RAID10 explanation and mirroring setup

I've been following the RAID discussions and I'm having a bit of 'DBA block' when it comes to understanding RAID10. RAID01 is pretty easy: stripe, then mirror. Somehow the 'mirror, then stripe' explanation seems a bit vague to me. Could someone point me in the right direction? On another note: Sometime in the near future (within 2 months), my SysAdmin and I are going to implement mirroring (as a minimum) on our production system. His idea was to set up the mirrors using the current rawspace configuration, restoring an Informix backup, thus populating the mirror. I was thinking about planning a system reorg in order to take the opportunity of correcting some table / disk placement issues. (Some tables have grown out of their 2G rawspace and the only place where I could put then was on another disk.) We'll be running under some time constraints. Would the benefit of some table placement changes outweigh the time involved to export, then import on a mirrored system? Sorry for all the questions . . . this mirroring stuff is a bit new to me (and a bit intimidating). -- John Carlson Informix DBA WHSmith USA #include std_disclaimer.h /* These are my opinions, not my company's opinion */

"Carlson@WHSmith" wrote: > > I've been following the RAID discussions and I'm having a bit of 'DBA > block' when it comes to understanding RAID10. RAID01 is pretty easy: > stripe, then mirror. Somehow the 'mirror, then stripe' explanation > seems a bit vague to me. Could someone point me in the right direction? You SEEM to get it, what's the problem? RAID10 is: build N mirrored pairs then create an N-way stripe set out of the mirrored pairs. The advantage of RAID01 is that if a single drive fails only a single drive needs to be recovered, that one mirrored pair. If a drive fails on a RAID01 stripe the entire failed stripe must be recovered from the other and until it is you are at risk of catastrophic data loss if ANY drive on the remaining stripe should fail before recovery can complete. With RAID10 you are only at risk for catastrophic data loss if the specific defunct drive's mirror drive fails. You can minimize the risk of this by making sure no drive's mirror drive is the same manufacturing lot as it is so that if there is a manufacturing defect affecting the lot the mirror is safe. (Don't scoff, lost over 100 drives from one lot during the course of a 2 week period because of such a lot wide defect. We were lucky as we had over 250 of those drives online. Seventy of the failures occurred after we discovered the defect and before all 250 drives could be replaced.) > On another note: > > Sometime in the near future (within 2 months), my SysAdmin and I are > going to implement mirroring (as a minimum) on our production system. > His idea was to set up the mirrors using the current rawspace > configuration, restoring an Informix backup, thus populating the > mirror. I was thinking about planning a system reorg in order to take > the opportunity of correcting some table / disk placement issues. (Some > tables have grown out of their 2G rawspace and the only place where I > could put then was on another disk.) We'll be running under some time > constraints. Would the benefit of some table placement changes outweigh > the time involved to export, then import on a mirrored system? I'd say better to get the system back online, add a few extra dbspaces and then reorg the system as time permits to improve performance and load balancing. BTW RAID10 over RAID1 will give you some of that load balancing without the work. Art S. Kagel