Methods for mapping scalable video over differentiated services networks

Differentiated Services (DiffServ), which are Quality-of-Service (QoS) based networks for the Internet, provide diverse classes of reliable delivery of Internet packets at a paid price. In this thesis, we propose methods for mapping scalable video streams over the Assured Forwarding (AF) class of DiffServ. A key objective of the proposed methods is to improve the overall video quality as measured by the packet loss ratio (PLR) of the video layers in a scalable video stream. Furthermore, a ratio of the weighted throughput of video packets to the associated cost is employed as a metric to evaluate the performance of the proposed mapping methods. Two high-level mapping strategies are described: a one-to-one strategy for mapping each video layer to a distinct AF priority level, and a multiple-to-one strategy for mapping multiple video layers to one AF priority level. In order to achieve the desired weighted throughput of video packets in the later strategy, we propose a linear pricing model (LPM) and an exponential pricing model (EPM) for network bandwidth commitments. The aforementioned mapping strategies were simulated using the Network Simulator (NS). These strategies were also analyzed for various network overload conditions, which result as a consequence of bandwidth over-commitment. Some of the key conclusions of this work include: (1) the EPM provides the most desirable (weighted) PLR performance under zero over-commitment conditions; (2) the LPM provides a desirable overall performance under a variety of conditions.