Selection Mechanism With Improving Mutual Assistance for Piece Downloading Request in Peer-to-Peer Video-on-Demand Systems

• Main Authors: TSAN, YI-CHEN, 昝易辰
• Other Authors: LU, CHUN-HSIEN
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/77208070830363848279

碩士 === 輔仁大學 === 資訊工程學系碩士班 === 105 === As the Internet becomes popular, the video traffic on the Internet grows rapidly as well. In the traditional Client-Server architecture the server is likely to become the bottleneck, therefore the Peer-to-Peer streaming services gradually takes over as the mainstream architecture. In a general Peer-to-Peer multimedia streaming system, each produced data piece must meet a playback deadline, thus every peer will download the pieces sequentially, and usually send its download requests to its most powerful neighbor. However, in this situation most peers will try to request the same pieces if these peers join the system at about the same time, and they all will have to download from the server because none of the peers have those pieces, which will increase the server's load significantly. When some of the neighbors have the piece, the powerful peer will receive a large amount of requests from its neighbor, which will overload the peer, and slows down its service. Eventually, the quality of viewing experience will be greatly affected. In this thesis, we propose a piece selection strategy and a request transmission strategy. To determine the next piece to download in the Un-urgent area, a peer calculates the probability of selection for all the candidate pieces according to their importance among the neighbors. In selecting which neighbor to transmit the request to a peer will consider both the neighbor's service response time and its potential contribution, and then give priority to a neighbor with less potential contribution to balance the load between neighbors. Simulation results show that our method apparently improves the assistance among the neighboring peers. Compared to the previous approach, our approach can reduce the load on the server by at least 40%, and the average viewing time can be reduced by about 60%. Keywords: P2P system, P2P VoD, piece selection, request-peer selection