Collaborative Research: NBD: An Abstraction Driven Approach to Characterizing and Designing Networks with Analyzable Properties

本项目研究一类新的抽象，以简化网络设计和配置，它是(I)任务驱动的，即捕获网络设计的预期性能、安全性、可管理性或弹性；以及(Ii)网络范围，即捕获整个网络的需求，而不是单个设备的需求。这与以前的工作完全不同，以前的工作只是对底层协议和机制进行建模。重点是企业网络，这是一个研究人员基本上没有探索的领域。通过对实际的企业网络设计进行自下而上的研究，该项目可以深入了解运营商的网络目标。这些研究采用了独特的？白盒？该项目制定了抽象概念，以及在三个领域衡量其有效性的理由和标准：(1)实施安全和复原力政策；(2)使用虚拟局域网简化管理；(3)通过计划维护实现网络演进。它展示了这些抽象在简化自上而下的网络设计和网络属性验证方面的威力。更具体地说，它发展了配置器的理论，即可以从高级设计需求生成箱级配置的系统，以及一组预先选择的协议和机制，并研究了语义审计。用于正式验证网络的现有箱级配置是否产生预期网络行为的技术。更广泛的影响：该研究旨在为将网络设计和配置变成一门科学提供基本知识和原则。它还将有助于提高网络管理在本科生和研究生课程中的覆盖面。普渡大学将开设一个新的网络管理研究生课程，并将使用受控研究来验证抽象的可用性，这些受控研究涉及对学生进行培训以成为IT专业人员。

This project investigates a new class of abstractions to simplify network design and configuration which is (i) task-driven, i.e., captures the intended performance, security, manageability, or resilience of a network design; and (ii) network-wide, i.e., captures the requirements of the network as a whole rather than of individual devices. It is a radical departure from prior efforts that merely model the underlying protocols and mechanisms.The focus is on enterprise networks, an area largely unexplored by researchers. Through bottom-up studies of actual enterprise network designs, the project obtains insights into the goals operators have for their networks. The studies employ unique ?white-box? methodologies involving extensive and iterative interactions with operators.The project develops abstractions, along with the rationale and criteria for measuring their effectiveness in three areas: (i) implementation of security and resilience policies; (ii) use of VLANs to simplify management; and (iii) network evolution through planned maintenance. It demonstrates the power of these abstractions in simplifying both top-down network design, and validation of network properties. More specifically, it develops the theory for ?configurators?, or systems that can generate box level configuration from high-level design requirements, and a pre-selected set of protocols and mechanisms, and it investigates ?semantic auditing? techniques for formally verifying that a network's existing box-level configuration produces the intended network behavior. Broader Impact: The research aims to produce fundamental knowledge and principles for turning network design and configuration into a science. It will also help enhance the coverage of network management in undergraduate and graduate curriculum. A new graduate class on Network Management will be created at Purdue and the usability of abstractions will be validated using controlled studies involving students training to become IT professionals.