To reduce log file sync you need to reduce number and size of IOs per physical device used for redo logs and remove all possible contention (i.e. dedicate disks).
Reduce number and size of IOs per physical device: - add more devices - reduce frequency of commits - check you log_buffer if it's not too smal and gets filled to 1/3 too often - usually not the case with OLTP anyway. - reduce amount of redo (don't update when not required - like columns with the same values and rows that can be skipped). - "do less work" on application - you would be surprised how much can be saved and optimized there. I would look on these points from bottom up.
Removing contention - are your backup destination/datafiles/archive log destination sharing the same physical disks? If yes than your IO subsystem gets more saturated. Consider optimizing your IO layout - dedicating seprate disks to redo logs.
RMAN backup itself doesn't influence LGWR unlike online backup without RMAN.
I am using 10.2 on linux for OLTP system. I am getting log file sync as top event. I can not change disk or combine transactions in one or some thing other as suggested in metalink.
Does rman backup has any effect on this event as I see this event going up during backup time and performance goes down the drain.
How can I reduce this wait event?
Sandeep -- http://www.freelists.org/webpage/oracle-l
-- Best regards, Alex Gorbachev