| In recent years, with the sustained development of broadband Internet infrastructure, the Internet based applications have gained rapid growth and enrichment. Among them, the video-on-demand (VoD) service, with the features of rich content and convenient to access has got rapid development and attracted the majority of users, becoming a "killer" application over Internet. However, due to the huge number of users, vast amount of multimedia data and users' ongoing pursuit of the improvement of quality of service, the construction of high efficient and low-cost distribution system to meet the needs of users becomes one of the most urgent requirements in development of VoD service.Recent years, the rapid development of Peer-to-Peer (P2P) technology and its huge success in file distribution and live streaming applications brings a new opportunity for the distribution of VoD services. To study how to build low-cost, highly efficient and scalable VoD system with the consideration of flexibility, scalability and other advantages of P2P technology has great theoretical and practical significance. However, in VoD system, asynchronous user behavior, the critical data arrival deadline, peer dynamics and their interactive operations bring a big challenge for the study above problem.In view of this, we focus on the study of video data distribution strategy in this paper, aiming to improve the data distribution efficiency and quality of service with the cooperation of users. Our work starts from the point of service and expends on three aspects:when a user requests some specific video clips, how to help him find the service provider quickly and accurately; how to improve the service capacity provided by users who are watching the same video so that sharing efficiency between them can be improved; how to reduce the requirement of media server bandwidth by the aids of idle peers via effective control and management. Main contents of this paper are listed as follows:In P2P VoD system, due to asynchronous user behavior, the critical deadline of data arrival required by continues video playback, peer dynamics and their interactive operations; the traditional structured and unstructured service location strategy cannot meet the demands. DHT, a representative structured approach cannot cope with high peer dynamics and the frequently data cache update, while the unstructured strategy is unable to meet the fast service location requirement when users take interactive operations. In chapterâ…¢, a semi-structured index structure is designed to help users improve the efficiency of service location. The backbone of this index structure is composed of relative stable peers that are adaptively selected from the system, while other peers will connect to the backbone to take part in the construction of the index structure and share the location service. Meanwhile, based on the consideration of efficiency, accuracy and robustness of service location, a simple but efficient index management strategy has been proposed for construction and maintenance of index relations. Simulation results reveal that due to the collaboration between the relatively stable peers and the efficiency of index organization, the proposed index structure can efficiently help users for quick service location.The design purpose behind P2P technology is to fully enhance the service capacity that provided by users, so that their requests can be efficiently served via mutual sharing. However, In VoD system, the asynchronous user behavior, the demand required by continuous playback and the requirement for supporting interactive operations make the users'demand (or interest) on video clips are only depend on their current playback position which are scattered at a time. Such features make the scheduling policies which can efficiently work in P2P based file sharing system or live streaming system becomes inadequate, because users'interest are relatively concentrated in such systems. Therefore, in chapter IV, we first give an in-depth analysis of the relationship between service demand and supply in VoD system. We then propose and analyze the problem of how to improve the service capacity with minimum cost. After that, we present an exact algorithm and an approximation algorithm which can work in polynomial time. Taking system deployment issues into account, we design a service capacity optimization oriented data scheduling scheme based on above approximate algorithm, which will satisfy users' playback requirement in prior, then try to optimize the peers' service capacity with downloading a small portion of video clips. Simulation results reveal that compared with the strategy that widely deployed in existing on-demand system, the proposed strategy can achieve more efficient use of users'upload bandwidth.When users'mutual sharing cannot satisfy their demands due to access network restrictions or affected by other network applications, system scalability will be seriously affected. In chapter V, we study the system scalability improvement problem under these conditions. In VoD system, the common user behavior is that, when a user reaches system, it chooses to watch a movie and exit the system when completed. If a reasonable mechanism is designed to incentive users for longer stay after they finished watching, or join the system even they do not watch any video, then we can take advantage of these free peers to enhance the service capacity with their upload bandwidth:with reasonable scheduling strategy to allow these idle peers to download a small portion of video clips, then these peers can serve the requests with their upload bandwidth. With this in mind, we first present a simple model of the VoD system with idle peers and study the relationship between system parameters. Then, taking the instability of idle peers and their resource scarcity issue into account, a control policy is designed for management of the idle peers in the system and plus with an efficient data scheduling strategy, aming to improve the system scalability.This work is supported by National Natural Science Foundation of China, under grant of "Interactive streaming media distribution over wireless mesh network" (60773193) and 863 project "Key Technologies of Knowledge Element Mining" (2008AA01Z127-1).
|
PDF Full Text Request