Bandwidth estimation for multiplexed videos using multinomial model

Real-time variable bit rate (VBR) video is increasingly dominating the bandwidth use of broadband networks. Many videos can be simultaneously transmitted over a channel in a broadband network. Given the usual capacities of communication links in use for transmission, the number of videos simultaneously transmitted over a broadband communication link for commercial purposes is not too large (not in thousands). Modeling relatively fewer number (less than a hundred) of videos is difficult because the use of asymptotic models for multiplexed videos cannot be justified. Although the literature has plenty of acceptable models for bandwidth requirements of single video traffic as well as of asymptotic multiplexed traffic, models for intermediate number of videos, which are of practical relevance, have remained relatively unexplored. This paper presents multinomial models for bandwidth estimation of both homogeneous and heterogeneous multiplexed videos using a Markov-modulated Gamma-based single video traffic model. An appropriate combination of exact and approximation schemes is used to contain computational overhead. Apart from bandwidth estimation, the model also offers a fast and accurate multiplexing gain estimation that may be used in call admission control (CAC) algorithms. Experimental results indicate the models have a high degree of accuracy.