Re: performance issue after upgrade to oracle 188.8.131.52 linux 32 bit.
Date: Sat, 7 Nov 2009 11:19:04 -0800 (PST)
On Nov 7, 10:31 am, "Jonathan Lewis" <jonat..._at_jlcomp.demon.co.uk> wrote:
> I've had a quick read through the posts I can see at the moment.
> The critical issue is (as Charles and Randolf have pointed out)
> the use of first_rows_100, combined with Randolf's observation
> that the expected result set is 40 million rows.
> >> First K Rows: K = 100.00, N = 40640900.00
> >> First K Rows: Setup end
that was one of my first thoughts as well but I was a bit reluctant to draw this conclusion because - probably due to the "dynamic proration" feature - there are more ALL_ROWS plans further down the trace file, leading e.g. to this one:
First K Rows: K = 100.00, N = 73255.00
which makes the sort operation very likely look much cheaper, so I wasn't entirely sure that this is the root cause of the problem, given that we obviously don't see the complete trace file.
Oracle related stuff blog: