Minutes of the ITC-Research
Computing Standing Committee
Meeting on
Members:
Alice, Bill, Brian, Dawn, Ed, Hamp, Kathy G., Katherine H., Jim, Mark
S., Martha, Michael, Robin, Sue Ellen, Steve, Terry, Tim S., Tom S.
Chair: Tim
T.
Recorder:
1. Note: Bill will amend PBSPro script so that is
echoes where the user has logged in (lc1 or lc0) and their temp space location.
2.
Aspen Xeon Cluster rollout:
· Ed described the results of three test problems
– in general the Xeons are faster than the AMDs (processor-wise, but also
better in their other components (faster memory, motherboard, and disk)). Ed and Katherine will do some more benchmarking
during the next few months.
· The plan is to get some friendly users trying
this cluster during the next few weeks – and then to make a “general availability”
announcement around April 23-24. Initially
the cluster will just be open and we will see what happens – parallel jobs
will not be given any preference over serial jobs. We will observe what users do before directing
them to a particular cluster or the SMP box.
3.
4.
SP Migration: this was announced
to the LSP conference; working on an announcement urging users to migrate;
users are aware of this and are starting to move.
II. New Topics:
1.
ANSYS licenses: currently we have
a mid-level license. Tim found out from Jerry O’Leary that Robert Hull in the
2.
PBSPro: taken over by
Altair. We are making contact to ensure
that all our arrangements will still be honored – have already received good
technical support from them. New version
has been released, version 5.3, which has some features we may find useful.
3. IP
filtering on jeeves and VPN clients: we already do IP filtering on jeeves.
ESRI wants assurance that we are not
distributing their product “to the world”.
Since there is no password validation provided in the license managers,
it is difficult to expect us to control their licenses – we need a way for
remote users (i.e. from home) to validate themselves to the license manager.
PC users can use UVa-Anywhere, but what about those who use Macs or
Linux? Actually we do “due diligence” when they initially
get the software license – so we should be OK.
4. Security
in Unix Labs: it is possible for someone to remove a workstation from the
network - plug in a laptop in its place - get the workstation network address
and then do most "anything". So to address this problem, we could wait for
authentication to the network or end NFS access to home directories and
substitute a way for mounting to the home directory (note: the latter is in
place now after this was exploited in April). Also could install
SecureNFS or IPSEC on the NetApps and might be able to install it in our labs
over the summer. (Note: this item, as given here, is the corrected version
from review of minutes at the 4/28/03 meeting.)
5.
User in Psychology Dept. (Ben Darling in Denny Proffitt’s lab) with a
large RAID drive and many large image files wants a way for his Windows (XP)
drive to be mounted to the Aspen AMD and Xeon clusters (securely and
routinely). He could use PC NFS or
Windows file services for UNIX. Hamp
reports that Stuart Berr has been trying to get these to work. There are issues with UID and GID mapping. May not be worth trying to do. For now he
will need to move his files with FTP or just use his own WinTel workstation.
6.
Hamp wants to upgrade the Linux OS, applications and compilers on the
AMD cluster so that they are at the same level as the ones on the Xeon cluster
– this will take about 1 day of downtime – to be scheduled after graduation
7. Continue
to support two compilers (Intel and PGI)? Or settle on one? Currently have to provide libraries for each
compiler and maintenance is a nightmare – potentially confusing to users also. On the other hand some use the PGI compilers
on their own equipment – and there may be some incompatibilities between the
compilers. Need to experiment – and see
if we can demonstrate that one compiler (Intel) is better – before we decide. One of many questions is does PGI do basic
profiling?
Next regular meeting:
Go to: ITC Research Computing
Standing Committee Home Page