Re: Account DB CPU vs OS CPU

From: Christo Kutrovsky <kutrovsky.oracle_at_gmail.com>
Date: Thu, 24 Sep 2009 14:34:41 -0400
Message-ID: <52a152eb0909241134v743cd366tbd202d883769180e_at_mail.gmail.com>

How did you measure CPU usage over the period of time ?

I can't tell you for sure, but here's a few thinking points.

- dbwriter and other background processes - are these accounted in DB CPU ?
- LMS processes - are these accounted in DB CPU ?
- listener, clusterware, process initialization - those are not accounted for

-- 
Christo Kutrovsky
Senior Consultant
The Pythian Group

On Thu, Sep 24, 2009 at 1:07 AM, LS Cheng <exriscer_at_gmail.com> wrote:


> By the way I understand that when doing I/O CPU is not accounted by Oracle,

> so it might be a the cause of such observation?

>

>

> Thanks

>

> --

> LSC

>

>

> On Thu, Sep 24, 2009 at 6:55 AM, LS Cheng <exriscer_at_gmail.com> wrote:

>>

>> Hi

>>

>> I have a 4 nodes RAC running on Solaris 10 each with 5 quadcore CPU (or

>> Dual with Threading not sure). Shows 20 CPU in psrinfo.

>>

>> A few days ago we were running with two nodes only and one node during

>> peak time was consuming an average of 95% CPU. There are 15 instances per

>> node so I obtained DB CPU from time model for each instance during the peak

>> time (one hour), added them up and got around 45000 seconds. Considering 20

>> CPU there should be around 72000 seconds CPU so from Oracle point of view

>> 62.5% of CPU was being used in average but from OS point of view 95%.

>>

>> There arent any more application running except Patrol agent and Grid

>> agent, so how trustie is DB CPU statistic?

>>

>>

>> Thanks!

>>

>> --

>> LSC

>>

>

>




-- 
Christo Kutrovsky
Senior DBA
The Pythian Group - www.pythian.com
I blog at http://www.pythian.com/blogs/
--
http://www.freelists.org/webpage/oracle-l

Received on Thu Sep 24 2009 - 13:34:41 CDT