On Thu, May 28, 2009 at 9:21 PM, John Hornbuckle <[email protected]> wrote: > Just make sure it's not RAID 5: > http://blogs.zdnet.com/storage/?p=162
His point seems to be that if you have a single disk failure, the system needs to read all the other disks to reconstruct the failed disk. But at the same time, you're likely to encounter a previously-unknown bad block on another disk, so you won't be able to completely reconstruct the failed disk. This really isn't a new concept. It may be more likely to happen with bigger disks; I dunno. A RAID system which does patrol reads -- regularly reading/comparing all blocks on all disks, to make sure everything is still working -- should compensate for this. As soon as that block goes bad, that disk will be failed and you'll know to replace it. Before the next failure develops. Unrelated to the author's concern: RAID-5 can be a performance issue for any database, including Exchange. Especially with an under-performing RAID controller, or when the array is degraded due to a failed disk. Still, I've successfully run Exchange (and several other things) on a single RAID-5 array in a small business (<50 mailboxes) before without any real trouble. But it will become a limiting factor at some scale (100 users, 500, 1000, I dunno). RAID 10 (a stripped array of mirror sets) was the traditional solution. Since the popular assumption seems to be that gigantic SATA disks are the program to expect, it may be a mirror of two 1TB+ disks would give you all the storage you need. (Hah! How many times have we each said *that* before? :) ) -- Ben ~ Finally, powerful endpoint security that ISN'T a resource hog! ~ ~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~
