The Demise or RAID5

[lukehealey]

I just found this posted on Slashdot:

http://blogs.zdnet.com/storage/?p=162

makes some interesting reading -

[hddguy]

More work for me then lol

[HDD_MASTER]

It does make an interesting read but I am sure as all well trained and experienced recovery dudes we will find a way around these URE's.

I have two thoughts already!

[drc]

I think this article indicates a flawed understanding of statistics.

Basically he's saying that because a given drive has a posted sector failure tolerance of 1 per 10000000000000 sectors, if your RAID5 has enough disks to be more than 10000000000000 sectors big then OH NOES WHEN YOU TRY TO REBUILD THE LAW OF MATH GUARANTEES A FAILURE!!!1

If this was true then you would see larger HDs being more failure prone by sheer virtue of being larger, which I don't think is the case.

[cgallery]

If this was true then you would see larger HDs being more failure prone by sheer virtue of being larger, which I don't think is the case.

I respectfully disagree.

[lukehealey]

If this was true then you would see larger HDs being more failure prone by sheer virtue of being larger, which I don't think is the case.

It's not referring to general failure -just URE's during the rebuild of an array. with most RAID5 controllers, a URE during rebuild will cause the rebuild to fail - thus rendering a RAID5 set broken...

There is always the risk of a general failure of a disk during a RAID rebuild, - something else that RAID5 does not cope with well (at all??!!)

Statistically - the bigger the volume that is being rebuilt, the greater the chance of a URE (or disk fail) causing a rebuild fail. simple.
I

[drc]

Statistically - the bigger the volume that is being rebuilt, the greater the chance of a URE (or disk fail) causing a rebuild fail. simple.

So the more disks you have in your RAID the greater likelihood of it being damaged because more than one fails at once?

Wow, that's not really news at all.