Andrew,

what version of BACKINT/ADSM do you use ?
Beginning from BACKINT/ADSM 2.2.5.1 you can use MULTIPLEXING to merge the
data from different disks to one datastream which is then send to one tape
drive.
This code also supports the RL_COMPRESSION parameter which controlls a very
lightweight compression algorithm to compress the data on the client side.


Mit freundlichen Grüßen / Best Regards
Gerd Kehrer


IBM Deutschland Entwicklung GmbH
Schoenaicherstrasse 220
Dept. 3293/7103-05
D-71032 Boeblingen
Germany

Web: http://www.ibm.de/ide/esd.html
Hotline: (+49) (0)7031 16 2277
Fax: (+49) (0)7031 16 3619
EMail: backint AT de.ibm DOT com



ADSM-L AT VM.MARIST DOT EDU on 07/16/98 06:29:16 AM
Please respond to ADSM-L AT VM.MARIST DOT EDU
To: ADSM-L AT VM.MARIST DOT EDU
cc:
Subject: ADSM disk read performance ?


Hello,

Does anyone have any tips on maximising the ADSM's READ performance from
disk, when
transferring data direct to a single tape drive (no disk pool on this
particular system)

We're running ADSM Client 3.1.20.3 on AIX 4.3.1, to ADSM server 3.1.1.3
(server runs AIX 4.3.0).
Disk is Ultra SCSI.
We're using ADSM to perform an online backup of a SAP Oracle database (using
brbackup with backint)
Using a blocksize of 128k gives us the best read performance from a dd test.

In the init<SID>.utl we've included:
DISKBUFFSIZE    131072
ADSMBUFFSIZE    131072

WITHOUT the use of more tape drive, or disk pools, what else can be done to
maximise throughput ?
(drive is a DLT7000)

Andrew Swift
Unix Systems Consultant
Retail Technology Services
Coles Myer Limited
*andrew.swift AT colesmyer.com DOT au
*61 (3) 9483 7629