> Do a quick test. Create a small copy of the table, with appropriate FKs.
> Start tracing, insert some rows, end tracing.
>
> Drop the FKs and repeat the test.
>
> Report results to list. :-)
Tom Kyte in "Effective Oracle by Design" did such test. In Chapter 1
there is a topic "Test the Overhead of Referential Integrity".
According to Tom's simple test (and I reproduced it locally on a small
tables) every RI constraint adds one LIO per row for insert statement.
So, from 1092396 current reads there are 56400*13=733200 reads just
because of RI constraints.
But I don't think that it's a problem. In initial message we can see
that there are extremely high values for 'SQL*Net message from client'
waits. I think that 'Informatica' (never heard of it before) is just
not fast enough to provide Oracle server with data for inserts.
--
Egor
http://www.oracledba.ru/orasrp/
Free Oracle Session Resource Profiler
--
http://www.freelists.org/webpage/oracle-l
Received on Thu Dec 30 2004 - 22:22:43 CST
