FMitF: Transplanting Syntax-Guided Synthesis to Computer Networks

Purdue University, West Lafayette IN

Investigators

Abstract

Computer networks are difficult to manage since there exists a wide gulf between the high-level goals that operators have for their networks (e.g., security, Quality of Service etc.), and the low-level configurations in heterogeneous devices in which such intent must be expressed. While the advent of Software-Defined Networking (an emerging technology that centralizes control decisions regarding how traffic must be handled, and separates them from the devices that actually forward each packet) helps, the process of designing networks is still ad-hoc, leading to high operational costs, design faults that account for a large fraction of network downtime, and costly security breaches. This project is motivated by the vision of design automation for networking, inspired by the success of the approach in other domains such as chip design. The project will develop methods for network architects to express their intent at higher levels of abstraction, and techniques to automatically synthesize low-level switch configurations that realize this intent correctly and efficiently. The project will tackle network automation through syntax-guided synthesis, which has been a popular paradigm embodied in many program synthesis systems. Rather than directly synthesize low-level switch flow-table rules and switch configurations like current network synthesis efforts, the project will explore how to synthesize code from user-provided sketches and specifications, corresponding to a network programming language, which may then be translated into low-level switch configurations. This approach will enable human experts to convey their insights and hints, which are critical for applying modern synthesis techniques. The project will develop a network programming language that is expressive yet has the necessary formal semantics to enable synthesis, explore the design space of sketches for network synthesis, and develop ways to improve the performance of synthesis for computer networks. The broader impact of this project is to raise the level of abstraction for managing large networks, leading to much lower management costs, better performance, security and reliability. The project will extensively involve graduate and undergraduate students, and incorporate new curriculum material in programming language and networking classes. 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.