Network Models: Do They Matter?
by Matt Grossglauser

Abstract:
In my presentation, I'd like to share a few thoughts and concerns
about the role of modeling in networking. These are the result of some
animated recent discussions with colleagues from inside and outside
our field, and I thought it would be interesting to share these with
the workshop participants.

The body of knowledge of engineering disciplines is codified in their
models. In signal processing, life is not possible even for the
practitioner without at least a rudimentary understanding of the
Fourier transform and of wavelets; verification has finite automata;
electronics has circuits and systems; economics has the rational actor
and frictionless markets, etc.

In networking, there is a scarcity of such models representing our
accepted core problems and the boundaries of our discipline. This is
perhaps because networking is arguably more a technology domain rather
than a field of engineering, in that it overlaps with many of these
established fields, and borrows their models.

I'd like to offer a few consequences of this perceived state of our
field for both research and teaching, and propose two areas where I
see future models emerge that are proper to our field. The first area
is the frontier of wireless, embedded, mobile networks, where
understanding and dealing with randomness is becoming more
important. The second area arises out of the increasing complexity of
networking in practice, and the need for binding together routing
policies, security associations, instrumentation, quality-of-service,
etc.