Minutes of the ITC-Research Computing Group Meeting
March 29, 1999 at 10:30 AM in Astronomy 117

Members: Chris, Dawn, Dee, Ed, Hamp, Jacques, Jim J., Mark S., Sue Ellen, Tim S., Tim T., Tom S.,
Conveners:Tim T.
In Attendance: Dawn, Dee, Ed, Hamp, Jim J., Mark S., Tim S., Tim T., and Tom S.
Recorder: Tim T.
This webpage by Tim T.
Next Meeting: April 26 at 10:30 AM in Astronomy 117.


Agenda from the meeting.

To Dos & Old Business:

Numbers in parenthesis match the numbers on the To-Do Table
of January 25, 1999 that we reviewed at this meeting. 

Item 2: Orange Cluster
----------------------
Hamp: Orange cluster here, ready to put in.  Loadleveler for non-AIX Unix
OS has been withdrawn as well as loadbalancing software.

Names will be: orange#.unixlab for Suns, teal#.unixlab for SGIs. 
Account rules:


Jim: This list of orange/teal cluster rules are fine. These are
mid-range compute engine available to all by request.  This cluster will
also serve to move the long running jobs off the unixlab machines and 
created better interactive graphic machines.

Hamp: Will need loadleveling software; will need to consider if
other loadleveling software is multi-vendor, will we drop  LoadLeveler?
Ed: is looking into some.

TMS: Let's pick best loadleveling software and then see how it works
with the SP. 

Dawn: Condor has advantages, gives priority to console.

Hamp: Need loadbalancing software too, may write own.  Will look 
at Sun product, which works for one domain name, so may need to 
have different names for teal and orange clusters.

Hamp: Question about what to do with galen and dayhoff names.  Assign to
graveyard so that users can't telnet there?  Group agreed Yes, and have
message tell telnet users if trying to get to dayhoff for web services,
go to avery.


Item 8: SP upgrade
------------------
TMS: 8 new parallel nodes are mostly for the Engineering school.  Mitch
wants pre-emptive priority.  We need to be very active in getting the
E-School to use the parallel nodes in parallel. 
He hopes they'll naturally get most of the CPU cycles, without us having
to hardwire priority-but we will need to do so if needed. They did pay
for upgrade.
ToDo>Ed: Announcement of upgrade will say they get priority.  He'll draft
and send out for review.
Draft announcement about new SP job classes  

TMS: Finally concluded last order for 1998. Upgraded SP switch and all
adapters. Both switches (frame 1 & 2) and adapters are on order. 
New switches and adapters enable upgrade of software that wouldn't have
been possible with old ones.  New nodes needed new switch. 

Hamp: Installation of these will require system shutdown.  IBM will 
do upgrade.
Let's announce we'll bring parallel nodes online April 19th and will
switch to new queues then too.

Tom: TotalView for DecAlpha's MPI aware debugger, is $4 to $8 thousand,
is multi-platform.

ToDo> Hamp: IBM partner sent URL about parallel tools, he'll send to
group.
(which he did: www.myrias.com (called:PAMS)

Item #4: ADSM
-------------
Hamp: He's moving data into ADSM; still just ITC data, public, etc. Will
still be doing users one at a time.
Names for volumes:  Need two writable systems to get around root access
security issues.

  All agreed with names as proposed.  (And there was much rejoicing!)

Item 9: Research Advisory Committees
------------------------------------
Tom: UCIT research sub-committee has another draft proposal that takes
into account the ACAC research sub-committee's objections.

--------
Reviewed the "To Do" and "Done" table.

- - - - -
Meeting adjourned at 11:30, next meeting April 26, 1999, 10:30 AM
=====================

<== Go to: ITC Research Computing Committee Home Page


by Tim F.J. Tolson