Statistical transmission delay guarantee for nonreal-time traffic multiplexed with real-time traffic

Abstract

When nonreal-time sources are multiplexed with real-time sources in an ATM broadband network, the probability distribution of the transmission delay time of nonreal-time traffic cells provides a statistical delay guarantee for this traffic, which is a significant performance parameter of the Quality-of-Service (QoS) of the broadband network. In this paper we develop a stochastic analysis based on a semi-Markov model in order to compute the transmission delay distribution of nonreal-time cells. The nonreal-time sources are characterized as on–off Markov modulated Poison processes, and the real-time sources, like compressed video, are continuous-time Markov chains with multiple states. A priority-based schedule determines a differentiated QoS for each traffic type, and we concentrate the analysis in the transmission queueing process of the nonreal-time cells. This transmission process is modelled with a regenerative semi-Markov process that has a matrix-geometric solution for the limiting distribution of the embedded Markov chain. From the limiting distribution of the continuous-time queueing process, we define a sequence of delay distribution matrices that converges. The limiting distribution matrix of this sequence corresponds to the delay distribution of the nonreal-time cells.