Re: Do you still ned RAID 5 if you follow OFA?

In answer to a post about the pros and cons of using OFA compliant naming conventions on a raid 5 disk farm, Tom Pall stated >> The write penalty for RAID 5 is 1/n where n is the number of disks used to create the array. <<

Please explain. I always thought the write penalty for Raid 5 was that every time you write you have to do two I/O's. One to write the data and one to write the parity and both I/O's must complete before the I/O is done. So how do you get 1/n?

As far as using OFA on Raid 5. OFA is a naming and organization standard usually obtained through links to the underlying data files. Raid 5 is a means of protecting data from storage device failure. Do not allow the true underlying structure to confuse the issues. Yes, your temp, data, and index tablespace could all end up on the same real physical devices, but the OS level stripping should even the I/O activity across disk devices resulting in a more consistent sustained I/O rate. I have seen the database backup also placed on the same (and only) raid stripe. Raid 5 only protects against the loss of one storage device per stripe so if this is the case you need to get the backup copied to tape ASAP, to ensure recoverability.

Remember on some systems Oracle can not write straight to tape so some places have no choice but to put everything together. If you have 20 disks I would recommend that you set it up as 4 stripes of 5 disks rather than 1 stripe of 20, but this may not be possible. If there was anything I would want to move off the Raid 5 device to regular disk it would be at least one copy of the control file and one if not all copies of the redo logs. Received on Mon Dec 07 1998 - 08:40:40 CST