Originally Posted by
peterb
Assuming you have two identical drives (different manaufactur batch!) and each drive has a MTBF of 500,000 hours. Assuming too that the initial failures (the start of the 'bath tub' curve) have been filtered out by the mfr burn-in process, then with one drive, you are likely to get a failure some time in the next 500,000 hours. Add two drives to an array (RAID 0) and the chances are you will have two failures in the next 500,000 hours, or one failure in 250,000 hours. In other words you halved the MTBF and doubled the risk of losing all your data. Now it might be the reduction of risk from a small number to a another small number - but it is still a factor that you need to consider. Certainly for RAID 0 I would want to use enterprise grade drives to ensure the MTBF is as high as possible.