| Peer-to-Peer Video-on-Demand (P2P-VoD) is a popular Internet application which aims to provide a high quality video service to users. The advantage of using the P2P technology is that the system can utilize peers' resources so as to satisfy other peers' viewing requirement, thereby improving the system scalability and reducing the operating cost. There are two key design issues in P2P-VoDs. First, given the distributed resources of peers, what is the most efficient manner to utilize them in a distributed and dynamic fashion. Second, given the selfish nature of peers, how to incentivize the peers to contribute their local resources.;A P2P-VoD system is highly dynamic, asynchronous and heterogenous in nature. In addition, it requires a much higher bandwidth resource as compared with file sharing applications. These features make it challenging to solve the above technical problems, and hence motivate our work. In particular, we aim to answer:;• How to determine the optimal ratios of storage space that should be assigned to each video, such that the content server's workload can be minimized?;• How to design effective and efficient incentive mechanisms so as to stimulate the peers to contribute their local resources?;We first focus on the optimal replication strategy. In particular, we answer (a) what is the optimal replication ratio of a video in terms of its popularity, and (b) how to achieve these optimal ratios in a distributed and dynamic fashion. We formulate the video replication as an optimization problem, and show that the conventional wisdom of using the proportional replication strategy is "sub-optimal". The optimal replication ratios should be proportional to the "deficit bandwidth" which we define in the thesis. We utilize "passive replacement policy" and "active push policy" to achieve the optimal replication ratios and show how to greatly reduce server's workload and improve streaming quality via our distributed algorithms.;We next focus on incentive mechanisms design. The content providers need to incentivize the peers to contribute their upload capacity to delivery data, and local storage space to cache the videos. We decompose the problems and design reward-based incentive mechanisms for them respectively. (a) Incentivizing upload capacity. Each peer is rewarded based on its dedicated upload bandwidth. We analyze the interaction between a content provider and the peers using game theory. We derive a unique equilibrium, analyze the system efficiency and study the long term interactions under a repeated game setting. (b) Incentivizing distributed caching. Each peer is rewarded based on the videos they cache. We characterizes the optimal reward price using optimization. In particular, we first derive the optimal prices to obtain the desired amount of replicas in an asymptotic system, and then extend our results to adapt to various system environments.;To summarize, this thesis addresses the resource acquisition and allocation problems in P2P-VoD systems via mathematical modeling, game analysis, algorithms design and performance evaluations. |
