primary keys and dictionary overhead

I am working on a nearly 20 year old legacy system. The OLTP peaks at about
100,000 executions/minute and we expect this to continue to increase. When
the system first came out, they had to make sure there were no constraints,
unique indexes, triggers, synonyms or anything to make sure the system would
perform. I can understand that 20 years ago. We just started adding some
unique constraints. I wanted to make the  unique constraints into primary
keys. One of the guys is worried about dictionary contention from the
primary keys. I have worked on some pretty large OLTPs and never saw any
issues with this.
has anyone seen this? We are on 10.2.0.5 (I forget the PSU level) on HP-UX.
If I was going to test this, I'm not really sure what I would look for. I
don't have a way to simulate the volume of users and executions we get in
prod and then just check the waits.
Any suggestions on what to look for? I'll need to show some test cases.


--
http://www.freelists.org/webpage/oracle-l

Other related posts: