Optimal rate control over DQDB networks with applications to voice packet transmission

Based on the statistical multiplexer concept, this paper introduces a new mechanism for regulating the rate at which multiple burst sources, such as compressed voice and/or video sources, are allowed to transmit along a DQDB network in order to provide graceful degradation of source signal quality under overload conditions. A queueing model is established first whereby the probability distribution of the length of the system-wide multiplexer is determined by solving a set of differential equations following the well known fluid flow approximation methodology. Optimality is defined with respect to the average distortion that the system experiences across admitted voice sources as well as along time subject to the constraint that multiplexer queue length remains within limits specified a priori. The underlying objective function is optimised via dynamic programming, by invoking additionally well-known approaches from the rate distortion theory