How reliable are IO timing statistics in v$filestat/x$kcfio in Oracle 9.2

From: "Hameed, Amir" <Amir.Hameed@xxxxxxxxx>
To: <oracle-l@xxxxxxxxxxxxx>
Date: Tue, 23 Nov 2004 12:03:05 -0500

Hi All,
I am gathering some data on the amount of time that Oracle took to read
and write (IO statistics). I am using v$filestat/x$kcfio for this
purpose. I would like to know how reliable are the READTIM, WRITETIM,
AVGIOTIM, MINIOTIM, MAXIORTM, and MAXIOWTM statistics. Here is what I am
trying to achieve. I am in the process of implementing ODM (Oracle Disk
Manager, from Veritas) in my lab. We are running Solaris 2.8 and Oracle
9.2.0.4 (64-bit). Since ODM offers RAW like performance features, my
goal is to run tests as described below:
-- Run a test with the buffered filesystem first and collect IO
statistics then
-- Enable ODM, increase the SGA size to ensure that what was cached in
the filesystem buffer cache can now be cached in the SGA. Run the above
test again, collect the IO statistics and then compare to see the
improvement/degradation.
Is there a better way to gather IO statistics (sar, iostat), etc than
from v$filestat/x$kcfio ?

Any help will be appreciated

Thanks
Amir


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

How reliable are IO timing statistics in v$filestat/x$kcfio in Oracle 9.2

Other related posts: