Fluid-based models of communication network traffic
consider traffic behavior in terms of packet rates, rather
than packet instances. While fluid modeling is com-monly
used in mathematical approaches, these often re-quire
model simplification. Simulation of fluid models
may be used in order to retain necessary model detail;
it is beginning to receive some attention in the literature
[1, 3, 4, 6, 5]. We provide here a cost-benefit study of
such techniques, within a common framework.
We describe a simulation tool that targets simulation
of communication systems, and its use to simulate fluid-based
models. Within the same tool we are able to develop
and compare packet-based, and fluid-based models, assess
any performance benefits and assess inaccuracies intro-duced.
The specific study made here is of networks where
traffic is generated as a Modulated Markov Process and
is shaped using a Leaky Bucket flow-control mechanism.
Switches model FCFS service among all competing flows
at output ports. The principle contributions of this paper
are to demonstrate fluid simulation in the context of an
existing packet-oriented simulator, and to study the per-formance/
accuracy trade-offs between packet-based and
fluid-based models.
|