I don't think I've ever experienced five 9's on any telco
service, I have always assumed I must be the one customer
experiencing down-time, and the aggregate was somehow five
9's. How is network reliability calculated to end up with
five 9's?
-do not include any outage less than 20 minutes.
-only include down lines that are actually reported by customers.
-when possible fix the line and report 'no trouble found'.
-remember that your company is penalized by the FCC for bad ratings, so
don't report any problems that you do not have to.
-do not include any outage less than 20 minutes.
-only include down lines that are actually reported by customers.
-when possible fix the line and report 'no trouble found'.
-remember that your company is penalized by the FCC for bad ratings, so
don't report any problems that you do not have to.
You forgot my favorite :
Every trouble report from a customer must include at least 2 hours on hold before a ticket is opened.
Don't forget to schedule maintenance as often as possible. After all, if a
customer doesn't get any service because there is a maintenance they don't
mind, but if the same service disruption is because of an outage they get
upset.
This is the sort of thing that can be discussed forever, but here's an anecdote anyway:
At my previous employer, we hired a lot of people who had spent their entire careers either running or developing equipment for TDM voice networks. Their view of five nines for voice was that the network was "up" if the voice signaling infrastructure worked as designed -- not that you could actually get calls through the network. So, for example, if your long distance call could not be completed because bearer trunks were down, there wasn't enough capacity etc. etc. then the voice network was still "up" for five 9s calculation purposes even though you couldn't use it for its intended purpose.
How many times have you received the "All circuits are busy" message? Some would say that was the voice network failing to function -- the Bell-shaped heads said it was working as designed. They were clear that what mattered was the signaling integrity of the network, not the application of voice connectivity itself. So, if you can get dial tone but not place a call, that's five 9s reliability at work.
When applied randomly to the Internet, I suppose that means if you can dial into a RAS and establish a PPP/IPCP session, but the RAS' connection to the Internet is down, then the service is up
Even disregarding the issue of whether 99.999% network reliability is
possible, people have made it abundantly clear that they don't want
it.
In this case I define "to want" as "to be willing to pay even a little
bit extra for".
This is even the case in POTS telephony. I know lots of people who
are happy to use 2 cpm long distance from Priceline et al, even though
half the time the call doesn't go through.
These can all accomplished with one simple and elegant system policy rule.
It also has the advantage of "tuneability". If you hold off determining if
there is an interruption of service for X minutes, then NO interruptions of
service shorter than X minutes exist, since the service is functional when
tested. Also, a call center that first "routes" tickets to the "appropriate
area" can deliver 99.99 with little effort. ( in fact, the LESS
effort/clue, the BETTER the rating ! )
BTW - One of my best friends growing up ( and we took EE together )
grandfather was the V.P. in charge of AT&T's LongLines division while a lot
of the "wire was pulled". From what he said of his grandfather's remarks,
they didn't think about five 9's. The question was how much spare/redundant
capacity did you have, both for dependability and to support the countries
growth. Not exactly "this quarter's profit" thinking -sigh-.
Bruce Williams
"Two is not equal to three, even for large values of two"
5 nines is a myth, conjured up by sales cretins to have something to sell...If I remember, 5 nines translates to 6 minutes outage a YEAR..?
(Correct me if I'm wrong here)
It's a marketing ploy for liar sales people and CEO's, it has absolutely NOTHING to do with real-world conditions.
[stuff missing]
When applied randomly to the Internet, I suppose that means if you can dial
into a RAS and establish a PPP/IPCP session, but the RAS' connection to the
Internet is down, then the service is up
[stuff missing]
I seem to remember a large internet provider's service contract reading
something to the effect of. "Your server is considered down if customer
router cannot pass packets [or ping] with service provider's immediate
upstream router." This is a functional description of the above for
dedicated lines as customer aggregation routers never talked to the
internet, so if there was a problem at a transit router you weren't getting
anywhere.
A modern contract I saw recently defined "up" for colocation purposes as
"the customer's assigned gigabit port is available." Though available was
not a defined term, one could not easily apply that to a ports' willingness
to pass packets. One could say a congested port was not available though I
guess.