Hi Tom.

     As you didn't specify what size your primary ADSM disk pool size is,
     what you're seeing would make sense given the following scenario:

     Say your disk pool size is 200 meg, and the file being backed up is
     larger, and you haven't specified a maxsize parm (the default).  As
     the the file is being backed up, the data is going to your disk pool.
     Eventually the disk pool reaches HIghmig value.  When this happens the
     client goes into wait mode, while the server goes thru migration from
     disk to tape.  Let's ignore the time required to mount the tape,
     position, etc.  After the data is copied to tape, assuming the disk
     pool has reached LOwmig, the client is awakened to continue sending
     data.

     When going directly to tape, this stop/start issue is less likely.

     If the above is the case, you should notice a few subtle differnces:

     When going to diskpool 1st, the client begins move data as soon as the
     session is established.  Then wait, send, wait, etc.  until done.

     When going to tape directly, you will see a delay after the session is
     established before data begins to flow.  Chaulk these delays up to
     mounting and positioning the tape.

     The advantage of going to disk is concurrency.  More clients can be
     sending data to ADSM than there are tape drives.  Obviously they will
     all go thru the send/wait cycle, but your overall thruput (how much
     client data is backed up vs wall time) will be greater.  This takes
     some tuning and DISK SPACE to make it fly.

     Hope this helps.

     Al


______________________________ Reply Separator _________________________________
Subject: ADSM Backup Directly to 3590 Tapes
Author:  Tom Stewart <tstewart AT US.IBM DOT COM> at ~Internet
Date:    6/5/97 11:25 AM


Hello -

 I'm very new to the ADSM world and have a question concerning backing up
clients (in my case other AIX servers)
directly to high performance 3590 tape drives in a 3494 library.

I've noticed that when I backup to my disk storage pool located in a 8 way R30
with 2 gig of memory and 120 gig of
SCSI disk space, that backups take about twice a long as going directly to one
of my primary tape pools, has anyone experienced this?

Is there any advantage of going direct to disk rather than to tape (other than
speed of recovery of files, i.e. no mount
wait time)?

Thanks for any input.

Tom Stewart
HONE/IBMLink AIX Support
Boulder, Colorado
8/347-2786
tstewart AT us.ibm DOT com