Minutes of the
ITC-Research Computing Standing Committee
(ITC-2015 Ivy Road, Room 102)
Meeting on August 25, 2004
Members:
Alice, Bill, Dawn, Ed, Hamp, Kathy G., Katherine H., Jim, Mark S.,
Martha, Michael, Robin, Sue Ellen, Steve, Terry, Tim S., Joe S, Tom S.
Attending: Bill,
Ed, Hamp, Kathy G., Katherine H., Jim, Mark S., Martha, Steve, Terry, Tim S.,
Joe S, Tom S., Tim T.
Chair:
Tim T.
Recorder:
Joe_S.
Click here to see the agenda
from this meeting.
Chair: Tim T.
Recorder: Joe Simard
I. Corrections to minutes
from last meeting on July 26, 2004? None
II. Old Business:
- Schedule from Hamp on getting
these done:
- Splitting SGI (crick)
behind VPN firewall for HIPAA users.
-
Complete except for turning off NIS
service
-
As NIS does
not work through the firewall a slave server would be required if NIS
needs to be behind firewall.
- Teal Cluster: Put 2
processors removed from crick and put into one teal node; use Craylink hook them together. If
other 4 CPUs are compatible, link them as well. Put PBSPro
on them.
- Setup
"teal-login" machine. - Block interactive login on all but
"teal-login". Use one of SGIs from Unixlab as teal frontend.
- Per timeline these
two items were to be done by second week in August
-
Craylink has been ordered.
-
On receipt of Craylink,
reconfigure Teal and two processors from Crick as two nodes with four
processors each. The two nodes may have
different amounts of memory.
-
Configure an SGI from the Unix
lab as an interactive front-end.
- Make old "dump.itc"(RS6000 H70) the 64-bit frontend of SMP (mp0.itc). DONE
- Rename (re-number)
Suns in E225 to be sun1 to sun8.unixlab
-
Renumbering is planned for week of August 23rd
(this week).-
- Aspen & Birch
cluster upgrade to 2.6 kernel. Timeline
dependent on ROCKS upgrade.
-
Bill mentioned Red Hat has not announced their next enterprise
version yet.
- Additional disk space for /longtmp and script to automate checking usage and
reminding users...Waiting for disk storage to come available, likely
sometime in September
-
Still waiting on storage to become available from
Library.
- Schedule for adding
additional disk space to /common on jeeves for
pending software installs (e.g,. SAS, Maple).
-
Hamp ordering disks
- Update on SEAS SUR grant clusters.
-
Reinstallation is needed due to unauthenticated user
access.
- Proposal for Linux cluster
support as 'for-fee' service.
- Need guidelines for
how to handle support for researchers who don't purchase cluster support
but expect assistance and guidance from ITC for their cluster.
-
Need to set rate for Unix Systems cluster support.
- 102 Small Hall Unixlab closed August 1st to give space to CS. Praise
& thanks to Hamp & his staff and Sharon Drumheller & her staff
for working diligently to get everything out so quickly - The room is empty.
III. New Business:
- New Linux cluster and condo
scheme
- Each participating researcher/department
would pay for some number of nodes with a percentage of those nodes made
part of a common pool available to all cluster partners.
- Each participating researcher/department
would also pay a portion of the cost of the infrastructure (gigabit
switch, etc) and cluster administration provided by ITC.
- Nodes in common pool
would be available to all cluster partners on a first come first served
basis except when a researcher/department submits a job for the full
number of nodes they funded.
- PBS Fair Share would
be used to manage allocation of nodes. Although
the number of nodes which a researcher/department funded may not be
immediately available, Fair Share would provide the funded number on a
priority basis as soon as they became available.
- Need to determine if
there will be a minimum number of nodes funded by a researcher/department
in order to participate.
- Participation would
require a three year commitment after which nodes would be returned.
- Fair Share could
facilitate one or more researchers/departments pooling resources to
become a partner.
- ITC contribution is
expected to be about 32 nodes.
- Storage is expected
to be about 3 terabytes.
- Configuration is not expected
to include MyriNet interconnects.
- Timeline is to get
departmental commitments and order equipment by December 2004.
- Marketing to researchers/departments
-
Need to manage expectations
-
Consider offering higher % to early adopters
-
Notice in next RSCS Newsletter
-
Need pre-announcement notice to capture grant proposals
being written now.
- Rather than using old Sun
Ultra's in Carruther Machine room, use most of Unixlab budget to buy new solaris
rack mounted servers.
-
Unix lab budget is about $20K
-
Target users are serial users on Aspen
clusters of Matlab, SAS, etc.
-
$23k budget is expected to adequately fund 12 processor
cluster and rack.
- Other – Aspen
and Birch clusters.
-
Aspen 2yr warranty has lapsed. So prospect is that when
Aspen nodes fail, the cluster gets smaller, if warranty
extension isn’t funded.
-
Birch warranty expires in December.
-
Birch has 2/3 the number of nodes as the Aspen
cluster.
-
Third year of maintenance being considered for both Aspen
and Birch. If funding is not adequate for both (~$7k Aspen,
~$5k Birch), priority will be to fund Birch warranty extension.
V. Adjourned by 10:30 AM and next meeting
- Next Scheduled Meeting: Monday, September 27th, 10:30 AM in ITC-2015
Ivy Road, Room #102
Suggestions, additions, corrections to: Tim Tolson or Joe_Simard