On Processing and Communications in Fog Computing

The term "cloud computing" has become familiar to the general public in recent years. It refers to the shifting of computation from local devices to a remote computing device, typically in a data centre owned and operated by a third party. In a consumer context, the main benefit to users is that they can benefit from greater capacity and performance (implemented in the cloud) than their local device offers. Moreover, cloud computing resources are scalable in response to need. The flexibility and convenience are attractive.

There are however drawbacks associated with cloud computing. It is not always desirable to send data to a remote location; there may be security and privacy concerns. In certain applications, the latency associated with the transfer of data back and forth is an issue. And further, with the rapid expansion of wireless endpoints (and resulting pressure on bandwidth and infrastructure), is it necessarily helpful to transport all of this data back to some far-off location for processing?

Somewhere between the two extremes of local and cloud computing, another idea has emerged: "fog computing". This provides a compromise, wherein the local network includes one or more "super-nodes" that are equipped with a richer set of resources, and more processor-intensive tasks can be transferred as appropriate within the network - avoiding the need for external connections to "the cloud". This architecture is relevant to the Internet of Things (IoT), where low-cost nodes with modest resources may have occasional requirement for demanding tasks.

The realisation of fog computing has its own challenges, and this research project aims to identify and address them. As fog computing is a relatively new concept, there is much scope for further work.

The initial research questions (and thus objectives) can be identified as:

1. In the IoT, which applications would benefit from (or be enabled by) a "fog" architecture, and what would their requirements be?
2. How can data be sent privately between nodes in a fog network? Is it possible to design wireless communications such that messages are obfuscated from possible snoopers?
3. With possible congestion in the available wireless spectrum, can Dynamic Spectrum Access (DSA) techniques be incorporated to improve communications performance?
4. What protocols are necessary for nodes in a fog computing network to work together effectively?
5. How best can the super-nodes be implemented, to offer "processing as a service" within a fog computing network? (This aspect of the research will focus on low cost Xilinx System on Chip embedded devices.)
6. What steps can be taken to ensure the physical security of a fog network (e.g. preventing nodes from being tampered with)?

In researching fog computing, there may be other challenges that emerge, and these will also be considered where appropriate. It is intended that a small fog computing network will be created as part of the project, and used to evaluate some of the methods investigated. This could also be a useful demonstrator for engaging others in the work.