[Soekris] Mean Time Between Failures

[Robert Woodcock]

Chuck Yerkes wrote:
> Well, actually, the emPHAsis is on MEAN.
> And in stats, Mr Foster taught me that the MEAN is entirely meaningless
> without the Standard Deviation.  (the mean of 59 and 61 is 60.  The mean
> of 20 and 100 is 60.  The mean age of a group of 2 is 60.  In what
> age group are they likely to be?)

MTBF is not a life expectancy at all. What you want is Rated Life. MTBF
assumes that if you take an unlimited number of drives, run one until the
end of its rated life, replace it with the next one, run it until the end
of its rated life, etc., you can expect one failure every $MTBF hours.

Since rated life is never specified anywhere, it's safe to treat MTBF
statistics as meaningless marketing fluff.

If I was a lawyer suing a HDD manufacturer for false MTBF statements, I
would attempt to argue that 'rated life' was at least equal to the
warranty period.

5 years ago, I would have pegged rated life at about 5 to 6 years. It
clearly isn't that now.