Performance analysis and admission control in multimedia communication networks

We address the problem of efficient resource allocation and quality of service (QoS) provision in modern high-speed, multimedia, communication networks. We quantify QoS by the probabilities of loss and excessive delay of an arbitrary packet, and introduce the model of a multiclass node (switch) which accommodates sessions that may belong to multiple service classes. We treat such a node as a stochastic system which we analyze and control. The results in Bertsimas et al. (1996) provide us with a large deviations asymptotic on the buffer overflow probability per service class. We use the same approach, which is based on ideas from large deviations, optimal control and optimization, to derive a similar asymptotic for the probability of excessive delay. We exploit these performance analysis results by devising a call admission control algorithm which restricts the number of users accessing the network to guarantee a desirable level of QoS. Finally, we discuss practical issues and establish that the proposed algorithm can be implemented in real-time. Our conclusions are reinforced by experiments with actual traffic patterns.