Research On Data Scheduling In P2P Live Streaming System

With the rapid development and evolving of the Internet, ultra-large scalestreaming media application has become the mainstream of development of theInternet and multimedia technologies. Peer to peer (P2P) network supports the largescale streaming media with technology bases. With their advantages of greatextensibility and ease of deployment, P2P streaming media systems have quicklytaken up the market shares. However, there are still many problems lying within theP2P systems that need further studies.In P2P streaming media systems, large numbers of nodes take part in massivedata transferring, the latter of which stress heavily on being delivered timely.Therefore, proper data scheduling strategies have to be designed. By deep researcheson the problem of scheduling data in P2P streaming media systems, we come up withthe following:In order to adapt to the heterogeneous bandwidths of Internet peers, layeredvideo encoding is employed, in which the original video stream is encoded into onebase stream and multiple enhancement streams. The base stream supports the mostbasic quality of the video, thus is required to be decodable independently; On theother hand, the enhancement streams provide users with better video experiences viaenhancing the video quality. Therefore, users can request video layers corresponsiveto their own bandwidths. In this paper, we establish the layered data scheduling model,define the priorities of data blocks in the scheduling model, and put forward amulti-level Buffer Map (BM) model, with which we describe the basic procedures oflayered data scheduling. Finally, we examine the superiority in performance oflayered data scheduling.Segmented buffer is employed to further categorize the data blocks in the nodebuffer, thus the buffer is sliced into discarded, urgent, strategy selection, and data scheduling segments. According to RTT we set a threshold to give up to request someblocks which are very close to the playback point, we give a method to calculate theemergency selectionboundary, and drive the method to calculate the probability toselect different layers. Last but not least, the data scheduling segment selects differentscheduling policies in order to achieve the maximum throughput across the system.By optimizing the model, we analyzed the data scheduling problem in P2Pstreaming systems, discussed on the objectives of data scheduling optimizations, andtransforms the optimal scheduling problem into an equivalent problem of minimumcost flows, which can be solved within binomial times, according to analysis results.Then, we solve the optimization problems via finding the maximum flows of thevirtual networks. We also verified the results’ correctness and feasibilities via theoryanalyses and emulations.