The past 15 years have seen the rise of the cloud, along with a rapid increase in Internet backbone traffic and more sophisticated cellular core networks. Some of the responsibilities of these: (1) data centers, (2) backbone IP networks, and (3) cellular core networks, are now descending to be among, or near, the end users, i.e., to the edge of networks. Fog networking is an architecture that uses one or a collaborative multitude of end-user clients or near-user edge devices to carry out a substantial amount of storage, communication and management. Engineering artifacts and applications that reflect such an architecture include 5G, home/personal networking, and the Internet of Things (IoT). It has thus become both feasible and interesting to ask the question: "What can be done on the network edge?" Can it carry out a substantial amount of storage (rather than storing data primarily in large-scale data centers), communication (rather than routing traffic always through the backbone network), and network measurement and control (rather than controlling primarily at gateways like those in the LTE Core)? Potential benefits of fog networking include real-time processing, client-centric objectives, pooling of local resources, rapid innovation with affordable scaling, and feasibility to operate on encrypted and multipath traffic.
There is no shortage of challenges in fog networking, including the following to be tackled in this project: traversing the boundary between centralized and distributed system architectures, steering the global behavior caused by collective client actions, incentivizing client participation, and using redundancy to achieve resilience on the network edge. Among a large and diverse set of topics in fog networking, this project focuses on two themes: (1) Client-driven measurement and inference, including real-time inference of network congestion conditions. Clients can combine local measurements from multiple sources to infer congestion in real time and use these insights to, for instance, optimally preload content at uncongested times. Implementation is achieved by turning client-side SDKs from an app development tool to a network control element. (2) Client-based control and configuration, including control of network connectivity and secure storage. On a fast timescale, client/edge devices can actively optimize their network connectivity by switching between heterogeneous networks, while on a longer timescale, they can relieve network congestion by selectively throttling their data rates. Beyond network connectivity, client-based software can also enhance storage security and reliability by scrambling, shredding and spreading data to different storage spaces.
Broader Impacts: Among this project's industry collaboration, curriculum development, mentoring and outreach are the following highlights. (1) Fog Consortium. Fog networking has the potential to tip the balance of power in the overall IT ecosystem. The PIs help create a Fog Industry-Academia Consortium, which will promote the ideas of fog networking and host outreach events and internship matching open to academia and industry. (2) MOOC. The PIs will offer a new course on "Fog Networking and the Internet of Things," both online and in in-person classes. (3) Community outreach. The PIs will maintain a research website, including a database of papers on fog networking, that is currently hosted at fogresearch.org.
