RAID failures

RAID 5 and Uncorrectable Read Errors

As the size of disks in a RAID 5 array increases, the concern arises that a large array hasn't time to finish a rebuild before the second member disk fails. It is possible to calculate a probability of such an event if the disk reliability is given. The greater the array capacity and the less the disk reliability, the possibility of second disk failure is greater. Disk reliability is provided in a vendor specification where its value is usually between one error per 10¹⁴ and one error per 10¹⁶ bits read.

Although such an approach of calculation arouses reasonable criticism, we have still created the calculator which works using this formula:

P=(1-10^-14)^{8*109*(N-1)*S}

where N is the number of disks in RAID, and S is the size of a single member disk, in GB