On the dynamic estimation of multicast group sizes

This paper concerns multicast applications that are interested in the evolution of their membership over time. It covers optimal on-line estimation algorithms for determining the membership of a multicast group. The paper briefly reviews the related work and our own contributions to the field. Using a probabilistic acknowledgement scheme – its period and probability can be set to achieve high quality estimation while simultaneously avoiding feedback implosion – and signal processing’s filtering techniques, we have derived MSE-optimal estimators under the assumptions of Poisson subscribers arrival and either exponentially or hyperexponentially distributed lifetime of receivers. Our estimators have been tested through trace-driven simulations using data from real multicast video sessions over which they exhibit very good performance.