CRII: NeTS: Towards Measurement and Optimization of Internet Video Quality of Experience

流媒体视频是迄今为止互联网流量的最大组成部分。目前，视频占全球互联网流量的66%以上。视频流量预计将在2013年至2018年期间翻两番。互联网视频的增长是由免费的基于订阅的服务（如YouTube）和付费的基于订阅的服务（如Netflix）的流行所推动的。为了维持这些收入模式，业界和学术界的研究人员都在致力于重新构建互联网视频交付基础设施，以提高用户的体验质量（QoE）。首先，在接近真实的时间和规模上准确地量化视频QoE是具有挑战性的。主观指标通常是不切实际的，客观指标会导致计算开销并带来隐私问题。其次，由于视频服务器和终端用户之间的路径包括内容提供商、内容分发网络（CDN）和互联网服务提供商（ISP）网络中的一系列异构链路，因此难以隔离和测量各个链路及其互连（即，对等和转接链路）。第三，TCP/IP栈的各个层处的许多网络协议同时影响视频QoE。在不同层的协议之间分离依赖关系是很困难的。首先，它旨在设计准确、可扩展和客观的视频QoE度量。这些措施将有助于QoE预测和识别影响QoE的因素。其次，它旨在开发一个工具箱来测量，分析和优化异构链路和协议对视频QoE的影响。该工具箱将有助于识别QoE瓶颈并阐明缓解这些瓶颈的方法。该项目的成功完成将有助于弥合数字鸿沟，改善有挑战的消费者，特别是服务不足和农村地区的消费者的视频流体验。该项目的结果也可能为联邦通信委员会（FCC）提供感兴趣的见解。该项目将连接网络、信号处理和人机交互领域，并促进它们之间的互动。PI将与利益相关者密切合作，促进技术从研究环境转移到实际商业部署。该项目的教育方面将把研究成果纳入爱荷华州大学的本科和研究生课程。

Streaming video is by far the largest constituent of Internet traffic. Video currently makes up more than 66% of the global Internet traffic. The volume of video traffic is expected to quadruple between 2013 and 2018. The growth of Internet video is fueled by the popularity of free advertisement-based (such as YouTube) and paid subscription-based (such as Netflix) services. To sustain these revenue models, both industry and academic researchers are working on re-architecting Internet video delivery infrastructure for improving users' Quality of Experience (QoE).Many challenges arise when measuring and optimizing video QoE. First, it is challenging to accurately quantify video QoE in near real time and at scale. Subjective metrics are often impractical, and objective metrics incur computational overheads and pose privacy issues. Second, because the path between a video server and an end-user comprises a series of heterogeneous links in content provider, Content Delivery Network (CDN), and Internet Service Provider (ISP) networks, it is difficult to isolate and gauge the impact of individual links and their interconnections (i.e., peering and transit links) on video QoE. Third, many network protocols at various layers of the TCP/IP stack simultaneously impact video QoE. It is difficult to separate dependencies among protocols at different layers.The objectives of this project are two-fold. First, it aims to design accurate, scalable, and objective measures of video QoE. Such measures will be useful for QoE prediction and identifying factors that affect QoE. Second, it aims to develop a toolbox to measure, analyze, and optimize the impact of heterogeneous links and protocols on video QoE. The toolbox will be useful to identify QoE bottlenecks and shed light on ways to alleviate them. Moreover, it will reveal issues with the underlying network protocols and provide guidance on fixing them.Successful completion of this project will help bridge the digital divide by improving video streaming experience for challenged consumers, particularly those in underserved and rural areas. The outcomes of this project might also provide insights of interest to the Federal Communications Commission (FCC). This project will bridge the areas of networking, signal processing, and human-computer interaction and foster interaction among them. The PI will work closely with stakeholders to facilitate the transfer of technology from the research environment to actual commercial deployments. The educational side of the project will integrate research findings in the undergraduate and graduate curricula at the University of Iowa.