[MLB-WIRELESS] Connecting nodes/clusters and expanding the m esh via Internet tunnels.

[Paul van den Bergen]

Hi all,
don't stand...  

right.  some statistics that do tie in nicely with this little debate.

There is the Parito Law wich states that 80% of the cost is caused by 
20% of the users.  this is a fairly universally found to be true 
economic rule of thumb.  Ofcourse, repeated application of limits to 
your user population eventually results in a population size of zero, so 
mostly you just have to accept it.  you can taylkor your user plans to 
mitigate the overusage by the few, and infact doing so is a good thing 
because you get much fewer disgruntled users who suddenly find their 
normal (i.e. high) usage patterns results in a huge bill.... usually a 
lot less profitable than a steady income stream...

This ties in nicely with my second point about bell curves...  or 
Poissonian distributions.... It is well known that internet traffic 
usage, in addition to increasing in size continuously, is also chaotic 
in nature - i.e. inherently unpredictable.  this can be adequately (yeah 
right! try it some time!) modelled by Parito Distributions... which are 
heavy tailed...  does that make sense? No? OK. basically, baselines, 
bell curves and so forth are inherently bogus in the internet domain 
because the traffic always has a chaotic component => unpredictable. 
 You can never have enough buffer space to ensure non-congestion (I am 
talking about the whole internet, not just individual networks). one day 
the internet will break under it's own load.  it'll get better, but it 
is inevitable that it will break. Eventually.


BTW, Brendan?  since I started this thread and I don't mind if it goes 
off topic, please, people, keep talking about DSL technical stuff.  I 
like it and it is what I wanted to expose when I started the thread...

remember, this is a wireless networking group.  therefore any discussion 
about computers, hardware, netowrk technologies, the law as it relates 
to comms, the way people behave in this domain, etc. are all on topic 
AFAIC....

evilbunny wrote:

>Hello Doug,
>
>I stand corrected...
>
>5% is a generalisation used everywhere... typically you will get a
>small minority approx 10% or less (usually about 5%) that will indeed
>use more then the other 95%, likewise you will get 5% that will use
>next to nothing, typical bell curve...
>
>This is where Optus got the magical 5% either end and removed it from
>the equation... Of course once you start imposing limits on a system,
>the bell curve can get out of whack...
>

-- 
Dr Paul van den Bergen
SERC, RMIT University
paul at serc.rmit.edu.au
+613 9925 1624 phone
+613 9925 5699 fax
goofey: bulwynkl




To unsubscribe: send mail to majordomo at wireless.org.au
with "unsubscribe melbwireless" in the body of the message