Introduction

The modeling and analysis of computer network traffic has been an area of extensive research over the last ten years, as new multimedia applications over the Internet become common.

In order to conduct a performance study, several steps are needed [#!tools2000-tmod!#]. First, one must understand the characteristics of the traffic competing for the resources under investigation. We can: (a) analyze measurements taken from the actual aggregated traffic of an existing network; (b) and/or focus on traces of traffic, for instance a trace from a given application such as voice and video transmission; (c) or concentrate on different ``classes'' from a given application. (For instance, for video transmission application, the classes of action movies, or the class of lectures, etc.)

A large number of models have been proposed in the literature. They include Markovian models, and models that possess long-range dependence. The user should be able to use these traffic models as input to a model that includes the system resources in order to conduct performance studies.

The TANGRAM-II environment provides a set of tools to , obtain descriptors, and experiment with different models. The modeler is able to: (a) use statistics from real traces; (b) choose from different (which includes Markovian, FBM and FARIMA models); (c) calculate descriptors from the models to be able to match parameters and/or verify statistical differences from the model to the measured data; (d) create a ``complete'' performance model which includes the traffic model and the resources under study; (e) solve it via simulation or analysis and; (f) conduct experiments with traffic generators over a laboratory environment.

The TANGRAM-II modeling environment includes traffic modeling tools and a , that are integrated with the simulation and the analytical modeling tools.