Scalable Network and Transport Supports for QoE Fairness in Streaming Media Services

视频流应用已经无处不在，预计到2021年将占全球IP流量的82%。Netflix和亚马逊（Amazon）等提供商越来越多地寻求取代传统电视体验，而不是补充传统电视体验，这得益于多种设备的客户端可用性，允许多个用户同时流式传输视频内容。为了确保用户的体验质量（QoE），有一些相互竞争的行业标准，其中一个例子是基于HTTP的动态自适应流（DASH）。DASH是一种编解码器无关的技术，用于从HTTP服务器交付内容，Netflix和YouTube等公司都使用DASH。媒体被分割成等长、不重叠的片段，以不同的比特率编码，并从服务器提供。基于当前网络条件，客户端应用程序以尽可能高的比特率请求媒体。请求的比特率可以根据可用带宽的波动从一个段变化到另一个段。DASH和类似的标准旨在确保应用程序用户获得最佳的QoE。然而，鉴于用户习惯的转变，视频流现在经常在一个常见的网络瓶颈（最常见的是家庭路由器）后面争夺带宽。最近在该领域的研究表明，当前的自适应机制并不能改善QoE，反而会导致用户的不稳定和竞争应用之间的不公平。为了提供这个问题的长期解决方案，有必要探索一种新的全网架构，该架构实现了类似于传输独立路径层状态管理b[4]的方法，该方法建立在使客户端应用程序能够在网络级别公开优化QoE所需的信息的基础上。

Video streaming applications have become ubiquitous and are predicted to account for 82% of global IP traffic by the year 2021[1]. Providers such as Netflix and Amazon increasingly seek to replace rather than complement the tradition television experience, aided by the availability of clients for multiple devices allowing multiple users to stream video content simultaneously.There are competing industry standards aimed at ensuring quality of experience (QoE) for users, one example is Dynamic Adaptive Streaming over HTTP (DASH). Used by Netflix and YouTube among others, DASH is a codec-agnostic technique for delivering content from HTTP servers. Media is broken into equal length, non-overlapping segments, encoded at various bitrates and made available from the server. Client applications request media at the highest bitrate possible based on current network conditions. The requested bitrate can be varied from one segment to another depending on fluctuations in available bandwidth.DASH and similar standards aim to ensure the best possible QoE for application users. However, given the shift in user habits, video streams now frequently compete with each other for bandwidth behind a common network bottleneck, most often a home router. Recent work in this area has shown the current adaptation mechanisms do not improve QoE, but instead lead to instability for users and unfairness between competing applications.In order to provide a long term solution to this problem it is necessary to explore a new network- wide architecture that implements an approach similar to Transport-Independent Path Layer State Management[4] This builds upon the work in to enable client applications to expose the information necessary to optimise QoE at a network level.