RE: os cache vs. db cache

From: Hameed, Amir <Amir.Hameed_at_xerox.com>
Date: Mon, 9 Jul 2007 18:19:28 -0400
Message-ID: <77A4D80DB2ADD74EB5D7F1D31626F0C004F822A6@usa0300ms03.na.xerox.net>

http://blogs.sun.com/glennf/category/Oracle look at the section "Where do you cache Oracle data?"

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Robyn

	Sent: Monday, July 09, 2007 5:34 PM
	To: Allen, Brandon; oracle-l
	Subject: Re: os cache vs. db cache
	
	
	Thank you Mark and Brandon,
	
	This is the kind of information I'm looking for; I've read

Steve's stuff but it's been a while and the AIX paper is new to me. I *think* we need to make some changes in our approach, but right now, I just want to gather and study as much information on the different options and approaches as possible.

So, if anyone has additional links, documents or experiences, I'd appreciate the input.

thank you ... Robyn

On 7/9/07, Allen, Brandon < Brandon.Allen_at_oneneck.com <mailto:Brandon.Allen_at_oneneck.com> > wrote:

I'd be curious to hear anyone's reasons for preferring OS cache to DB cache.

It seems pretty clear cut to me that it is better to allow Oracle to manage its own cache since it has much more knowledge available internally to help predict which blocks are most likely to be needed again. I've had good results with CIO (Concurrent, a.k.a non-buffered, non-inode-locking I/O) on AIX, but I did increase db_cache_size to make up for the lack of filesystem buffering - in one case from 600M to 1500M, in another I just used CIO from the beginning so there was no before/after comparison, but performance has been excellent with CIO. In the case where I switched from regular, buffered I/O to CIO and increased db_cache_size from 600M to 1500M, the performance of a fixed set of batch jobs improved from an average runtime of 166 minutes to 129 minutes - so a 22% reduction in runtime, but it's difficult to say how much of that improvement was from switching to CIO and how much was just due to the increase in db_cache_size alone.

Here's a great paper specifically on AIX CIO for more info: http://www-03.ibm.com/servers/aix/whitepapers/db_perf_aix.pdf

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Robyn

Over the years, I've read differing opinions on balancing the os cache and the database cache.

Privileged/Confidential Information may be contained in this message or attachments hereto. Please advise immediately if you or your employer do not consent to Internet email for messages of this kind. Opinions, conclusions and other information in this message that do not relate to the official business of this company shall be understood as neither given nor endorsed by it.

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

Received on Mon Jul 09 2007 - 17:19:28 CDT