Re: redo log on veritas filesystem

Are you seeing your performance hits only when a log switch is occuring? I personally would size your redo logs so that you only switch every 20-30 minutes. I use ufs on solaris (with logging) for all oracle files except data files. My log destinations are hardware raid 0+1 with only 2 redo logs per disk (and each log disk is never read/written to at the same time. ie When log1a switches it then writes to another disk for redo2a and the archiver process is then reading from disk 1 at that point). I also have 2 arch destinations and mirror my redo logs accross 2 diffderent SANs so I'm covered in the event 1 SAN goes per-sahped. Hotbackups are stored on tape and 2 different independent raid 5 volumes.

Also, if you really need the I/O boost, I would turn to raw devices before using vxfs. We use storage foundation for oracle but really only use the vxvm functions and then we only use vxvm to move disk groups between servers. We moved away from vxfs about 1 year ago and haven't regretted it one bit. Received on Wed Mar 02 2005 - 15:31:17 CST