But surely in a sense it is. Perhaps I haven't understood the concept, so put me right if I haven't. The MTBF figures appear to be saying something like, 'If I use a 100 drives under the same conditions and to the same extent for, say half a million hours in total, then one of them can be expected to fail.
Another manufacturer might claim, 'If you use 100 of our drives under the same conditions, then they should last in total, one million hours before you would expect a drive to fail'.
The second scenario, with a quoted MTBF of 1 million hours would imply that the individual drives must be more reliable than in the first scenario where the MTBF is only half a million hours. Is this not the case?
The point is that MTBF can't be used directly for approximating the reliability of a single drive.
It's a number derived from the testing of a group from a batch of drives with that particular manufacturer.
It
may be used for approximating the expected failure rate of a group of drives within their product life time.
As the IBM article said:
5. If I purchase 1000 drives with an MTBF of 1,000,000 hours, how many can I expect to fail over a five-year period?
[...]
In this example, because of statistical variation, there is approximately a 90 percent probability that the actual number of failures will be between 33 and 55.
If we use the same numbers for 100 drives, that gives us between 4 and 6 failures with approximately 90% probability. (Having 0.3 drives fail is the same as having 1 drive fail.)
In addition, it's worth mentioning that several drive manufacturers are on record stating that their "desktop" drives are
not constructed for 24/7 use(!), but rather 8/7 or something like that. I believe IBM was one of those manufacturers (the DeskStar, AKA "Death Star" series).
This may skew MTBF numbers if the same drives are used 24/7, and the risk of failure is likely not simply tripled (but I'll refrain from speculating whether it will be worse or better than triple the risk).
