To Dos & Old Business:

Numbers in parenthesis match the numbers on the To-Do Table
of November 24, 1998 that we reviewed at this meeting. 

(1) Discussion on Orange Cluster and (2) SGI Origin200s policies.
------------------------------------------------------------------

Dee (for Jim): Remember that dollars for these computers are general
purpose computers - not just research computing, so access has to open.

TMS: Should accounts on blue automatically get orange? 

EHC: If we say that, they will come and compute.

TAS: 3 classes of rules, I thought:
New Suns -- cluster of batch machines, the existing ones became interactive
	with "police" running on them.

EHC: We have $55K for cluster expansion.  One issue is how many more nodes
blue needs; may need 2-3 nodes @ 5K each (Model 140).

TAS: We have 3 orange 200's - now called SP - in NIS domain for SP put
into load leveller, for long run jobs. GCG will be interactive with one
being front end to others because GCG licensed only for SGI.

EHC: concern about SGI200's, for folks doing NPACI stuff NCSA that's all
SGI - need some place here for development.  If we split them up into
interactive and batch, then hard for them to do work.
What if merge SP and unixlab, Sun, SGI accounts into one account.

TAS: need place to run parallel jobs.

TMS: The SP upgrade that's coming will have 8 parallel nodes.

EHC: Only 1 person needed more than 1 hours (of users moved from watt)
 moved him to the SP, which has newer nodes that are faster.

TAS: OK, two classes of users: 
	any one can get a blue.unix 
	anyone needing SGI, Sun for visualization work or the SP for
	 long running jobs. Graduate and Undergrads need faculty sponsor.

EHC: Simple and easy to do, will reduce the number of accounts
databases.  Have SPUnix lab be a subset of blue.unix so 1
userid/password. This is Unix Systems goal.

TO DO: (TFT) need to write-up this definition of classes and users
	for review. 
		
DMA: How can we map disk space for those with 2 accounts to shared disk
space.  There's about 250 SP users and 1000 Unixlab accounts to merge to
the Home Directory. Do we need to migrate those to Home Dir?

EHC: Planning on doing it Jan 5.

TFT: Timeline for ordering computers.

EHC: First determine blue.unix nodes needed. I need to order by mid
January.  Plan to get rid of SP/Unixlab distinction in late spring.

(8) SP upgrade
--------------

EHC: New nodes are going in on the SP right now. 12 done so far.
replacing slower ones with 160 MHZ chips and more memory and
re-distributing disks. So extra disk on each node for local storage on node
that's running job,  so won't need /bigtemp.

All nodes will be will have at least 512K memory. The wide nodes will
have1 GB memory. 

When upgrade is done will have 4 nodes at 120 MHZ, and the rest at 160
MHZ.  These "new" nodes are three to four times faster than existing
ones.

EHC: For now we'll designate the four 120 MHZ nodes as parallel and can
add to these and let them be ones with mix of 120 and 160 MHZ nodes.

TMS: We told Engineering we will create parallel pool, so have to
reserve nodes; create parallel pool of 8 nodes.

TAS: It'll be better to advertise what update we've done, but not the
number of nodes. Talk in terms of throughput, megaflops, etc. 

TAS to TMS: The Biomedical Engineering parallel pool not being used.
 TODO: (TAS) will write Mitch about this.

EHC: If/when we get the new SP switch we will have to take the whole SP
down.  Could create interactive nodes then.

(4) ADSM
---------

EHC: ADSM - new IBM agreement, it's now under "TIBLY" (spelling) suite.
#1->We know we have to get to get rid of maggie. and
#2->We've got to use something better than we had.
So let's get the latest/greatest before consortium runs out October,
1999. That way we can proceed. 
 Start in January to move from maggie to HSM  And let's not call
it "archive".

TMS: (to EHC) let's plan to add a tape robot to the budget for
1999-2000.

(6) CGC/SGI users
-----------------
No GCG/SGI issues

(3) Home Directory
-------------------

EHC: Home Directory, can lease disk space for an annual fee.  How much
is not back from cost accounting yet, likely to be about $63 per year
for 250MB.  Are in process of annnouncing.

EHC: need to advertise the new HSM correctly so users know what it is.
It is tape; so know users know they are getting files from tape.  As
differentiated from disk wedges, diskspace is yours - never run out, if
you do, you can buy more.

TMS: Have we considered sell backend storage to the library?  

EHC: Possibly.  When maggie proved unacceptable; they put their files on
CD and use a CD tower for PC access; so Unix may no longer be
appropriate for them.

 TODO: Need to document differences between /longtmp, HSM and disk wedges
that uses can purchase.

TMS/EHC: just leaves these as they are, users can go to either HSM or
disk wedge. Don't expand the "/longtmp" pool of disk.

 TODO:(EHC) will move /longtmp to HSM server; to HSM SSA array from
SEAS SSA array.  

EHC: The HSM disk cache is RAID array. Will continue public dataset
(/public and ICPSR). They'll go to HSM on separate file system from the
users.

TODO: (TFT) Reconvene ADSM group in January.
EHC: Think it'll be ready to go in January.
EHC: Possibly use ADSM as way to handle unix backups.

Report from UCIT and ACAC research computing subcommittees. 
-----------------------------------------------------------

TAS: UCIT sub-committee proposing to set up ACHS style research computing
support center. Working on first draft of this proposal.

TFT: ACAC sub-committee discussed issues long term and short goals for research computing
support. Headed up by Steve Stern.  Meet again in January.

Jump to III New Business:

Discuss Research Computing Support Center:

TEH: Let's consider using Small lab as Research Computing Support Center
(RCSC) place. The front of room is mostly empty. Lab has console and 28
workstations.  Seems like it's only about half used.

TMS/TAS/EHC: Problem is backroom is not available space to us, basement
isn't either.

EHC: configure old hobbes, poe, and ptolemy to be switchless serial
nodes?

Agreed to move this meeting to 10 AM Mondays for 1999. Last Monday of
the month.

Go to: ITC Research Computing Committee Home Page