Re: RMAN, vanilla or "tweaked"
- From: Joseph Amalraj <joseph@xxxxxxxxxxxxxx>
- To: Michael.Kline@xxxxxxxxxxxx, oracle-l@xxxxxxxxxxxxx
- Date: Thu, 12 Jan 2006 14:14:23 -0800 (PST)
Setting the BLKSIZE does help, I have set it upto 512K
also SIZE_DATA_BUFFERS and
NUMBER_DATA_BUFFERS on the netbackup size also needs to be adjusted.
Increasing the number of channels helps to increase the backup speed but
reduces the restore speed, because multiple channels are multiplexed on the
tape.
Number of channels for the backup should match the total of multiplex
configured in the Netbackup policy.
If disk i/o is tuned using SAME, maximum speed can be attained by using 1
channel per tape drive.
Filesperset is another RMAN parameter that can be tuned.
Oracle had released a white paper related to RMAN tuning and this is very
informative.
Joseph Amalraj
"Kline.Michael" <Michael.Kline@xxxxxxxxxxxx> wrote:
st1\:*{behavior:url(#default#ieooui) } We are
trying to squeeze out more performance on RMAN. The network folks said our back
channel is only 60% used and NetBackup has found quite a few errors and
corrected them.
What other settings other than more tape drives have you found to be very
useful?
For instance, how many are tweaking the blksize on the allocate channel
command and did it really help?
How many are messing with the backup_tape_io_slaves and what numbers worked
and what numbers provided no help at all. I know on some things 4 may be WAY
better than zero, but 10 provides maybe 2% more, thus useless.
Also, what kind of throughput is possible? We are doing about 22 to 30MB per
channel thus far. But on one test pushed that to 47MB. How come we can?t push
47MB all the time?
We?ve got several databases in the multiple TB and at least once a week will
need to do a full backup on.
Other related posts:
