[geeks] Re: Speaking of Bad practice in the Win/X86 world

[Gavin Hubbard]

>> One of my clients was concerned that year-end processing was
>> coming up soon and the VAX had been up for over 500 days.  The
>> client wanted the system rebooted.  Never mind that there weren't
>> any problems.
>> -- 
>> Eric Dittman
>> dittman at dittman.net
> 
>	This does actually have a sick basis in probability math. Let me see
>if I can explain this the way they are seeing it. If the chance of a server
>failing on a given day is X, and the chance of it failing the next day is X
>* 2, so at 500 days it is x * 500 and growing towards certainty. Now if we
>look at the case of, if it has worked this far once before it will work
>again this far, and define that as Y with and set it to the maximum value
>the last value of days that was attained. you then subtract that from the
>day value. you end up with (X * (days - Y)) which give you a negative
>failure chance until after you pass the maximum time you have ever ran. 
>	While this is not actually true, people who have not played with
>systems a great deal and don't know thing like MTBF and the fact that
>electronic parts do wear out will think it is. This is not even a stupidity
>thing, it is just basic human nature to assume that if it has worked once it
>will work again. Figuring it has been up for 500 days, they are worried it
>will fail on the 501st, but if they reboot they get 500 days of error free
>operation before they have to worry again. 
>	It may not be right, but I can sympathize on why this is there
>reaction.

With the exception of memory leaks and mechanical degradation, I would have
said the the chance of failure in an ideal system was a statistically
independent probability. In other words the chance of failure after an
arbitrary number of days would still be X.

Regards,

Gavin