Software Coordinating Committee Conference Call
		      March 24, 2005 1:00 PM EST
			  Recorder: C. DeTar

Present:

   DeTar, Maresca, Osborn, Brower, Edwards, Mawhinney, Holmgren, 
   Renner, Jung, Tom Schlagel (BNL), Eric Blum, Watson,
   Pochinsky, Efstratios Efstathiadis (Stratos)

Absent:

   Zhang, Petrov, Simone

** action items
=====================================================================

0.) ILDG File Formats

    DeTar: [Presented ILDG proposal.  See March 17 minutes]
    Edwards: How would the API change to allow writing an ILDG
      formatted file?
    DeTar: That hasn't been decided.
       [There were no further comments.]

0'.) Accounts at BNL

    Mawhinney: We need to get BNL accounts set up.  It takes time.
      I'll interact with Claudio to be sure the principals are 
      notified.

1.) Road Map to Kerberos:
             (Eric, Chip and Don)

    Blum: The BNL ssh gateway will change to a new ssh gateway.
        Then we will install Kerberos as a client.
        All traffic will pass through the gateway.
        The gateway will have about 1 TB for intermediate storage.
        Two hops to move a file.  Perhaps it can be tunneled.
**  Holmgren: I'll look into the tunneling possibility.
    Blum: We haven't considered Kerberos logins.

    Brower: How does it work at JLab?
    Watson: We may be able eventually to log into the QCD machines
        directly.  It would be ideal to have a single command
        that moves files with a single command.
    DeTar: Currently there is no Kerberos client installed
       on the JLab GigE hosts.  Don had to install his own.
    Watson: Installing the Kerberos RPM is on my list. But not at the top.
       I have asked for a sysadmin.
    Stratos: I have an scp tunnel and tested it between
       Columbia and BNL.  Instructions are on the web site.
    DeTar: We need to test it between the labs as well.
**  Stratos: I am applying for an account at FNAL.

1') QCDOC readiness

    Brower: When will a 1000 node machine be ready for alpha
      testers?
    Mawhinney: Four 1000 node machines are ready for initial
      testing.  I hope in two weeks they will be ready for alpha
      testers.
    Brower: I would like to have Dru and James get their codes
      on as soon as possible.
**  Mawhinney: The groups need to decide what initial runs
      they will do.  e.g. reproduce previous runs?

2.) Progress on Asqtad Level 3 tests/interface to MILC? 
             (Dru, Chulwoo and Carleton)

**  Renner: I just received a new level 3 library from Chulwoo.  It seems
       to have solved previous problems.  I am starting to
       test it.
    Jung: We fixed a problem with the wrapper.  It now works against
       the test cases Carleton provided.  The odd dimension version
       works.

   Jung: There may be a memory problem if we want to run
       40^3 x 96 on the initial 1000 nodes.  The local volume
       would be 10^3 x 6.  The DDR memory available on the heap
       to application code is only 32 MB.  The page size is 16 MB.
       The heap is allocated by pages.  So other system processes 
       consume a fair amount.
** DeTar: So Dru needs to be testing such local volumes on 64 nodes. 
       We need to understand how the limitations affect the physics
       we can do.
   Jung: I may need to make some changes in allocation in Level 3
         to accommodate memory requirements.
   Watson: Could we change the page size to 1 MB to make memory
        allocation more efficient?
   Edwards: I believe Peter was going to look at this.
   Watson: When would an 8000 node machine be available?
   Brower: Even UKQCD is running at most 2000.
   Blum: I will be a while before we can run on multiple racks
         since we need to weed out the bad processors first.

3.) Progress for Chroma on QCDOC
             (James and Robert)

    Brower: James has been making progress.

4.) What are the actual disk storage requirement for LHPC Structure
    function at the USQCD? (Future discussion)

    Brower: Norman got the message.  There may be enough $ to provide
      20 TB temporary storage.

=====================================================================
Committee conference concluded at 2:15 PM EST.  
Next plenary call March 31 at 1:00 PM EST