A new approach to modeling of ATM networks

Switching and multiplexing are the two most important concepts of ATM networks. Statistical multiplexing, which is based on asynchronous properties and fixed length cells, is a special characteristic of ATM networks and the key factor affecting the performance. In this paper, we propose a novel idea for the performance evaluation of ATM networks that we call "transformations from cell overlapping to interleaving in logical channels". A SINK concept is introduced into the switching node for statistical scheduling. Then, a deterministic and stochastic Petri net(DSPN) based model reflects the completely random and correlative multiplexing action. Finally, the DSPN model is solved using an embedded Markov chain based on regenerative theory. Complexity is also defined according to the analysis of the reachability graph. We argue that, it is necessary and possible to design a new scheduling algorithm in order to reduce the complexity of statistical multiplexing.