In the Internet of things (IoT), is there a difference between edge and fog computing?

By now we all understand the cloud computing paradigm in which end devices (anything from sensors and smartphones to industrial robots or cars) connect to highly scalable computing facilities run by someone else and accessible through classic Internet protocols. It is also easy to grasp that for applications with a strong demand for real-time, low-latency, high-bandwidth, or very secure communications, sending each request up to the cloud and waiting for a response will not provide a satisfying user experience.

Today’s most prominent architectural pattern to overcome this shortcoming is to move computation (and storage) closer to the end devices, toward the so-called edge of the cloud or the network. So it seems quite natural to call this architecture “edge computing” and the actual systems performing the computation in proximity to the end devices “edge nodes.”

The paper also subscribes to this view, albeit defining it as “fog computing.” Consequently, it faces insurmountable difficulties when trying to separate the fog from the edge, arriving at extraordinarily strange if not flatly wrong characterizations, such as edge computing “is located in end devices” or unable to support “multiple IoT applications.”

Amusingly, in many places, the actual technical content openly hints at the true differences, for instance, when talking about software-defined networking (SDN), network function virtualization (NFV), cross-node management capabilities, and various prototypical fog applications. The authors simply (and utterly) fail to correctly analyze their own data and to formulate a correct definition consistent with it. This is also egregiously evident in the two abysmally trivial architecture diagrams and the dreadfully incoherent definition attempts of various terms (fog platform, fog layer, fog node, edge node, end node).

Disregarding this, and also discounting the fact that the paper is too long by some 30 percent, the reader will find a lot of features correctly attributed to fog computing that are typically absent in pure edge computing. If you do not want to embark on this search and rescue mission, I would recommend looking up the National Institute of Standards and Technology’s (NIST’s) “Fog computing conceptual model” instead [1].