Originally Posted by
PAVB
As far as I can see wait events on db_file_sequential_read account for about 20% of the total elapsed time therefore even if you manage to get rid of 100% of this particular wait event - which is not possible - you are looking at a performance gain of about 20%
What are the chances of rewritting this Informatica mapping as a PL/SQL block? - just for testing purposes :)
I would write the code without using explicit cursors a.k.a. set-based and, I'll eventually use APPEND hint during insert. On the first try I'll commit only once at the end of the job.