NeTS: Small: Collaborative Research: Studying and Improving the Performance of Access Networks

由于国家宽带计划和相关的基础设施投资，美国将看到高速连接越来越多地渗透到家庭中。同样，世界各地越来越多的用户将宽带接入网作为一种重要的（如果不是主要的）互联网接入手段——例如，这些网络在撒哈拉以南非洲的发展中地区变得越来越普遍。研究人员、运营商和政策制定者需要了解宽带接入网络的性能，网络性能如何影响人们通常使用的服务，以及最终如何提高这些网络和服务的性能。从用户对家庭网络设备的选择到网络链路的拥塞，再到互联网服务提供商之间的业务安排，许多因素都会影响用户对不同应用程序的体验。然而，尽管我们依赖宽带接入网络的互联网连接，但我们对各种因素如何影响流行服务的最终性能知之甚少。更好地了解用户从网络边缘体验到的性能和性能差的原因——由实际部署和可靠的经验方法的数据驱动——可以更好地为我们关于政策和技术投资的讨论提供信息，并最终改善全球数十亿互联网用户的体验。理解和提高两个家庭网络的性能需要一个灵活、可扩展和易于管理的底层网络基础设施。我们开发的算法和工具可以帮助用户和网络运营商排除故障并改善用户的网络性能，最终将带来更好的性能和更低的运营成本。为了实现这一目标，该项目正在开发平台和首个同类数据集，以帮助用户、运营商和研究人员更好地了解接入网性能的本质。该项目分为两个部分：(1)对接入网的性能进行表征和改进；(2)表征和改进接入网的应用和业务性能。以往对住宅接入网络的研究主要关注家庭网络和互联网服务提供商之间的有线接入链路的性能，与之相反，这项研究采用了一种整体的方法，也探索了家庭无线网络的性能和家庭内部的一系列因素，这些因素最终可能会影响用户体验的性能。此外，该项目正在开发更好的方法来帮助服务提供商确定这些性能下降发生的原因（以及在哪里），例如问题是否由用户的家庭网络或设备引起，服务的某些方面的故障或降级(例如，用户？可能是服务器运行缓慢，或者其他原因。在与几家大型服务提供商现有合作的基础上，研究人员将从网络边缘探索服务性能下降的潜在原因，以及减轻这种下降的方法。研究人员还将与发展中地区的政策组织合作，以更好地了解其网络和应用程序性能的性质，以及改善这些地区性能的可能机制和建议。该项目的成果将以多种方式传播，包括整合到校园课程中，以及在适当的情况下，大规模开放在线课程。研究人员与几个国家的互联网服务提供商、内容提供商、政策组织和监管机构有着密切的关系，并计划通过出版物、讲座、教程和研讨会将研究结果传达给这些团体。他们也有让本科生和少数民族参与研究的历史，并将把这些努力作为这个项目的一部分。例如，考虑到在发展中地区测量宽带的重要性，这些科学家将做出具体的协调一致的努力，通过讲座、研讨会以及在适用的情况下直接提供建议，让发展中地区的学生参与进来。

The United States will see increasing penetration of high-speed connectivity into homes as a result of the national broadband initiative and related infrastructure investments. Similarly, more users around the world are using broadband access networks as an important (if not primary) means of Internet access-for example, these networks are becoming increasingly prevalent in developing regions of sub-Saharan Africa. Researchers, operators, and policymakers need to understand the performance of broadband access networks, how the performance of the network affects the services that people commonly use, and how ultimately to improve the performance of these networks and services. Many factors, ranging from a user's choice of equipment in their home network to congestion on network links to business arrangements between Internet service providers, can affect the performance that users experience for different applications. Yet, despite our dependence on Internet connectivity from broadband access networks, we have little understanding of how various factors affect the ultimate performance of popular services. A better understanding about the performance that users experience from the edge of the network and the causes of poor performance -- driven by data from real deployments and sound empirical methods -- can better inform our discussions about both policy and technology investment and ultimately improve the experience for billions of Internet users around the world. Understanding and improving performance in both home networks will require an underlying network infrastructure that is flexible, extensible, and easy to manage. The algorithms and tools that we develop to help users and network operators troubleshoot and improve network performance to users will ultimately lead to better performance and lower operational costs. Towards this goal, this project is developing platforms and first-of-their-kind datasets that will help users, operators, and researchers better understand the nature of performance in access networks. The project has two parts: (1) Characterizing and improving the performance of access networks; and (2) Characterizing and improving application and service performance from access networks. In contrast to previous studies of residential access networks, which have focused on the performance of the wired access link between the home network and the Internet service provider, this effort takes a holistic approach by also exploring the performance of the home wireless network and a range of factors inside the home that may ultimately contribute to the performance that users experience. Additionally, this project is developing better ways to help service providers determine why (and where) these performance degradations occur, such as whether problems are caused by the user's home network or device, the failure or degradation of some aspect of the service (e.g., the user?s domain name system (DNS) server may be slow), or some other cause. Building on existing collaborations with several large service providers, the investigators will explore the underlying causes of service performance degradation from the edge of the network, as well as ways to mitigate such degradation. The researchers will also work with policy organizations in developing regions to better understand the nature of their network and application performance, as well as possible mechanisms and recommendations for improving the performance in these regions.The project's outcomes will be disseminated in many ways, including integration into on-campus courses and, where appropriate, massive open online courses. The researchers have close relationships with Internet service providers, content providers, policy organizations, and regulators in several countries and plan to communicate the results of the research to these groups, through publications, talks, tutorials, and workshops. They also have a history of involving undergraduates and minorities in research and will extend these efforts as part of this project. For example, given the importance of measuring broadband in developing regions, the researchers will make specific concerted efforts to engage students in developing regions, through lectures, seminars, and, where applicable, direct advising.