Who has experience running ADSM with more than 1 Gig of memory?  I would
like to understand any architectural limitations  I have found that 1 Gig of
memory on 3.1 causes the server to crash in the buffer pre-fetching routine.
With 4.1 how much memory will ADSM use?  4 Gigs? 10 Gigs?

Jeff Bach
Home Office Open Systems Engineering
Wal-Mart Stores, Inc.

WAL-MART CONFIDENTIAL


        -----Original Message-----
        From:   Richard Sims [SMTP:rbs AT BU DOT EDU]
        Sent:   Friday, August 03, 2001 3:31 PM
        To:     ADSM-L AT VM.MARIST DOT EDU
        Subject:        Re: bufpoolsize and AIX

        >Our tsm server is a hand-me-down with lots of memory - 12GB.  I
        >decided to double the bufpoolsize parm to 2048576.  When I tried to
        >start tsm it complained that it couldn't acquire this memory and
        >failed to start.  I tried to set it to just a few hundred mb more,
        >but had the same result.  I finally just put it back to the
origional
        >value and got our server up and running.
        >
        >Any idea what limit I might be hitting?  I think I must be hitting
an
        >AIX (4.3.3) limit of some kind.

        Rick - Good intuition.  Your prevailing Unix Resource Limits are
probably your
               hindrance.  Do 'ulimit -a' to check.  See the ulimit man page
as a
        lead-in to information.  Also look into those pesky AIX Segments.
Here are
        some notes I have:

                If a Unix system, check Unix Limits values.  Assure that the
system is
                not running out of virtual storage.  If AIX, you may be in
need of more
                than the single memory segment that AIX allows by default,
which you can
                go beyond via Large Program Support conventions.  The scoop:
                 The amount of memory that the process needs may exceed the
size of one
                 data segment (256 MB), which is the default number of
segments a
                 process may use.  The process is in this case killed by the
system.
                 The work-around for this is to enable the program to be
able to use
                 more than one data segment by enabling Large Program
Support, using the
                 following commands:
                        cp -p <Pgm_Name> <Pgm_Name>.orig
                        cd /usr/lpp/adsm/bin
                        /usr/bin/echo '\0200\0\0\0' |
                         dd of=<Pgm_Name> bs=4 count=1 seek=19 conv=notrunc
                 which causes the XCOFF o_maxdata field (see <aouthdr.h>) to
be updated.
                 This allows the program to use the maximum of 8 data
segments (2 GB).
                 Choose the string to use for a given number of data
segments from
                 the following table:
                        # segments    vm size       string
                        ------------------------------------------------
                        8             2 GB          '\0200\0\0\0'
                        6             1.5 GB        '\0140\0\0\0'
                        4             1 GB          '\0100\0\0\0'
                        2             512 MB        '\040\0\0\0'

         Richard Sims, BU


**********************************************************************
This email and any files transmitted with it are confidential
and intended solely for the individual or entity to
whom they are addressed.  If you have received this email
in error destroy it immediately.
**********************************************************************