Have a look at the script lgwr_stats.sql. If the sync cost ratio reported by that script is approximately 1, then you can do nothing other than attempt to reduce your commit frequency and optimize log file writes at the operating system level.

If the ratio is higher than 1.5, then you may have an instance tuning problem worth investigating. If the log_buffer is too small, it can increase log file sync times indirectly via redo allocation latch contention. If so, in addition to the latch contention itself in V$LATCH, the above script would show that the average log write size is of the same order of magnitude as the background write threshold. If not, then increasing the log_buffer size is not likely to have a significant impact, but if anything it would in fact increase the average log file sync time rather than reduce it. That is explained in tip on the Ixora web site called Tuning the Log Buffer Size.

At present, our production (Oracle EE 8.0.6, solaris 2.7) is using 156K for log_buffer. The performance tunning for log buffer is ok. but the log file sync is bad. I cannot do anything to relocate the redo log files. I think it might improve the performance to increase the log buffer or tuning the log checkpoint interval.