RMAN - Disk IO Buffer size not what is expected - We want 128K

From: <paul.porcelli_at_scottishwater.co.uk>
Date: 22 Feb 2006 01:35:20 -0800
Message-ID: <1140600919.961212.279860_at_o13g2000cwo.googlegroups.com>

Hi there,
please bear with me as I am not a DBA or RMAN expert.

We are performing some RMAN backups of a 9.2.0.4 Oracle DB to an NFS share.
Our performance could be better so we are using a tool (Ortera - Highly

recommended) to try to optimise.

The tools shows that all RMAN I/O to the share is 1024K. We would like to change this to 128K as we know this will immprove performance based on other tests.

Reading up on RMAN I see that if we have multiplexing over 8 then we should see 128K IO blocks.
We have the following lines in our RMAN script:

---

   $RMAN target=/ nocatalog <<EOF >>$LOG_FN 2>&1   configure controlfile autobackup format for device type disk to
'${W_DIR}/%F';

  configure channel device type disk MAXOPENFILES=16 format
'${W_DIR}/%U';

  configure snapshot controlfile name to
'${W_DIR}/snapcf_${ORACLE_SID}.f';

  configure controlfile autobackup on;
  run {

    allocate channel c1 type disk;
    allocate channel c2 type disk;
    allocate channel c3 type disk;

    backup database filesperset 16 format '${W_DIR}/%U';

    release channel c1;
    release channel c2;
    release channel c3;

  }
  report need backup days 1;
EOF

---

I have basically changed filesperset from 3 to 16 and added the MAXOPENFILES=16 to the configure channel command. From my understanding

this should mean multiplexing is 16. (N.B. We are backing up around 50 files.)

After these changes we are still seeing 1024K blocks !!!!

Could someone explain why this is and if it is possible to get the 12K blocks we desire.

Very many thanks. Received on Wed Feb 22 2006 - 10:35:20 CET