I am running Oracle 8.1.7.4 on AIX 5.2 on IBM hardware and connecting to
a SAN for the backup directory (250GB disk directory), backup is through
RMAN scheduled through cron initiated shell scripts.

Database is 165GB and takes up to 2 hours to backup to disk using RMAN.
During this time the system is VERY slow to respond.

Users are complaining about slow response of the system when the level 0
RMAN backups are running at night. To address this, I would like to make
some changes but HAVE TO read up on the changes before proposing them
officially. I am not finding the information in www.tahiti.oracle.com,
nor in Robert Freeman's RMAN book, or any of the Oracle Press books I
have for reference.

I would like to either make the backup more efficient so the users are
affected for a shorter time, or restrict the resources RMAN can grab to
make the backups not affect the end-users as much but over a longer
time.

Can I restrict the number of IO (disk, not tape) writes RMAN is allowed
over some time period? How? Readrate setting is for tape input for
restore, is there a simular parameter for disk output?

I currently have DB_WRITER_PROCESSES parameter set to 1 and
DBWR_IO_SLAVES set to 0. I understand I can not set both parameters. Is
this true, why, and which should be set - and what should the
DBWR_IO_SLAVES be set to if that is the parameter that should remain - I
thought any non-zero number of DBWR_IO_SLAVES allocated 4 channels, but
would like to know for sure.

Currently the LARGE_POOL is allocated 0 bytes. Can I set the parameters
above before changing the large_pool size? Not sure why they would want
to, as changing either parameter requires a database bounce to take
effect - or so I understand it - but I was asked to include the
question.


Oracle LazyDBA home page