NBD: Controlling Applications by Managing Network Characteristics

该项目研究使用局域网性能的受控降级来控制可以在网络上有效运行的应用程序的类型。网络管理员可以使用该项目开发的技术来防止不良应用程序(对等文件共享、网络游戏等)。仅通过调整网络性能即可在网络上运行。例如，一些网络游戏可能会因抖动变化而无法忍受，或者实时音频可能会因高丢失率而中断，而不会影响Web浏览、远程文件访问或电子邮件等应用程序。即使你不能登录到连接到网络的计算机，这种技术也有效，用户更难逃避。该项目将构建实用的工具，使网络管理员能够以这种方式控制他们的网络，并提供有关如何在常见情况下使用这些工具的见解。该方法的主要挑战是找到控制重要应用程序的基本所需网络条件，使得程序员或用户在面对这些降级条件时使用应用程序的任何尝试都不可能取得成果。该项目还将调查网络性能的分析问题，特别是带宽和延迟等常见网络性能指标的导数和积分的使用。更广泛的影响：这一分析将有助于为项目建立控制机制，并总体上增加研究社区对网络行为的理解。该项目开发的软件和分析工具将向研究界发布。

This project investigates using controlled degradation of local area network performance to control the types of applications that can run effectively on the network. Network administrators can use the techniques developed by the project to prevent undesirable applications (peer file sharing, network games, etc.) from being run on the network merely by adjusting network performance. For example, some network games can be made intolerable by varying jitter or live audio can be disrupted by high loss rates, without impacting applications like web browsing, remote file access, or email. This technique works even if you cannot log into computers connected to the network and it is much harder for users to evade. The project will build practical tools that allow network administrators to control their networks in this manner and provide insight into how to use those tools for common cases. A major challenge for this approach is to find fundamental required network conditions that control important applications such that no attempt by programmers or users to work with the applications in the face of those degraded conditions is likely to be fruitful. The project will also investigate analytic issues of network performance, in particular the use of derivatives and integrals of common network performance metrics like bandwidth and delay. Broader Impact: This analysis will both assist in building control mechanisms for the project and generally increase the research community's understanding of network behaviors. The software and analytic tools developed by this project will be released to the research community.