NeTS: Large: Collaborative Research: Design Principles for a Future-Proof Internet Control Plane

当今互联网面临的一大挑战是与日益多样化的使用模式和日益增长的用户对服务质量的期望相关的日益增长的复杂性。互联网中使用的核心技术Ipv4和BGP是基于20世纪70年代的S和80年代初的S开发的技术，当时互联网的重点是在简单的边缘网络和应用之间提供连接。相比之下，今天的用户和服务提供商需要高质量的体验，例如流畅的高清视频、响应迅速的电子商务商店、交互式社交网络等。应对这些挑战很困难，因为事实证明IP和BGP都非常僵化，很难修改。虽然已经有很多关于互联网数据平面的从头开始的提议的研究，但控制平面的研究主要集中在特定的协议上。该项目通过使用整体方法来设计互联网控制平面来填补这一空白，该方法可以有效地适应对互联网日益增长的需求。根据我们在XIA未来互联网体系结构项目中的经验，我们定义了三个独立但相关的需求。首先，控制协议必须是可演变的，这样它们才能适应未来不断出现的需求。其次，互联网路由协议必须是资源感知的。最后，互联网控制平面必须支持与内容交付网络(CDN)和云基础设施等日益复杂的边缘网络的控制平面进行协调。该项目有一个广泛的研究议程，包括研究互联网中对可进化性和资源感知的协议支持，定义用于跨边缘和转换域的协调的接口和技术，以及提供对常见控制协议功能(如发现和控制通信)的支持的控制平面体系结构。这项研究使用了几个现实世界的驱动因素，包括流量工程、云计算和大规模分布式拒绝服务(DDoS)防御服务。最后，该项目将包括使用CloudLab、GENI和PEERING等大型测试平台的端到端实验评估工作。智力优势：该项目的主要智力优势是，它以集成的方式解决了互联网控制协议(可进化性、资源感知和跨不同控制平面协调的接口)方面的三个基本挑战。这三个领域的成果将被整合到一个新的互联网控制平面架构中。更广泛的影响：该项目也将产生重大的更广泛的影响。首先，将向研究社区提供在CloudLab、Geni和Peering等试验台上运行大规模网络实验的工具。其次，该项目将吸引本科生和硕士学生参与该项目，并将成果整合到CMU和威斯康星大学麦迪逊分校的课程中，并将组织DIMACS研讨会，让社区参与讨论未来互联网的控制面设计这一重要主题。最后，重新思考网络控制平面的整体方法可能会对网络研究社区和行业产生巨大影响。

One of the big challenges faced in the Internet today is the ever growing complexity associated with the increasingly diverse usage models and growing user expectations about quality of service. The core technology used in the Internet, IPv4 and BGP, is based on technology developed in the 1970's and early 1980's, when the focus of the Internet was providing connectivity between unsophisticated edge networks and applications. In contrast, today's users and service providers want high Quality of Experience, e.g., smooth HD video, responsive e-commerce stores, interactive social networks, etc. Addressing these challenges is hard because both IP and BGP have proven to be very rigid and hard to modify. While there has been much research on clean slate proposals for the Internet data plane, research in the control plane has mostly focused on specific protocols. This project fills this void by using a holistic approach to the design of the Internet control plane that can adapt to efficiently meet the growing demands placed on the Internet.Based on our experience with the XIA future Internet architecture project, we define three separate but related requirements. First, control protocols must be evolvable so they can adapt to future emerging requirements. Second Internet routing protocols must be resource aware. Finally, the Internet control plane must support coordination with the control planes of increasingly sophisticated edge networks such as content delivery networks (CDNs) and cloud infrastructure. The project has a broad research agenda that includes research in protocol support for evolvability and resource-awareness in the Internet, the definition of interfaces and techniques for coordination across edge and transit domains, and a control plane architecture that provides support for common control protocol functions such as discovery and control communication. This research uses several real-world drivers including traffic engineering, cloud computing and a large-scale distributed denial of service (DDoS) defense service. Finally, the project will includes an end-to-end experimental evaluation effort using large-scale testbeds such CloudLab, GENI, and PEERING.Intellectual merit: The main intellectual merit of the project is that it addresses three fundamental challenges in for Internet control protocols (evolvability, resource-awareness, and interfaces for coordination across heterogeneous control planes) in an integrated fashion. The results in these three areas will be integrated in a novel control plane architecture for the Internet.Broader Impact: The project will also have significant broader impact. First, tools to run large scale networking experiments across testbeds such as Cloudlab, GENI and PEERING will be made available to the research community. Second, the project will engage undergrad and MS students in the project and integrate results in courses at CMU and UW-Madison, and will organize a DIMACS workshop to engage the community in discussion on the important topic of control plane design for the future Internet. Finally, the holistic approach to rethinking the network control plane can have tremendous impact on the networking research community and industry.