If you include leap years in the calculation, ie, 525,960 min/yr
then yes, availability of 99.999 comes out to 5 min of downtime.
The techniques the telcos follow to achieve 5 9s are amusing but here's
a few comments. If I had the level of availability of my home
phone on my frame/atm networks, I could probably get my handicap
down to around 2 or 3. As it is, I have alot of trouble breaking 80.
Another thing, some of the comments in this thread make me understand
why the availability of the voice network infrastructure is so much
superior to that of the data side of things. For one, alot of network
engineers that I know don't know the ins and outs of how to calculate
availability. This is troubling since it means that even if they
accidently design a highly available network, they are left guessing and
not predicting how available it will actually be. Every person who calls
a network *engineer* should be familiar with calculating 3 sorts
1) Device availability calculated by combining the availabilities
of all the significant components of that device
2) The availability of serial vs parallel systems/designs
3) End to End availability calculated by combining the availabilities
of groups of devices which function together to achieve a common purpose
Some networks do not merit such analysis but you can bet that the
911 emergency services network does and should have this and much
more detailed analysis.
Another rant if I may. An engineer who can't calculate
availability also can't calculate ROI. The cost of employing a certain
design is largely a factor of how available it's going to be. If
you build a $2mil network where hour long outages cost the business
$400k, maybe you should have built a $4mil network....especially
if you know that you can count on 20 hours of outages during normal
If you've read this far, you may be interested in a book tip
on High Availability Networks: Cisco Press: "High Availability
Network Fundamentals" by Chris Oggerino. Seems to cover enough
detail...to have an impact on awareness...