It’s strange to see gundreds of nerds trying to calculate probabilities to try to prove incorrect interpretations caused by this somewhat unclear article. Endless discussion and not a single person who understands what they’re talking about, and no clear conclusion or consensus. Geeze. It’s actually very simple and I will explain it to you all in simple terms without any need for calculating a single probability.

It is THIS simple:

* First, you must understand: The maximum possible storage space in both RAID 10 and RAID 01 is 50% of the total size of all disks. * In RAID 10, this comes automatically, as you set up the RAID 0 on top of any number of individual two-disk RAID 1 (mirror) pairs. The mirroring in each pair is what halves the total available storage space since half the disks are used for redundancy. * In RAID 01, you set up a RAID 1 (mirror) at the TOP LEVEL instead, and since mirroring halves the available storage space, this means that to get the maximum possible storage space (50% of the total disk capacity), you must create only TWO RAID 0s, by putting half of the disks in each RAID 0, and then mirroring across those RAID 0s, thus giving you the optimal 50% capacity. * To put this into numbers: For a 24 hard disk example, the maximum, most efficient storage space layout would be: RAID 10 = A single RAID 0 stripe consisting of 12 x two-disk RAID 1 mirrors. RAID 01 = A single RAID 1 mirror consisting of 2 x TWELVE-disk RAID 0 stripes. * Do you see the problem yet? * If a single disk dies in RAID 10, the other disk in that particular two-disk mirror is still chugging along, and ALL other mirror pairs in the whole array are perfectly fine. * If a second disk dies in RAID 10, you are still perfectly fine, as long as the failure is in a disk in another one of the two-disk mirror pairs. You can keep losing loads of disks and as long as you never lose both disks in the same mirror-pair, you lose NO data. * If a single disk dies in RAID 01, ALL TWELVE DISKS IN THAT TWELVE-DISK RAID 0 STRIPE GO OFFLINE, leaving only a SINGLE REMAINING RAID-0 twelve-disk stripe alive. * If a second disk dies in RAID 01, you are done. All data is lost forever. Goodbye RAID. * It gets even worse for RAID 01: Let’s say you ONLY lost a single disk and the other RAID 0 is still running fine. Well, all data on your offline RAID 0 is now out of sync as further writes have taken place on the online RAID 0. So when you insert a new disk and being the second RAID 0 back online, it must now erase ALL TWELVE HARD DISKS in the briefly-offline RAID 0 array that broke, and instead fill (re-mirror) them with the latest contents from the disks in the other RAID 0 stripe. This takes an EXTREMELY long time (copying from 12 disks to 12 disks) and puts extreme stress on the remaining healthy RAID 0 and probably kills it too before the re-mirroring can complete. * In RAID 10, replacing a single disk is easy and safe regardless of how many hard disks are in your array: It just needs to re-mirror a single replacement disk from its mirror twin. This is fast (just a 1-to-1 copy) and SIGNIFICANTLY less likely to fail than the 12-to-12 copy in the RAID 01 example. * These differences are true even in simple consumer-level 4-disk arrays: If you lose a single disk in such a RAID 10, you only need to re-mirror as a 1-to-1. If you lose a single disk in such a RAID 01, you need to re-mirror as a 2-to-2 to restore the de-synced mirror. And if you lose 2 disks in a 4-disk RAID 10, you can get lucky and only lose one from each mirror pair, which means all data is still safe. But if you lose a second disk in a RAID 01, you kill the only remaining RAID 0 and thus lose all data. * In short: RAID 01 is only used by clueless people who have no idea how extremely dangerous it is. It doesn’t matter how many hard disks you have in RAID 01; lose two disks and all data is lost, BAM! In RAID 10 you can lose up to half of the disks and as long as none of the two-disk mirrors are totally destroyed you will never lose data. And as mentioned, repair is never more than a simple 1-to-1 copy. * There is no performance reason to choose RAID 01 either. Performance between RAID 10 and RAID 01 is identical: Both have NX read and NX/2 write performance, where N is the number of disks and X is the individual disk read/write speeds.

There you go. Feel free to delete the hundreds of useless comments before this one. 😉