Statistical rate control for efficient admission control of MPEG-2 VBR video sources

Variable bit rate video (VBR) allows statistical multiplexing in asynchronous transfer mode networks (ATM) while offering constant visual quality. A key issue for achieving efficient video multiplexing and therefore efficient admission control is the selection of a VBR video characterization specifying accurately the rate distribution of video sources. We show that the gamma distribution is a good tool for modeling the aggregate traffic of homogeneous as well as heterogeneous VBR MPEG sources. Therefore, we propose to characterize the VBR video traffic by the use of simple descriptors based on the gamma modeling of the sources: the mean rate and variance. We show that the proposed descriptors allow for the implementation of an efficient connection admission control (CAC) function that agrees well with simulation results. We then show that by making use of rate shaping algorithms (variance control) for the enforcement of smaller variances an even more efficient CAC can be designed.