Modeling and analysis of peer-to-peer video-on-demand system

Abstract

Peer-to-peer video-on-demand (P2P-VoD) system has been recently a promising platform for delivery of multimedia streaming services. In this system, optimal selection of peer groups to be jointly dedicated for the requested video segments is one of the important problems. Traditional peer assignment approaches are not suitable due to the architecture and characteristics of the system. In this thesis, we propose a novel peer selection approach. The ultimate goal is to effectively utilize resources of peers to achieve better performance of the overall system. For this purpose, we formalize the peer assignment process as a linear optimization model constructed according to the real P2PVoD framework. The proposed assignment procedure is compared with a classical P2P peer assignment procedure strictly based on contribution level under different scenarios. It is seen that the proposed approach provides much better performance compared to other. Moreover, extensive statistical data analysis is carried out to reveal the characteristics of the educational video systems. This study is essential for a possible simulation investigation to evaluate the performance of the proposed system. Our data set is obtained from YouTube EDU, currently the largest education video sharing site. The data set contains statistics about 20000 unique educational videos. Using this data set, it is found that videos on YouTube EDU have different static and popularity characteristics from traditional online videos.