NeTS: Small: Meta-Networking Research: Analysis, Partitioning, and Mapping Tools for Large Experiments

该项目解决了大规模网络实验的规模化和半自动化这一紧迫的研究课题。由于互联网中协议和设备的巨大规模和异构性，研究人员需要进行实验，因此网络研究在技术上具有挑战性。为了解决这一问题，该项目将调查和解决资源限制以及测试平台和扩展技术的实验人工制品的问题。到目前为止，工作要么集中在控制平面协议(例如，路由模拟器)上，要么仅仅考虑数据平面。该项目的目标是以前所未有的规模进行联合控制平面和数据平面的实验。该项目包括三个相辅相成的工作，通过设计来解决实验规模挑战：(1)实验映射工具和分类：该项目将设计一个通用框架、分类和一套工具，以弥合当前试验台用户和使用多种缩放技术的大型试验台实验之间的差距。用户可以就不同组件的期望保真度提供提示，这些提示将被用于确定实验的高保真映射。(2)实验分析和划分工具：该项目将设计方法来模拟大型实验组件之间的复杂依赖关系，以便于规划和映射。这些模型还允许将大型实验划分为可以按顺序执行的最大独立的较小实验。(3)案例研究的应用：该项目将使用一系列大型实验作为应用，重点放在有问题的实验上，包括(A)实验以了解错误配置、攻击和防御对互联网基础设施的影响(例如，RPKI的可扩展性、蠕虫或DDoS对BGP的影响、BGP策略冲突)、(B)异常检测实验和(C)云计算实验。研究将有助于识别和部署可扩展的协议，使互联网能够安全地容纳增加的流量。这项研究的影响包括开发和公开传播通用实验工具、大规模测试技术、使用测试框架的方法以及相关的研究生级课件。PI将对模拟和试验台团队进行重要的外展工作，例如，威慑/Emulab、GENI、AutoNetkit、ns-3和工业。国际计算机协会将积极让计算机科学领域中代表性不足的少数群体的本科生和研究生参与研究和教育工作，并将组织一次关于项目专题的DIMACS讲习班。

This project attacks the pressing research problem of scaling and semi-automating large network experiments. Networking research is technically challenging due to the great scale and heterogeneity of protocols and devices in the Internet that researchers need to experiment with. To address this the project will investigate and tackle the problems of resource limitations and experimental artifacts of the testing platforms and scaling techniques. Work to date has either focused on control plane protocols (e.g., routing simulators), or solely considered the data plane. This project's goal is to conduct experiments that are joint control plane and data plane at a scale not previously possible. The project project includes three complementary efforts to address experimentation scale challenges by designing:(1) Experiment Mapping Tools and Taxonomy: The project will design a general framework, taxonomy, and a set of tools to bridge the current gap between testbed users and large-scale testbed experiments that use multiple scaling techniques. The user can supply hints on desired fidelity of different components, and these will be used to determine a high fidelity mapping for the experiment.(2) Experiment Analysis and Partitioning Tools: The project will design methods to model complex dependencies between components of a large-scale experiment to facilitate planning and mapping. These models may also allow partitioning the large experiment into maximally independent smaller experiments that can be sequentially executed to mimic the large experiment.(3) Applications to Case Studies: The project will use a range of large experiments as applications, focusing on problematic experiments including (a) experiments to understand the effect of misconfigurations, attacks, and defenses on Internet infrastructure (e.g., scalability of RPKI, effect of worms or DDoS on BGP, BGP policy conflicts), (b) experiments for anomaly detection, and (c) experiments with cloud computing.The research will help identify and deploy scalable protocols that will enable the Internet to securely accommodate increased traffic volumes. Impacts of the research include the development and public dissemination of general-purpose experimental tools, large-scale testing techniques, methodologies for the use of testing frameworks, and related graduate-level courseware. The PI will undertake significant outreach efforts to simulation and testbed teams, e.g., DETER/Emulab, GENI, AutoNetkit, ns-3, and to industry. The PI will actively involve undergraduate and graduate students from under-represented minority groups in computer science in the research and educational efforts, and will organize a DIMACS workshop on project topics.