Workload Write vs. Read Ratio
From: Steven Andrew <postora_at_gmail.com>
Date: Fri, 18 Nov 2011 13:16:51 -0800
Message-ID: <CAE6QNMtBuhARG825S+KdXbP-=4Fj7tiGTbyiD=wiwYjOzAkEmA_at_mail.gmail.com>
All,
I have this little thing nagging me for a while and would like to get opionion from others. For capacity planning excercies, we generally monitor the metric 'physical read total IO requests' and 'physical write total IO requests' over a period of time to get IOPS. I see that Read requests to be around 200 and Write request around 2000. Does this mean the application workload is write intensive with write vs. read ratio of 90:10. What about all the caching that's done at the SGA for reads, likewise SAN caching for write. Shouldn't that be accounted as well when it comes to knowing the workload pattern of the application. How would you go about in knowing the workload ratio for an existing system.
Date: Fri, 18 Nov 2011 13:16:51 -0800
Message-ID: <CAE6QNMtBuhARG825S+KdXbP-=4Fj7tiGTbyiD=wiwYjOzAkEmA_at_mail.gmail.com>
All,
I have this little thing nagging me for a while and would like to get opionion from others. For capacity planning excercies, we generally monitor the metric 'physical read total IO requests' and 'physical write total IO requests' over a period of time to get IOPS. I see that Read requests to be around 200 and Write request around 2000. Does this mean the application workload is write intensive with write vs. read ratio of 90:10. What about all the caching that's done at the SGA for reads, likewise SAN caching for write. Shouldn't that be accounted as well when it comes to knowing the workload pattern of the application. How would you go about in knowing the workload ratio for an existing system.
TIA,
Andrew.
-- http://www.freelists.org/webpage/oracle-lReceived on Fri Nov 18 2011 - 15:16:51 CST