i would like your opinion on what would get me better performance. a little background. we were having slow performance issues with the database and it looked like the redo logs were are problem. upon investigating, i noticed the db was setup with groups of redo logs pointing to 3 different file systems, and the logs were multiplexed. the setup was: group 1 pointed to r01 and r02, group 2 pointed to r03 and r01, group 3 pointed to r02 and r03, and group 4 pointed to r01 and r02. (archiving was turned on). i indicated we needed to either take multiplexing off or point to different file systems because each time we did a switch, the next group would be writing to a redo log file at the same time the archiver process was trying to read the log to write out the archive log. to me, we were killing ourselves. At the same time, the dba decided to increase the redo log size from 200mb to 800mb. we cut our switching time from 5 minutes a switch to 15 minutes a switch. since we made both changes at the same time and we increased performance drastically, we would like to know what made the biggest performance gain: making the logs bigger (from 200mb to 800mb) or eliminating the i/o contention of the multiplexing and separating the file systems out so we would not be writing and reading to the same log at the same time.
thanks for you input!
I would have thought that the i/o contention on the switch due to having the old and new logs on the same file system was the major source of the problem. Increasing the log size would actually increase the time for the log switch as it has a bigger file to copy (though obviously you would reduce the number of log switches by 4).