Click here to see the agenda from this meeting.

To Dos & Old Business: Numbers in parenthesis match the numbers on the To-Do Table of September 27, 1999 that we reviewed at this meeting.

Item 1: Research Computing platforms
Ed spoke with Chip Smith - did preliminary test with PBS on orange cluster; had problems that involve rhost file. Chip disabled use of rhost for security, but still getting "work-around" to work because PBS expects to use rhost files.
Hamp said that one issue is that PBS makes a local copy of files during the job run, and then fails when it tries to copy them back to the user's directory. Chip is checking into how/if can configure PBS to run without rhost files; once he's figured it out, he'll duplicate to SGIs (teal).
Tom: What about MPI - doesn't it use rhost files?
Ed: IBM's version of MPI doesn't on SP.
Tom: Bill Pearson is against using IBM's MPI on the "bme" cluster, he wanted to use Mpitch (Argon, public domain version) and it users rhost files. Is there a way to use TCPIP to mask/imitate rhost functionality.
Hamp: Maybe using ssh - shost file does host key authentication.
Dawn: Is it worth contacting other sites/vendors to ask how they do it?
TODO:Ed> Contact PBS and other sites to see there approach to these issues.

Discussion on rhosts files versus host equivalent files. Host equivalent files create more management work for systems.
TODO:Hamp> Look into host equivalent files, see how big a security hole they create, how much management required.

Mark: Does the teal cluster exist yet?
Hamp: No clusters exist yet, machines are being used individually, not widely known, not advertised. Need to get batch queuing system setup (PBS) and have one CPU available for interactive login. Don't have cluster management software for load based interactive logins--> we'll have to develop our own.
Mark: Teal cluster is 3 machines each with 2 CPUs, one interactive, 2 batch. Can we do something simpler/quicker now and then change up the road if need be? Tim S.: Web page to display load?, pick the lightly loaded machine?
Ed: Have something like this now for the Suns in UnixLab. Have had login problems implementing for SGIs.
Tim S.: Can we do nameserver based load balancing/searching?
Hamp: No, not SGI or Suns. We license cluster software for mail servers, may be able to utilize the load balancing part. Vendors have it only for their line, available for their big servers, can't license for workstations like we're using.
TODO:Hamp> Will see if we can use load balancing piece of mail server management software on clusters.
Hamp: Think we ought to turn UnixLab back into interactive workstations and not allow long batch jobs to run on them.
Use Webpage for cluster machines, Orange/Teal cluster same password database as SP/Blue, but NOT Unixlab password. These are not in research password database. If you can log into SP, you can log into Orange/Teal.

Item 2: HSM
Discussed Space allocations/policy page.
How can/do we charge on the HSM for additional space; is use cost of hardware, staff, etc. hard time justifying to cost recovery.
Tim S.: Want to charge to control usage; capping usage is reason to charge.
Hamp: What if we set an upper bound that we won't relax? The only person that had more than 10 GB on maggie was David Seaman, Alderman Library. So what about making 20 GB upper limit.
Tom: Check with library to see what they're doing for disk space for faculty.
TODO:Tim T.>Check with Martha (MRS@) about what library does about disk space quotas.
Hamp: Two accounts with most space is "Public" - which is ours and Alderman's Special Collections.
Tim S: We can't provide Alderman with all their disk space needs within this policy.
TODO: Hamp:> Talk to Martha (MRS@) about Alderman's disk space needs, possibly getting a HMS of their own.
Hamp: Down to last 10 users (not counting these 2 biggest users). Have some 3-5 year old datasets that can't be read, tape is physically damaged and attempts at repair have failed.

Item 5: NPACI/Legion
Ed: No change; They don't have PBS on Intel/Linux yet.

Item 8: Research Computing Support Center
Dawn reported Open House went very well, about 50 Plus folks attended.

Item 4: Software: IDL & PV-Wave evaluation
Ed: Doing a month long testing of IDL vs. PV-Wave starting today. Both current IDL users and other research computing users have been notified via e-mail. This will help us decide which to license. Evaluation copy available on Unix and microcomputer platforms. CDs for PC installs to be available at Res. Computing Support Center (RCSC). Will have two "forums" to discuss evaluation period, one half-way through and one at end, both at RCSC.
Currently have 22 IDL licenses, and 2 PV-Wave, but one of them has moved to IDL already.
Tom S: There's no demand for these at ACHS.
Discussed pricing, PV-Wave pros/cons versus IDL, problems with Data Explorer.

Tim S: what about last spring's ACHS focus groups that asked about visualization tools? Tom S: David Drake is working on this topic. They need more 'feature recognition', IDL wouldn't be mainstream for ACHS users. He has talked with Ed about "ISee"? products that are more medical research specific.

New Business
"Outside" printing at RCSC:
Hamp:ITC doesn't charge for color printing anymore, since got Tektronics 740.
Tom S: ACHS doesn't either; but do have some problems with damage. Do charge for disublimation printer.
Hamp: InR maintains color printers in Small hall, E225, and GIS lab in Alderman. Basically there's no good way to charge. Have to pre-process each file to get a page count and this slows the printing down way too much.
Sue Ellen: Charging is also a issue for Labs, printing PDF files, Files from Toolkit.
Dawn: Prefer not to have public printing to RCSC, don't want Center to just be printout place with lots of associated traffic.

UnixLab policies on multiple jobs/multiple CPUs and locking console
There is a written policy, have to run one at a time and can't lock console. The solution to these being violated is to get the clusters up and running.

Mark: Have had 3 people e-mail Unix Consult asking about web database on Unix.
Tim S: There is a CDP working on this, they're going to implement mySQL, should be available by next semester.
Hamp: Concerned that going to need some kind of "database" police/administer.
Tim S: Do intend to do this, initially target tools for web application development, use minerva and www.people.
Tom S: Please add avery to first rollout list.
Dee: Remember there's also the Faculty web database CDP working on database solution from NT side.

Meeting adjourned around 11:50, next meeting November 29, 1999, 10:30 AM

Go to: ITC Research Computing Committee Home Page

by Tim F.J. Tolson