Multi-parent selection via Naïve Bayes in P2P layered video-on-demand system

With the increasing network speed, online video-ondemand system (like YouTube and PPS) is becoming more and more popular. Thus, many researches are about VoD system. We notice that among all the systems proposed, there is more or less space to make improvements. For example, seldom systems break the rule that one peer only transfers data with one single parent in the streaming service. As for the parent selection, most papers focus on the optimization on network topology, such as building a subset based on either content relevance or geographical features. In this paper, we gather the history information of each peer and rate each peer with the statistics, in order to select most reliable peers to play as parent peers. After the parent peer is chosen, the playing peer downloads the necessary video data from all the parent peers simultaneously to speed up the caching process. In this way, not only the server load is reduced, but also the playing peer gets better quality service.