RE: log file sync

<snip>

First verify that the switching order is what you're reporting, as opposed to the apparent order from the names.

Then, yes, it could possibly help by eliminating read/write competition between "ARCH" and "LGWR". Whether it will actually help depends on whether those disks operate independently enough to improve throughput and whether i/o to online logs is actually your gating resource. The effort to do this is really quite low and it is very unlikely to cause a problem. While you are correct this may not be the root cause, you might as well get it out of the way. If nothing else that will remove the annoyance of the possibility of this being a contribution to the problem from consideration, and probably cheaper than measuring whether it is a part of the problem. So the worst case scenario is that you cheaply do no harm and the best case scenario is you get to the best possible throughput given the existing devices.

Regards,

mwf

I noticed that there are 6 groups, 1,2,3 on disk1 and 4,5,6 on disk2 (shortened for brevity). I'm beginning to think that part of the delays are due to group 2 being written to before 1 is finished, then getting backed up even more as group 3 is being written to while group 2 is still getting flushed and group 1 is still behind as well. I haven't tried it but i'm wondering if I should rearrange the groups so that each group would write to a different set of disks on each switch. I have some doubts that this is the ultimate fix but it wouldn't be extremely difficult to change and try.

Thanks!
--

http://www.freelists.org/webpage/oracle-l

--

http://www.freelists.org/webpage/oracle-l Received on Mon Oct 16 2006 - 11:10:13 CDT