[5158] in North American Network Operators' Group
Re: Fractal models of Big-I Internet
daemon@ATHENA.MIT.EDU (Vadim Antonov)
Wed Oct 9 22:41:25 1996
Date: Wed, 9 Oct 1996 19:31:38 -0700
From: Vadim Antonov <avg@quake.net>
To: michael@memra.com, nanog@merit.edu
>Given that real world systems that exhibit fractal behaviors can now often
>be modelled mathematically, do you know of anyone who has attempted to
>apply such fractal models to Internet traffic?
Self-similar traffic patterns are hard to work with (and were discovered
only recently). For one, inter-arrival times in self-similar stream
have infinite variance. That sure screws up a lot of math generally
used in queueing theory. It is also hard to generate (there are no
known efficient methods of producing self-similar streams) and stochastical
models using it are not reliable (infinite variance again).
>When designing protocols
>do researchers take this fractal nature into account?
Researchers do not design things. They smash things and watch what
resulting particles do :)
On a more serious note, at least there are some indications of what
will NOT work with fractal traffic. Packet shredding for example.
>I suppose the second question would be somewhat moot if there is not yet
>an accepted fractal model to math the Internet...
It is still a research topic, to a large extent. There are results to
the effect that even connection arrival processes are self-similar and
not Poisson as was previously thought, which may mean that connection
oriented schemes like ATM will go bust in large-scale networks.
--vadim
