CAREER: Bug Tolerant Networking: Enabling Highly-Available Network Infrastructures through Semantic Transformations.

软件定义网络（SDN）是促进向下一代软件定义基础设施过渡的关键使能技术，该基础设施结合了虚拟化网络功能、第五代蜂窝网络和先进的无线通信技术以及边缘计算。尽管SDN将发挥至关重要的作用，但网络研究界仍然需要关键技术和工具来维护高可用性和弹性网络；sdn容易出现软件设计和实现错误，从而导致灾难性的后果，例如网络故障。本项目通过引入一种新的范例来解决这些需求，该范例主张通过将触发错误的输入转换为“安全”输入来克服软件的“缺陷”。为了支持这一范式，该项目将开发设计、抽象和算法，以系统和原则的方式识别、隔离并将触发输入转换为安全输入。该项目将研究几个研究重点，以实现实际的容错网络，包括：(1)引入框架和设计，通过将触发bug的输入转换为不同但等效的输入，保证bug恢复和向前进展；(2)设计一个抽象和跨层网络事务，在触发事件恢复期间和之后提供原子性和一致性；(3)通过静态分析控制和数据平面源代码，以及对协议交互建模，开发算法，使所提出框架的效率最大化；(4)将当前SDN范例（例如OpenFlow）之外的技术推广到具有有状态数据平面和表达接口的新兴和未来范例（例如P4）。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。

Software-defined networking (SDN) is a key enabling technology facilitating the transition to a next generation of software defined infrastructure incorporating virtualized network functions, 5th generation cellular networks and advanced wireless communication technologies, and edge computing. Despite SDN's crucial role ahead, the networking research community still needs key techniques and tools to maintain highly available and resilient networks; SDNs are prone to software design and implementation errors that can have catastrophic consequences, e.g., network failures.This project addresses these needs by introducing a novel paradigm that advocates overcoming software `bugs-by transforming the bug triggering inputs into 'safe' inputs. To support this paradigm, the project will develop designs, abstractions and algorithms to identify, isolate, and transform the triggering inputs into safe inputs in a systematic and principled manner.This project will examine several research thrusts to enable practical bug-tolerant networks including to (1) introduce frameworks and designs that guarantee bug recovery and forward progress by transforming bug-triggering inputs into different, but equivalent inputs; (2) design an abstraction and cross-layer network transactions that provide atomicity and consistency during and afterrecovery from triggered events; (3) develop algorithms that maximize the efficiency of the proposed frameworks by statically analyzing control and data plane source code, and modeling protocol interactions; and (4) generalize techniques beyond current SDN paradigms (e.g., OpenFlow) to emergingand future paradigms with stateful data planes and expressive interfaces (e.g., P4).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.