Collaborative Research: CNS Core: Medium: High-performance Network Stacks for the Edge

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

Edge computing is instrumental to the growth of many sectors of modern digital economies, ranging from enterprise IT and communication, to smart cities, manufacturing, and healthcare. By bringing resources closer to end users in the form of regional or micro-datacenters (uDCs), edge computing enables new applications with extreme low-latency and high-bandwidth demands. This project considers the design of the end-host networking stack for edge servers. Maximizing utilization of the limited uDC resources is therefore critical for both cost and performance reasons, motivating us to revisit the software and hardware layers that are used in the edge today. This project will follow a software/hardware co-design approach. The project will focus on reducing the overhead of resource scheduling and state management functions, revamping the way these functions are distributed across the stack. The ultimate objective is to minimize the CPU time wasted performing these functions, conversely maximizing the compute resources available to handle application logic. With the right innovation to overcome current obstacles, edge computing will enable new classes of applications that the current paradigm of utility computing in the Cloud is ill-suited for. These include applications that are highly sensitive to latency (e.g., 5G, situation-awareness applications, augmented reality, autonomous vehicles, and online gaming) or have heavy bandwidth demands (e.g., camera networks and video analytics). It is therefore expected that edge computing will follow the trajectory of Cloud computing and grow to a multi-billion-dollar industry within the next five to seven years. The project’s outcomes have the potential to transform the way modern societies operate and use this critically important and rapidly growing piece of internet infrastructure. The project aims to dramatically improve edge performance and cost efficiency, allowing providers to maximize the utility of scarce edge resources and developers to perform computationally demanding tasks that were previously infeasible. The research developed in this project will be integrated into undergraduate, graduate, and online courses. 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.