99.9% uptime

[Hall, Ken (ECSS)]

Sorry, forgot to change the subject on my first posting.

I was reading a comment this morning about something Microsoft had published
to the effect that there were vendors guaranteeing 99.9% uptime for NT.  The
guy who wrote the reply did the math for what that means, and the results
are very interesting.

Quote below:

	OK, now what does a 99.9% uptime guarantee mean? Well, it means that
at bottom,
	 a guarantee that the machine will not be down for more than one
one-thousandth
	 of the time. If we assume that we have a stable system, ie., one
where the
	 system is not taken down for software upgrades, etc., then we are
looking at
	 the time from a system crash (strictly, I think, from the time that
the crash
	 is actually noticed) to the time that the system is running again.
So, below is
	 a little table that shows the best guaranteed up-times for various
values of
	 the above restart time, rounded up.

	 Restart time   Uptime
	 10 mins        7   days
	 30 mins        21 days
	 120 mins      84 days
	 etc   .....

	 The conclusion I draw is that either (a) NT crashes very often and
is quick
	 to restart or (b) NT crashes less often but takes a long time to
restart.. 

Considering I've had Netware servers and UNIX boxes that stayed up for 6
months at a time, maybe we need to re-think what constitutes acceptable
availability. "99.9%" sounds great, but if you look at it this way, it's
really pretty abysmal.

Anyone care to do the math for their mainframe systems for comparison?