Summary PHYSGI Meeting 4/11/94

                                                         B. Gibbard


    Next meeting will be Monday, April 18, 11:30 - 12:00, in Room 2-160.


    Major users for the week beginning 4/11/94 are expected to be:

        1)  E881 (Hackenberg et al.) - Full out production, multiple data 
                                       analysis jobs. (up to 6 CPU's) 

        2)  E818 (Weygend et al.) - Multiple CPU bound partial wave anlysis
                                    jobs each of moderate duration for 1st
                                    couple of days of week. (up to 6 CPU's)

        3)  E891 (Saulys et al.) - One CPU intensive job at a time.

        4)  E850 (Gushue et al.) - Occasional few hour multi-CPU intensive
                                   running. (up to 6 CPU's)

    Plan for week beginning 4/11/94:

        On the basis of the above summary it seems that all of these users can
        run as described without degrading system performance.  Experiments not
        represented at the meeting should not run more than one CPU intensive
        job on PHYSGI at a time.
        

    Left over problems from the Upgrade to IRIX 5.2:

        CASEVISION debugger is out of date and broken:  although Version 2.2 is
        recommended and we don't have the software immediately available 2.1
        has been reported to work and we have it so we will immediately install
        it.  We will try to get version 2.2 as soon as practical.

        Clock is running slow:  we will monitor it and follow SGI recommended
        fixes (parameter changes).

        Problems with tape drive device files affecting BackUp, users should
        report similar problems, requires looking into.

        X-terminals:  May still be some problems but not well defined.  Users
        should report such problems.

        Tapeman is broken but is thought to be quickly fixable (question of its
        long term desirability remains).


    Other issues:

        Someone physically damaged a tape drive apparently trying to remove a
        tape.  People should be aware that physical force is not productive in
        work with tape drives and should be generally on the lookout for abusive
        treatment of tape drives.

        The problem which results in an E881 job hanging PHYSGI for lack of
        resources, may or may not behave the same under IRIX 5.2.  At a time
        agreed upon by Bob Hackenberg and Tom Nepsee a test to determine the
        new operating systems vulnerability to this problem should be made.

        Options were discussed for handling excessive demand on PHYSGI.

        1) Reduce the number of jobs allowed for each production user from 6
           to some lower number.

        2) Assign two numbers of allowed jobs for each user, one for jobs which
           are to run at full priority and a second for jobs that can be run
           at reduced ("niced") priority to fill in if the system becomes idle.

        3) Use the new IRIX 5.2 capability to assign the use of processors, in
           particular perhaps setting one aside for interactive use.

        4) Using a UNIX batch system.

        It is our intention to get information relating to these weekly
        meetings into the PHYSGI WWW tree within a week so that mailing will
        Emailing will not longer be necessary.