PS the waits are all IO related - log buffer , log file sync
etc

... are you absolutely sure the I/O component of these waits is what is
actually slow.
You know there is more to the anatomy of a log file sync and log buffer
wait than
I/O right? A process get posted though the post/wait mechanism (port
dependent)
and THEN it comes out of the wait. In short, there is IPC wrapped around
the LGWR
loop. I wonder if there is something in the 10gR2 post/wait driver that
is busted?
Quite possible. How do they compare with _logging_disable=TRUE? When
this is set there
is still LGWR posting and lfs waits, but LGWR doesn't actually write to
the log.




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