CI-New: Collaborative Research: An Open Platform for Internet Routing Experiments

数据从其源到目的地通过Internet所采用的路径是用户体验的服务质量的关键决定因素。路径通过路由进行控制，受流量工程和策略的影响。因此，路由问题--不稳定、故障、迂回路径--对性能有重要影响。尽管存在这些问题，研究人员和网络运营商仍难以研究互联网路由。可用的有利位置几乎无法了解网络如何互联，因此很难评估现有路径的恢复能力和性能。除了有限的能见度，研究人员缺乏尝试替代路径的手段。该项目开发的基础设施将允许研究人员和网络运营商观察互联网路径，并试验替代的路由方法。该基础设施由两个协同工具组成：中转门户允许实验者发布影响当前路由的路由公告，而反向跟踪路由允许实验者测量得到的路径。PI使用早期版本的基础设施来提高互联网性能、可靠性和诊断能力。该项目开放了供网络社区和教室使用的基础设施，为理解和改进互联网提供了基础。关键组件是1)增加运输门户网站的数量；2)增加对运输门户实验的支持，将完全模拟的网络连接到实际的互联网；3)升级和强化反向跟踪路由代码，使其可被更广泛的社区使用；以及4)通过更高效地调度和重用测量来改进实验工作流程。

The path that data takes across the Internet from its source to its destination is a critical determiner of the service quality that a user experiences. The path is controlled via routing, which is influenced by traffic engineering and policy. Therefore, routing problems--instability, failures, circuitous paths--have a critical impact on performance. Despite these issues, researchers and network operators have difficulty studying Internet routing. Available vantage points provide little visibility into how the networks interconnect, making it difficult to assess the resilience and performance of existing paths. On top of this limited visibility, researchers lack means to experiment with alternate paths.The infrastructure developed by this project will allow researchers and network operators to observe Internet paths and to experiment with alternate approaches to routing. The infrastructure consists of two synergistic tools: Transit Portal allows an experimenter to make route announcements that influence current routing, and Reverse Traceroute allows the experimenter to measure the resulting paths. The PIs have used early versions of the infrastructure to improve Internet performance, reliability, and diagnosis. The project opens the infrastructure to use by the networking community and in classrooms, providing a basis for understanding and improving the Internet. The key components are 1) increasing the number of Transit Portal sites; 2) adding support for Transit Portal experiments that join fully emulated networks to the actual Internet; 3) upgrading and hardening the Reverse Traceroute code so that it is useable by the broader community; and 4) improving experiment workflow by scheduling and reusing measurements more efficiently.