- I think a "log file parallel write" wait event is recorded by LGWR
along with a REDO write statistic whenever it is writing redo (not
file headers).
Even for all the instances that i manage, the "redo writes" is equal
to "log file parallel write".
I think the LGWR does not record a "log file parallel wait" when it is
writing to log file headers.
- Seeing that your log file sync is almost equal to user commits
which is almost equal to redo writes, i think most of the work is done
is a single session.
You dont have many concurrent sessions.
- Seeing that the wait time for "log file sync" is bit smaller than
"log file parallel write" i think you can avoid unnecessary check
pointing by setting log_check_point_timeout to 0 and
log_checkpoint_interval to 0 to disable check points.
Please correct me if i'am wrong.
regards
Srivenu
Received on Fri Nov 08 2002 - 01:05:49 CST
