
ECI/TAS: Think Robin did for Suns.
TAS: Is there upside to NOT having Umenu on Suns?
ECI/TMS: Don't have to give users all UMENU choices or "old" things
like Upenelm.
TFT: Policy for multiple logins or multiple node logins?
JAJ: Can break files with multiple logins, just like now.
JAJ: Money for upgrade is funded NOW, so Hamp/Dee need to get going.
TODO> Get equipment ordered this fiscal year.
Convene group (cluster group) again for fall.
TAS: Have "installer" to change mail configuration for people.
View as migration from any RS6000 to cluster.
EHC: Current WATT becomes Minerva replacement.
ECI: Same migration steps for this as we used for this cluster
migration.
TMS: Do some sizing; Small Solaris cluster and add AIX nodes.
Maybe poll users on how many might use Solaris cluster.
EHS: Has budget figures.
TAS: Software on only one platform, e.g., ANSYS on AIX versus take to
Solaris - move Khoros to Solaris? Other savings or costs?
JAJ: Is E-school aware? Mitch knows its coming but no details.
ADD TO TABLE AS NEW ITEM #12 - Non-IBM Research Computing Engines (both
SUN and SGI)
EHC: Proposal for next fiscal year. A cluster for researchers, that
is not "entitlement users" (e.g., cluster). A research cluster -
not blue or orange cluster. A kind of extension to SP. NOT SP
nodes - but would run LoadLeveler. Run GCG - visual piece, so
need interactive use greater than allowed with SP.
(2) Archive (hardware and software)
TMS: Archive ADSM is at 3.1 (not 3.0).
TFT: Are we preparing for Dave Saunders departure?
TMS: The ADSM development will be handed off when Dave leaves. Should
have working in few weeks. Test run for month or two for
performance issues, etc. Then plan to migrate to it. HOW?!
TAS: Is it still invisible to SP?
TMS: No - SP can see it now.
ECI: Name?
TMS: Could keep /archive after migration.
EHC: Will take long time to migrate (a couple hundred users).
Doing D. Seaman now - doing it in tape reel order and still takes
weeks! Tape data format is VERY obscure/difficult - basically is
disk format.
TMS: could we outsource it?
TAS: Cornell might be able to do it - They have HPSS wich can read
directly.
EHC: If exclude public and special collections, may eliminate about
1/3 to 1/2 of data on archive. D. Seaman - 25,000 files; 16 GB.
DMA: Are some '94 or '95 datasets that no longer have tapes (Compustat).
So may need some.
Some smaller issues with CRSP.
EHC: We can transfer anything we want. Have to do it user by user.
Contact each one and way will move you but can't touch archive
while we're moving them.
TMS: Can we start to pull stuff off onto disk now?
EHC: Yes.
TFT: Need to create separate archive group - TMS and EHC;
TODO> TAS volunteered to coordinate.
(3) "Research News"
- Tom got draft out.
- Add dialin link, from TMS's e-mail on this topic.
(4) Software Installations and Updates (including GIS/ESRI and GCG)
RLW: GCG - Hamp's new multiprocessor SGI for next year, migrate GCG there.
In interim, keep users on similar platform - So put them on Crimson.
Jacques and Rob testing GCG on Crimson.
60 people NOT using GCG on Avery.
Move the medical one to ACHS.
Move the avery users of GCG to ACHS's crick
Move rest to a Unix lab machine, but only one CPU. So it's same
binary as crick - so support is easier.
72 hours CPU usage total for the year for GCG users, so not a
huge drain on system currently.
JAJ: Alternatives to GCG?
RLW: No - it's THE gene sequencing software and they are NOT
supporting AIX any longer.
EHC: Rather than move faraday GCG users to cluster, move to Crimson
(this is the idea that RLW described).
TAS: SGI rep giving us first refusal on DEMO units.
He just got sheets of stuff at 40-60% off.
Maybe we could avoid transition at all with this.
DMA: ESRI/GIS - Steve Heldreth spoke with Dawn and she gave him info on ESRI.
Haven't heard from Steve H. about mtg. of Univerisity and college purchasing.
EHC: Gauss working.
John Elder using on both SP and nodes.
10 user floating license fine.
RLW: Gaussian (G94) - got Biochem user working.
Runs 3 @ time jobs and she's happy!
(5) NPACI (National Partnership for Advanced Computational Infrastructure)
and SUR Grant
CPR: Web page done!
Will send notice to research group. Already sent notice to SP users.
EHC: SUR - All hardware here and being sent out to users and reconcile
list.
TMS: May have some equipment to send back. So we don't have problems
that had last year with property accounting.
(6) SP
EHC: Just management of users and jobs.
Ongoing issues - users don't estimate resource need very well.
TMS: Anybody using checkpointing?
CPR: She's checked around country.
Bill says it does whole image each time - so lots of disk space.
San Diego not using - everyone writes a checkpoint routine.
TODO> TAS: Will get back on checkpoint issue.
EHC: Over next year may have to adjust nodes and queues.
Not sure how much move to parallel computing.
CPR: Identifying come users for this.
(7) ITS Research Advisory Committee
TMS: First meeting time fell through so trying again.
RLW: Any medical school reps?
TMS: Pearson (Bill) is; maybe Levy?
(8) GRA Training --> "Met"
(9) Departmental Outreach
- TAS talked to MRS - can tread lightly but will share credit.
- He's not sure they see us as equal partners.
(10) Y2K Issues
- TFT met with Lori; following items discussed
Group presenting to Leonard S. with plan week of 2/23.
Task force/Steering focus group on Research - Committees and PI's
Gene Block is Co-Chair of this Y2K effort..
Jeff S. working with Vonda S. on the research part of this.
Getting questions on what OSP gives is Y2K compliant.
TODO> TFT let Jeff know so he could sit in on our meeting and copy Lori.
TAS: Look at Tom's Research News draft on this issue.
JAJ: We have Y2K plan for unix - it's under control.
5 packages will be tested.
EHC: Mail, news, whois, user database - will test fix.
All vendor software identified and commercial packages.
TAS: Our web page has list of compliant software on the Y2K page.
ITC is covered on Y2K - it's spectrometer's etc. that's potential
problem and difficult to test/verify.
JAJ: Netware sheet?
(11) Process Simplification
- Forgot to discuss.
- New item from Jan mtg.
- Need to add to TO DO list.
- Meeting minutes from Jan mtg.
------------------------------------------------------------------
TMS: process simplification - research oriented piece of it. OSP
"G.A.M.S." quarter-million dollar package that tracks grants
dollars, positions, etc. Not really research computing support,
sort of, Melvin Mallory is ITC person on this task force.
------------------------------------------------------------------