Re: Performance metrics

From: Karl Arao <karlarao_at_gmail.com>
Date: Sat, 14 Apr 2012 00:27:02 -0500
Message-ID: <CACNsJnf=pKJ8U-DTPb=kWReN3tXCb8i2dA05XR2hEGSeeFj-dA_at_mail.gmail.com>



And BTW, from this issue here
https://lh6.googleusercontent.com/-INmcLLwj-9g/T4gLGZwIIfI/AAAAAAAABkg/L5AtWmhi-Ts/s2048/20120412-template-PGA3.png I
don't think that resource manager (limit the amount of CPU) will be of any help. The boxed green are the known Oracle CPU cycles and if you start to limit stuff you may be affecting the encircled green workload on the left side that is not doing any harm on the server... causing another problem altogether.
Unless we dig into the "usage tracking" of BIEE on the time slice where the issue happened and have a graceful way of mapping those jobs/users to a consumer group just like what rittman did here http://www.rittmanmead.com/2010/01/oracle-database-resource-manager-and-obiee/ but
then again, by enabling that feature it's not addressing the problem.  kswapd will not care if you have resource manager and it will still do tremendous swapping which translates to "CPU wait IO" and high load average.

So fixing the SQLs (rewrite,etc.) and drilling on what can we do on the BIEE side (cache seed,design review,etc.) would be the best thing to do here.

--

Karl Arao
karlarao.wordpress.com
karlarao.tiddlyspot.com

--

http://www.freelists.org/webpage/oracle-l Received on Sat Apr 14 2012 - 00:27:02 CDT

Original text of this message