Re: High Number of undo writes

From: Purav Chovatia <puravc_at_gmail.com>
Date: Mon, 21 May 2012 13:49:18 +0530
Message-ID: <CADrzpjE7_HVyaj_G1a9Lsj=jci97nMjQJjOiY1Z7uPJ9XiZA3A_at_mail.gmail.com>



So finally to confirm that it was the SGA values and nothing else, I set the values same as were set in the 10.2.0.5 env. in which we were seeing a huge difference (high no. of undo writes to disk). I was expecting to again see a huge difference (high undo writes to disk). But surprisingly I did not see any difference appearing to mean that the problem was version difference and not SGA difference. I am lost!! Also, in this case we had a test and report where undo writes were substantially less otherwise somebody who sees the test and report with high undo writes would take it as correct because there are no errors, no change of plan, the workload is exactly same and works fine without any issues, etc. So the question arises - how to verify whether such internal statistics (no. of undo writes to disk) are appropriate/optimal or no? Because the difference between 2 was huge - IO statistics were very different and it could have lead to substantial over sizing!

Regards

--
http://www.freelists.org/webpage/oracle-l
Received on Mon May 21 2012 - 03:19:18 CDT

Original text of this message