TRAFFIC CONTROL IN ATM NETWORKS

This paper presents a framework to study the problem of traffic control in ATM networks. Bursty traffic sources are modelled by Interrupted Bernoulli Processes. Probability of violation and dimensioning issues are addressed for the leaky bucket mechanism and the (L, M, T) mechanism, proposed by the authors. We present two schemes to demonstrate that improved statistical multiplexing can be achieved if the traffic control mechanisms are not used in isolation for each session, rather they are used on groups of sessions. In the first scheme, traffic control is performed on the aggregate traffic of a group. In the second scheme, traffic control is performed on single sessions with information about the status of all the sessions of the group being available and used at each traffic control mechanism. Numerical examples that illustrate the improvement in statistical multiplexing as a result of the above schemes are presented.