基于对等网的可扩展流媒体分发模型研究

The research project is intended to provide a low-cost high availability of high-quality scalable video distribution of new model.Firstly, popular video request prediction model used to estimate the maximum amount of requests at different times according to the video sampling theory to determine the demand for different periods of cooperation nodes appointment; Then, Signaling Game Theory is applied to establish video distribution cooperative node selection model, with added economic incentive award compensation mechanism to attract nodes to cooperate and to choose the high reliability of the appointment of cooperation nodes ready to participate in the service side distribute appointment slots video; By applying queueing theory and mathematical programming theory, to effectively reduce the packet loss probability of aggregated bandwidth，a new algorithm with the pattern of multiple senders and single receiver is proposed, and to minimize transmission delay of streaming videos, a new algorithm with the pattern of single sender and multiple receivers is proposed; To meet the needs of large-scale streaming distribution applications and reduce cross-regional traffic, based on the video request landmark users are grouped;On this basis, the scalable streaming media distribution model to combine advantages of CDN and P2P, is proposed. By applying model, service side demand for video distribution cooperative nodes, can be adjusted according to dynamical changes in the users of request video,and to guide the cooperation node provides optimal video stream to a specified group of users.With reliable cooperation node assisted, service side can provide high quality and resilient scalable service, but also reduce the cost of service.

本项目拟研究一种以低成本提供高可用高可扩展高质量的视频分发新模型.首先根据抽样理论建立视频请求流行度预测模型,用于估计不同时段对视频的最大请求量,从而确定不同时段对预约合作节点需求量;然后应用信号博弈原理建立视频分发合作节点选择模型,用激励加经济奖赔机制吸引节点合作,并从中选出高可靠性的预约合作节点预备参与服务方在预约时段的视频分发;应用排队论和数学规划理论分别设计多对单低丢包率和单对多低延迟动态数据分配算法,用于视频副本和视频流高效分发;为满足大规模流分发应用需要和降低跨地域通信量,基于地标对视频请求用户群分组;在此基础上建立融合CDN与P2P技术优势的可扩展流媒体分发新模型。模型能根据不同时段视频请求用户量动态变化而实时调整服务方对视频分发合作节点的需求量,并最优地引导合作节点向指定的用户群提供视频流,在可靠合作节点辅助下服务方既能提供高质量高可扩展的弹性服务,也降低了服务成本.

本项目围绕高可用高可扩展高质量的视频分发新模型，从视频分发合作节点选择模型、单对多低延迟动态数据分配算法和多对单低延迟低丢包率动态数据分配算法、用户群分组、视频请求流行度预测模型、弹性服务策略方面开展了深入的研究工作，取得多项创新性成果，包括：混合流媒体分发系统中可信合作节点选取模型、对等网端到端多路径选择模型及算法、多服务节点合作直播系统的设计与实现、基于Markov预测模型的视频流自适应算法设计与实现、一种H.265视频带宽实时预测模型、基于DASH标准的视频自适应传输方案、P2P流媒体服务器负载均衡算法、基于组合定权法的负载评价模型、多指标用户群分组算法、基于视频特征及历史数据的流行度预测算法等。. 迄今为止，团队在该项目的资助下发表论文13篇（期刊论文10篇，国际会议3篇，2篇被EI收录）；申请发明专利4项，完成计算机软件著作权登记8项；参加国际学术会议3次；培养从事流媒体技术研究与应用的研究生12名, 超额完成了预期研究目标。