Present:  Harvey Newman, Jim Amundson, Greg Graham, Dimitri Bourilkov,
Rick Cavanaugh, Jorge Rodriquez, Ian Fisk, Peter Courvares, Miron Livny

1. Progress since last meeting:

   a. Florida

      - has set aside 5 machines into a mini test cluster

        Globus 1.1.3    installed  (will change to Globus 2.0)
        Condor 6.3.1    installed
        DAGMan 6.3.1    installed
        Objectivity 6.1 installed

      - has had problems with the head node and this has slowed
        progress a bit (eventually traced to a local NFS problem)

      - has not received any requests for accounts.  There are no special
        procedures to follow for setting up an account--just email Rick.

   b. San Diego

      - Ian remarked about problems installing RH 6.2 on a PIV 1.7 GHz
        processor (see mail posted to uscms-grid-testbed@listserv.fnal.gov
        on Tue, 18 Dec)

      - has not received any requests for accounts.  There are no special
        procedures to follow for setting up an account--just email Ian.

   c. Fermilab

      - has set aside two machines for the testbed

      - has not received any requests for accounts.  Fermilab has
        special procedures that users must follow for an account.  Jim
        will detail them and post them on the web site.

   d. Wisconsin

      - The big (and famous!) Condor Pool will soon change from running RH
        6.2 to RH 7.x.  Hence, the original idea of carving out a
        cluster from the Condor Pool is no longer feasible.  Peter will
        set aside 10 (or so) machines in a separate cluster for the
        testbed.  The issue was raised that finding processors which
        support RH 6.2 will be difficult.  It was recognized that CMS
        need to be more pro-active with respect to porting CMS software
        to current OS's.  Miron and Peter indicated that some of
        the machines in the Condor Pool already running 6.2 could be used.

      - has not received any requests for accounts.  As the testbed
        cluster will be separate from the Condor Pool, there are no special
        procedures to follow for setting up an account--just email Peter.

   e. Caltech

      - has carved out a cluster of RH 6.2 machines that will be
        used for the testbed

      - Harvey mentioned that Caltech is investigating the use of a 1 U
        Dual Athalon server, but is afraid of possible problems
        like those Ian had encountered with the PIVs.  Ian
        pointed out that the some groups in Europe have already had
        (positive!) experience with installing RH 6.x on Dual Athalons.

      - has received some requests for accounts.  There are some special
        forms to fill out and fax back to Julian. The instructions are
        now on the web.

2. Grid goals

   - Rick explained that the original goal of the Test Bed was to create
     a "play ground" where developers from the PPDG and GriPhyN
     could *test* software for CMS production, CMS remote data
     analysis, grid architecture, etc.  However, several people had
     indicated that the goals should be more long term and the eventual
     integration with US-ATLAS and the iVDGL (Harvey reminded us that
     GriPhyN and iVDGL are essentially the same) should be considered.  In
     addition full CMS production should be included as one of the
     goals.  Hence two CMS grids were proposed:
      * a US-CMS Test Grid (for software and infrastructure development)
     and
      * a US-CMS Production Grid (to facilitate CMS goals and milestones
        in meeting data challenges and milestones).

   - Miron felt this was a good idea and suggested that the Test Grid be
     used to develop tools for the Production Grid.  In this sense, the
     two grids might only have a logical separation.

   - Ian responded that the Production Grid should be rather more physically
     separate from the Test Grid so that maximum use of CPU power
     could be made available to meet the CMS production needs.  Ian
     mentioned that a large Monte Carlo production will take place in the
     first half of 2002 (*NOT* necessarily running on a grid!!) and that
     this is an indication for the future:  ever increasing production
     challenges will need to be performed.

   - Jorge was not convinced of the need to have two grids: more work,
     not well defined, etc.

   - It was agreed that the details of a "stable" Production Grid will
     be worked out later.  For now, we will move forward with the Test Grid.

2. Decide to use Globus 2.0 beta

   - Since the EAC time constraint was removed as motivation for building
     the Test Bed before Christmas, everyone agreed that Globus 2.0 beta
     should be used rather than Globus 1.1.3.   Hence the decision is
     to install the following:

        Condor          6.3.1
        Condor-G        6.3.1
        DAGMan          6.3.1
        ClassAds        0.9
        Globus          2.0
        GDMP            2.0
        Objectivity     6.1

     *** This requires some care in installing and configuring Condor-G and
         Globus (see Peter's email on Tue, 11 Dec.) ***

   - As no one is familiar with installing Globus 2.0 with Condor-G 6.3.1,
     the timeline for having an operational Testbed has been pushed
     back to mid Janurary.

4. Instructions for adding new grid site

   Greg asked if there are any instructions for adding new grid sites to
   the testbed (hardware configuration, software installation and
   configuration, policies, etc).  As there is no procedure written
   down for this, Rick and Greg will work on writing a preliminary
   list of procedures and posting it on the web.

5. Account Registration

   The issue of centralizing the procedure for registering for a
   USCMS Test Bed account was discussed.  Jim suggested we look at
   what the European Data Grid is developing (which uses LDAP).  A
   look to the (far?) future would include Globus's CAS (Community
   Authorization Service).

6. Next meeting

   It was agreed that the next meeting would take place on
   Friday, 4 Janurary, 2002 at 17:00 Eastern, 16:00 Central, 14:00
   Pacific