FMitF: Track I: Flexible Data Plane Programming

There has been significant interest in the industry and research community regarding the ability to program the actions taken by network routers when they process data packets. Such a capability (referred to as data plane programming) can unleash exciting applications such as network performance diagnosis, cyberattack detection, and fast training of machine learning algorithms. Unfortunately, it is challenging for network architects to program data planes naturally and efficiently. Existing approaches are too abstract, precluding necessary optimization, or too low-level and burdensome to programmers. The project’s novelty is a radically new approach for data plane programming, which enables a network analyst to describe their network design and optimization goals naturally, flexibly, and automatically. The project’s broader significance and importance are simplifying data-plane programming, thereby benefiting the networking and Information Technology (IT) industry, and training graduate and undergraduate students in data plane programming and formal methods in computing. The project develops P4Sketch, a domain-specific, sketch-based language that extends a representative data plane language Programming Protocol-independent Packet Processors (P4). P4Sketch allows network architects to express their high-level network-wide design ideas as a program sketch, describe their customized, quantitative, and even indeterminate optimization targets, and produce desirable and optimal P4 programs. The backend of P4Sketch is a domain-specific synthesis algorithm which effectively handles external objects and functions via model generation and the interplay between the model generator and the code synthesizer. P4Sketch is also supported by a suite of tools for testing, verification, debugging and optimization. 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.