Unanswered: Response Time worst after upgrade to 11g
Hi All ,
My Oracle database size is =921 GB
Physical Ram=132 GB
Recently the DB is been upgraded to 11g, From 10
Then Started experiencing the performance worst(Response Time) by 30%
Then our colleagues have enabled the Parameter Filesystemio=ASYSNCH
The original value was SETALL
Now how to get the normal response time from the database base
without parameter filesystemio=ASYNCH,
The application Running on the database is SAP ,
Please advise on this
Which are the Memory Consuming parameters in oracle i e
also , is there a way to find out , determine the should be values for
different oracle crucial parameters based on Physical Ram and Database size
This is my first post on here, but I may be able to help. In 11G the Oracle memory parameters have changed you now have:
You will still have the previous memory parameters such as drilling down to the sga and all of the areas that make up the SGA. It is often recommended though that during the testing stages that you just set the memory max and memory targets as the upper limit of Memory that you wish to allocate to Oracle, then allow the database to resize all of the other areas such as the SGA, shared pool, log buffer etc automatically.
Of course for this to take effect you will need to ensure all other memory parameters such as those mentioned are set to zero.
I assume that you've ran the production workload through on a test system using Database replay beforehand?
What you will need to look at is some of the performance views based on what you have set.
v$memory_target_advice - this will show you whether the memory allocated is ok or whether you can benefit from adding more.
Remember though that there will be a point where just chucking more and more memory at the instance won't make an improvement in performance and in fact could go the other way. The process will spend more time scanning the shared pool for a free block than actually doing work for example
You can also use the advisors in OEM and you should generate statistics as mentioned previously. I tend to invalidate any previous cursors.