多CDN自适应流媒体分发的用户体验与网络资源联合优化研究

项目名称： 多CDN自适应流媒体分发的用户体验与网络资源联合优化研究

项目编号： No.61472204

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 孙立峰

作者单位： 清华大学

项目金额： 88万元

中文摘要： 视频已成为互联网和移动互联网上的重要应用和主要流量，多CDN自适应流媒体技术是满足大规模视频内容高效分发需求的技术途径和应用发展趋势，其中的关键问题是在满足QoE的条件下最大化内容分发网络资源利用率。本项目采用数据驱动的研究方法，以自适应流媒体用户码率切换行为和异构环境下用户视频访问上下文为研究切入点，研究建立用户体验与用户激励、系统资源消耗之间的映射模型、视频内容不同版本的访问流行度分布模型，联合优化网络资源配置和视频质量匹配的问题，设计开销与性能优化的多CDN内容部署策略、用户体验与服务器压力约束的动态码率选择算法和QoE驱动的CDN节点视频转码与传输联合优化策略，达到视频用户体验、公平性、用户兴趣偏好以及视频分发费用之间的平衡，最大化用户体验质量，并在真实系统的数据上进行验证。

中文关键词： 自适应流媒体；内容分发网络；用户体验；资源优化；网络多媒体

英文摘要： Having been dominating the Internet traffic for over a decade, online video streaming is evolving itself to a multi-CDN and HTTP based dynamic adaptive bitrate paradigm, in which a fundamental challenge - how to maximize the network resource utilization in a multi-CDN architecture while satisfying good quality-of-experience (QoE), has to be resolved. Using a data-driven approach, this proposal will first study the bitrate selection behaviors and the video streaming contextual information of users. Then we focus on (1) the correlation between QoE, user incentive mechanism, and the system resource consumption, (2) the modeling of popularity distribution of the adaptive video versions, (3) the joint optimization of network resource allocation and streaming bitrate planning. Based on the above proposed theoretical frameworks, we design the following strategies: (A) A joint resource and performance awared video deployment strategy over multi-CDN; (B) A combined QoE and server-load oriented dynamic rate selection; (C) A QoE-driven optimization strategy with joint online transcoding and delivery. The theories and strategies are designed to satisfy an optimal balance between QoE, fairness, user preference and the cost for video delivery, which will be evaluated by real traces in the project.

英文关键词： Adaptive Video Streaming;CDN;QoE;Resource Opimization;Networked Multimedia