Greg Miller wrote:

> I'm not sure that overall MTBF is all that useful. It might be better 
> to say that x% of users should have at least n hours MTBF.
>
> If we have an average MTBF of 1000 hours, that may simply mean that 
> it's 2000 hours for half of the users and 5 minutes for the rest.

Yes, makes sense. It would line up with the experiences of me and 
chofmann - he claims to have had 1 crash with 0.9.1, while I had dozens, 
maybe even > 100.

As for concrete values, I'd suggest:
90% users: MTBF > 250 h
99% users: MTBF > 50 h
MTBF average: 500 h


Reply via email to