Agenda for the ITC-Research Computing
Standing Committee Meeting

June 26, 2002 at 9:00 AM in Astronomy 117

Members: Bill, Brian, Dawn, David, Ed, Hamp, Kathy, Jim, Mark S.., Martha, Michael, Robin, Sue Ellen, Steve, Terry, Tim S., Tom S.
Chair: Tim T.
Recorder: Alice

I. Review and Discussion of committee charge - Dr. Bob Reynolds

• Mission and role of committee and members
• Ground rules and process
• New research support maillist needed? (I've made itc-rescomp an unmoderated mail list).

II. Items on Linux Cluster from updates

• (EHC) Concerns about Aspen Systems replacement parts turnaround time. Getting a spare HP network switch card.
• (EHC) Temp space - how to manage temp storage during a job and how user gets access to it her job completes?
• (EHC) sendmail setup - need to configure the cluster to allow sending mail but not do local mail delivery
• (EHC) Do we need to lock users out of compute nodes? Current configuration allows interactive login. Restrict login to owner of running job on that node (as suggested in NCSA workshop)?
• (EHC) Do we need to use our ssh or Open SSH currently installed?
• (EHC) Downtime for moving Linux Cluster in unmanned machine room - projected date
• Who should be test users for Linux Cluster and date for this?
• (TFT) Testing Linux Cluster - has test suite provided by Aspen Systems been run and compared against their in-house results?
• (TFT/TEH) Cluster System monitoring: Hamp mentioned Linux Cluster uses Ganglia to display node status and Steve working on porting to other OSes. What about NetSaint (http://www.netsaint.org/) or BigBother (http://bb4.com/) ?
• (TFT) Cluster Management Software: We're using NPACI Rocks and Ganglia. Have we gotten new version of Aspen Systems Cluster Mgmt Software? How does it fit into this picture?
• (TEH) Should we install PBSPro rather then the OpenPBS that comes with NPACI Rocks? Universities get PBSPro for free and has the advantages over OpenPBS listed on the webpage http://www.OpenPBS.org/product_comparison.html . This would also be consistent with the version of PBS used on our other Unix Platforms.
• Installation of Maui Scheduler?
• (TFT/TEH) IMSL for Linux Cluster has been ordered. PGI compiliers being ordered, getting 10 simultaneous user license for $1,205

III. Items from Linux Cluster workshop

• (TEH) Should we reserve some nodes (say 4) to be used as a test cluster for new software and kernel patches?
• (TEH) Should we install kernel patches to enable hardware performance counters. The install Performance Application Programming Interface (PAPI)? See http://icl.cs.utk.edu/projects/papi/
• (TEH) Should we install M-VIA to improve the latency and bandwidth of the Gigabit Ethernet interconnect? http://www.nersc.gov/research/ftg/via/
• (TEH) Should we install the following Scientific Libraries?
  • LAPACK http://www.netlib.org/lapack/index.html
  • ScaLAPACK http://www.netlib.org/scalapack/scalapack_home.html
  • FFTW http://www.fftw.org/
• (TEH) Is PVFS installed? It provides high performance parallel I/O, and is included in the NPACI Rocks distribution.

IV. Other action items and issues?

• Overall picture of how the Orange and Teal clusters, the UnixLab workstations, the SP nodes, the SMP box, the new Linux Cluster can be a coherant, cohesive set of research computing resources, instead of distinct, disparate ones. For example, can and should we use PBS & Maui scheduler for all of these?
• Moving from Loadleveler to PBS/Maui on SMP box? And on SP nodes?
• (EHC) Unixlab Script/System Usage/Monitoring website: Linux Cluster uses Ganglia to display node status. Steve working on getting same software working on AIX, Irix, & Solaris so that users can use these tools for providing status on all research computing platforms. When fully functional, this will replace current Unixlab usage page.
• Need to keep in mind that we told ACAC they would have input and oversight to configuration and implementation of Linux Cluster and SMP box. How and at what level of detail do we go about giving them some choices?
• Maple 8 installation? Timeline for installation?
• Others?

V. Adjourn by 10:30 AM and next meeting

• Monday July 29, 2002 at 10:30 AM in 117 Astronomy

Suggestions, additions, corrections to: Tim Tolson or Alice Howard