-3

RAID 10 seems to me as a permutation RAID 01. Then how can it be more fault tolerant than RAID 01?

I see the description here but it explains by dividing disks into groups and one disk if fails in each group there is no data loss in raid 10 but there can be data loss in raid 01. But what do these groups physically mean? Aren't there basically just 6 disks in the example in the given link?

http://www.thegeekstuff.com/2011/10/raid10-vs-raid01/

I see this link http://www.tomshardware.com/forum/43220-32-raid-raid0-raid10-explained#. where the discussion ends concluding that in case of RAID 01 if any disk fails in one of the array/group consisting of stripes, all the disks become inaccessible by the controller. But for RAID 10 the controller can access the good disks of the array/group consisting of mirrored disks. Why is it so? When there is a read/write request why can't in case of RAID 01 it just access the good drives in array/group 1 and the drive which failed be accessed into the other group/array which is the mirror of group/array1?

Ruppesh Nalwaya
  • 1,409
  • 2
  • 14
  • 22
  • 4
    Who knew Wikipedia had a ``HUGE`` entry on RAID http://en.wikipedia.org/wiki/RAID – Noam Rathaus Dec 05 '13 at 18:24
  • Even better: http://en.wikipedia.org/wiki/Nested_RAID_levels – Simeon Visser Dec 05 '13 at 18:25
  • @nrathaus I see that my doubt is similar to this one http://www.tomshardware.com/forum/43220-32-raid-raid0-raid10-explained# Here the conclusion is that controller can't access any of the drive in the array/group if one drive fails in the group/array in case of RAID 01? Is it true? why can't it access other drives which are just fine in the array/group? – Ruppesh Nalwaya Dec 05 '13 at 19:16
  • I cannot help you with your comment, as I don't know that much about this flavor of RAID (i.e 10) – Noam Rathaus Dec 05 '13 at 19:20
  • @SimeonVisser Can you help? I am convinced that if there is limitation with the controller RAID 10 is more fault tolerant.But is it there any such limitation? – Ruppesh Nalwaya Dec 05 '13 at 19:26

1 Answers1

4

Please reference the wikipedia artice on 0+1 and 1+0 for images which can help with understanding this answer. http://en.wikipedia.org/wiki/Nested_RAID_levels

In RAID01, the RAID0 is on the lower layer and the mirroring RAID1 is on the upper layer. Should one disk fail on a RAID 01, you would be forced to replace both RAID 0 disks on that side of the parent RAID1. The RAID1 would then rebuild the two new disks. RAID0 is like JBOD in this sense, where if you lose one disk, the entire virtual sub-array has failed. RAID 01 would, therefore, only be practically able to lose one disk. Should a second disk fail on the other side of the array for some reason, even if it held the other half of the RAID0 data, the controller will fail the RAID0 portion on that side as well and you will have no disks left.

In RAID 10, the mirroring is on the lower layer rather than the upper layer. You would be able to lose one disk on one side of the RAID0 and safely replace just that disk. In fact, you would be able to lose up to two disks total (one from each side) and still have a functioning RAID array. For this reason, RAID10 is preferable to RAID01.

You will see that most large-scale RAID arrays will have striping as the outer layer (RAID100, RAID50, RAID60). This is for the same reason. You would keep more useful data if one disk failed in a RAID50 as opposed to a failure in a RAID05 (if such a thing were possible to create).

Andrej
  • 180
  • 1
  • 2
  • 10