[NTLUG:Discuss] MS FUD & innumeracy/inaccuracy

[Doug Shaw]

>Hewlett-Packard, IBM, and Unisys provide 99.9 percent system-level uptime 
>guarantees for Windows NT-based servers."
>
>The marketing boys haven't done their math.  99.9% uptime equates to 4
> minutes of downtime *per day*.  Who wants a system that is guaranteed to 
>crash once a day?

If we are to fault Microsoft for its inaccuracy, should we not attempt to
ensure that we are not ourselves being inaccurate?  People in glass houses
and all that.

First, the original calculations were faulty when used to determine the
conclusion of four minutes downtime per day.  In the following, values are
rounded where practical.

Annually:

 365.25 days per year
  24.00 hours per day
* 60.00 minutes per hour
-------
525960.00 minutes per year

525960 minutes per year * ( 99.9 / 100 ) = 525434.04 minutes/year uptime

 525960.00 minutes total per year
-525434.04 minutes uptime per year
----------
    525.96 minutes downtime per year
/   365.25 days per year
----------
      1.44 minutes downtime per day

Second, the statement that the customer is being "guaranteed a system that
will crash once a day" is drawn from no logical base.  I'm certain that the
guarantee simply provides a maximum quantity of downtime, not how that
quantity of downtime will or must be distributed.  If the server does crash
once per day and is back up in 1.2 minutes, they will still have exceeded
99.9% uptime and will not be in violation of the guarantee, although the
customer would probably abandon them long before that year was up!
Definitions as to how they are defining a system to be "up" are a different
issue.

Please double check my calculations above.  If I am shown to be in error
then I surely shall retract my statements.

Doug