RAID 5, we hardly knew ye.
by Brian Hershey on Oct.21, 2008, under General Info
We all knew this day would come, when RAID 5 will cease to be. OK, maybe not all, but I had a sneaking suspicion that super-large hard drives would be the end of RAID 5.
Check out this article about it.
Oh, and start hinting around about a budget for a fiber-channel SAN, while you’re at it.
November 4th, 2008 on 12:31 pm
Doesn’t a SAN just add another layer of complexity/hardware that that could fail? Yes, it would add some redundancy, but at what cost? The backend of the SAN will probably be some sort of RAID array that would have these same issues.
I don’t think this is just a RAID 5 issue. Wouldn’t RAID 1 be in a simular situation?
November 4th, 2008 on 12:55 pm
I think the real problem is that the algorithms used to regenerate lost data from a RAID 5 parity block are based on Pythagorean theorem. (if you know the values for any two variables, you can always figure out the value for the third in a+b=c) RAID 1 is just a bit-for-bit mirror, and so doesn’t make use of parity. If you have a really large SAN, and use RAID 1 for all of your LUNS, a failed disk won’t matter. (as long as the entire mirror doesn’t fail all at once.)
November 12th, 2008 on 10:17 pm
Did you just explain the Pythagorean theorem in one and a half lines there? Awesome.
I think back just 10 years and remember the cost of a single gigabyte was very high. Today, prices are so affordable, it hardly matters which configuration you want to do. Just as long as it is reliable.