NeTS: Medium: Collaborative Research: Automatic Network Repair

第1部分：现代计算机网络必须满足日益复杂的安全性、可用性和性能目标，以满足用户和应用程序的不同需求。因此，对网络进行编程以满足这些要求已变得极具挑战性，即使对高技能的网络操作员也是如此。虽然以前的工作已经帮助运营商生成了符合策略的网络配置，但网络仍然可能遇到意想不到的行为，如链路故障或流量激增，这可能导致违反策略，需要对网络进行修复。如今，维修通常是手工完成的，而且容易出错。该项目将开发一个基础框架，用于自动有效地修复不满足给定要求的网络。在先前工作的基础上，这项研究为“零接触网络”奠定了基础，在这种网络中，网络的行为是基于人类操作员提供的高级目标自动管理的。这项研究的结果将使没有经过广泛培训的个人能够修复网络数据和控制平面，从而为提高企业、政府和个人严重依赖的网络的稳定性和安全性铺平道路。第2部分：该项目建议开发一个正式框架，用于自动创建最佳网络修复，并为生成的修复质量提供正式保证。具体而言，该项目将在三个方面进行研究：1)为分布式控制平面设计正式抽象，为广泛的策略计算高质量的修复。2)为集中的控制平面设计正式的抽象，为大范围的策略计算高质量的修复。3)设计并实现在存在概率事件的情况下修复控制平面的算法（例如，符合交通工程策略）。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Part 1: Modern computer networks must satisfy increasingly complex security, availability, and performance objectives to meet the diverse needs of users and applications. Consequently, programming a network to satisfy these requirements has become extremely challenging, even for highly skilled network operators. While previous work has helped operators generate policy-compliant network configurations, networks can still experience unexpected behaviors such as link failures or traffic surges, which can lead to policy violations that require a repair of the network. Today, repairs are often done manually and are error-prone. This project will develop a ground-up framework for automatically and efficiently repairing networks that do not satisfy given requirements. Building on previous work, this research lays the foundation for "zero touch networking", where a network's behavior is automatically managed based on high-level objectives provided by human operators. The results of this research will enable individuals without extensive training to repair network data and control planes, thus paving the way for increased stability and security in the networks on which businesses, governments, and individuals so heavily depend.Part 2: This project proposes to develop a formal framework for automatically creating optimal network repairs and for providing formal guarantees on the quality of the generated repairs. Concretely, the project will pursue research on three fronts: 1) Design formal abstractions for distributed control planes that compute high-quality repairs for a wide range of policies. 2) Design formal abstractions for centralized control planes that compute high-quality repairs for a wide range of policies. 3) Design and implement algorithms for repairing control planes in the presence of probabilistic events (e.g., to conform to traffic engineering policies).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.