The primary activity of a network is that of store and forward operations, which current technology can carry out in a sub-millisecond time. In order to understand the detailed behavior of the network, we believe it is necessary to observe its characteristics with a fine time scale, one of the order of milliseconds and not seconds or minutes. However, as there are phenomenon which occur over a long time scale, it is also necessary to observe the network for long enough a time to capture these as well.
In 1992, we developed a low overhead approach to extracting network dynamics, based upon a fine-grained probing tool called NetDyn. We have used NetDyn in conducting several studies to date, and continue to improve and use it extensively.
Traditional approach to modeling the networks has been to use stochastic models. While such models have served us well in the past, we believe that the new models have to directly reflect the deterministic interactions among the components. To that end, we have been using deterministic modeling techniques, augmented by stochastic models only when necessary, and have developed a series of models which have been used to study the detailed characteristics of the network and to improve our understanding of the network dynamics observed through NetDyn studies.
The Maryland Routing Simulator (MaRS) is a network simulator which provides facilities for creating simulation models of complex networks with tremendous ease. This model has been distributed and widely used by the research community.