> Why would a query of all_constraints cause 7 million consistent gets
> and take nearly two minutes ?
[quoted text clipped - 28 lines]
>
> Elapsed: 00:01:56.08
(the following is easy in OEM and similar tools):
Look at the definition of the all_constraints view.
For each table mentioned there, in both instances, see if there are any
statistics, and be sure all indices are there and valid.
I'm guessing the CBO is being fooled by the recursive self joins and
funny statistics (or problematic indices) into repetitively full
scanning obj$. I'm predicting your faster instance won't have
statistics. Make wooeeewoooeee sounds now.
jg

Signature
@home.com is bogus.
http://www.janus-software.com/fb_fyracle.html