Re: converting from non-rac to rac performance issue
Date: Wed, 10 Dec 2008 13:50:04 -0500
It's a good practice to reevaluate (read increase) SGA/buffer cache size when database got moved from filesystem to ASM. In case of filesystem you had implicit advantage of filesystem cache, on ASM it mostly direct access, so you have to allocate more memory for Oracle buffer cache to mitigate cache misses.
On Tue, Dec 9, 2008 at 8:45 PM, Kumar Madduri <ksmadduri_at_gmail.com> wrote:
> Our environment was upgraded from 10.2.0.2 NON-RAC using SAN DMX storage
> on file-system to 10.2.0.4 RAC using ASM on SAN DMX.
> We are observing performance issues with high amount of elapsed time for
> dbfile sequential read.
> The same job before upgrade completes in 3hrs50mins but now taking 5hours.
> We did take the tkprof output and notice high amount of dbfile sequential
> reads/scattered reads.
> After we applied the patch 7243560 - RAPID INCREASE IN DBFILE SEQUENTIAL
> READ AFTER MOVING TO ASM the job has completed in 4hrs 30mins.
> Almost all queries run twice the time it runs on single instance NON-RAC
> NON-ASM all storage on DMX.
> SR with oracle support is taking its own time
> Did anyone have similiar issue.