Scalable Internet Video-on-Demand Systems

In this thesis on Scalable Internet Video-on-Demand Systems solutions are provided that allow scalable streaming in today's Internet. These solutions allow the optimization of existing Video-on-Demand applications, as well as the integration of new mechanisms in the video streaming area. With the increasing popularity of video and audio data in packet switched networks and the lack of quality of service support, mechanisms that increase the scalability of Video-on-Demand applications must be provided. The scalability introduced in this thesis is two-dimensional. The first dimension is characterized as system scalability which is realized by the integration of caches into the distribution infrastructure. The second dimension, the content scalability, allows the adaptation of the video stream to existing conditions on the network by applying scalable video. In the course of this thesis, it is demonstrated how these two dimensions of scalability can be combined to increase the overall scalability and performance of a Video-on-Demand application. First of all, the problems that arise from the combination of both scalability dimensions are identified. This problem identification results in a new architecture that is suitable to solve the problems identified. Since this new architecture is based on several new mechanisms, most of this thesis is dedicated to the development and investigation of these new mechanisms. In the initial phase, when first developing these mechanisms, an additional problem was identified. That is, the influence of quality variations in scalable video on the perceived quality of the viewer had not been investigated so far. The results of a subjective assessment, which was conducted in the scope of this thesis, are the first that provide substantiated information for the layer-encoded scalable video format. These results build the basis for the development of an objective quality metric, which is used to measure the influence of quality variation in scalable video. It is shown that this metric is better suited than an existing, well-known metric. In addition, this new metric is used to evaluate the newly developed mechanisms that are presented in this thesis. An important element of a distribution infrastructure for Video-on-Demand is the cache and, thus, it is investigated how new mechanisms (which make use of the functionality offered by the cache) can increase the scalability of the system and the quality of the delivered video stream. Simulations which are based on these new mechanisms demonstrate their applicability and show that the identified problems can be solved. These investigations on new mechanisms for Scalable Internet Video-on-Demand Systems are enhanced by the extension of transport and signaling protocols. The latter allows the extended communication between servers, caches, and clients in a distribution infrastructure that is necessary to optimally support the new mechanisms. Simulations and measurements based on our own streaming platform show that the combination of both increases the scalability of a Video-on-Demand system and the perceived quality for the viewer. Additionally, these results show that the Scalable Adaptive Streaming architecture is capable of supporting heterogeneous clients. Thus, Video-on-Demand applications making use of the new mechanisms presented in this thesis are not limited to a specific class of clients and are even capable of increasing the overall scalability of the system