In the beginning it was the Cloud. It looked like the solution to massive storage and computation, It was also a perfect support for distributed service architecture with some parts of the service run locally and some part in the Cloud. Then the fog appeared. The availability of storage and computation capability in many “pockets” of the network, some provided by network Operators themselves but mostly by service providers was pushing computation and storage closer to the users. These kind of storage and computation are in general cheaper, faster to get to and to respond since communications did not need to go across the whole network.
In some “interpretation” the fog could actually be provided by terminals. Their storage and computation capacity is individually much lower than the ones provided by cluster of servers but if you take the hundred of thousands terminals that could potentially be clustered in an area you get significant power overall. The problem with a massively distributed architecture involving mobile terminals is, today, the draw on battery power on those terminals making this solution difficult to implement (and accept by users) wherever there would be a significant use, hence a significant draw on power. It remains attractive in the area of using mobile terminals as aggregation points for IoT data but this is clearly, at least so far, a niche application.
A different story is the use of fixed terminals, like home media centres, where powering is not an issue. An example is the one of SoftBank, Japan, where media centres on their subscribers’ premises could be used as well as content distribution points.This solution hasn’t gained traction, moslty because the big content providers like Netflix are using a more centralised content distribution architecture and because a good portion of content remains realtime (like watching soccer games).
More recently we have seen a growing interest of edge computing, both requiring storage and processing capabilities at the edges. The novelty is in the possible allocation of these capabilities on the mobile towers hosting antennas for the wireless network. Towers Operators are finding easier and easier to get funding to expand their footprint and part of these funding can be used to transform a tower into a mini cloud, offering storage and computation capacity to services run on mobile terminals. This is clearly decreasing latency since the network latency goes down to zero. Powering is not an issue, since intelligent antennas require power and the cost of installing local storage and computation capacity has decreased enormously.
5G promises very low latency, but this low latency can usually be guarantee only in the radio part. Recent experimentations by Deutsche Telekom and Nokia have shown that meeting the low latency requirement of self driving cars relaying on the wireless infrastructure can only be met by processing the signals at the edges (towers).
Now, the graphic in the figure presents a quite clear partitioning across Cloud, Fog and Edge in terms of functionality each one can deliver to the maximum effect: the Cloud is fit to store long latency data, aggregate them and perform data analytics, the Fog is good for real time processing and low latency data whilst the Edge is better suited to capture interactions and activate feedbacks. However the graphic representation has two problems. First, it is not clear at all where the boundary is located, second it gives the impression that the Fog is between the Edge and the Cloud.
With regard to the first point what we will be seeing is that as technology, and its deployment evolve, the functionalities will drift from one to the other. Services that today are residing in the cloud might move in toto or partially to the Edge or to the Fog. Meta-data, resulting from data analytics may be created at the Edge, in the Fog not just in the Cloud. Low latency storage may drift between the Fog and the Edge.
The forecast that was made last year, seeing the speeding up growth of the Edge, to have it diminishing the role of the Cloud did not actually materialise. The drift of functionality among the three areas is actually occurring but at the same time we are seeing more exploitation of each one, to the point that each one is actually growing.
Hence, although I see that a number of services will progressively migrate (ending up in terminals for the most part) as well as their related data (this is going to be one of the migration forces, privacy protection) I see a coexistence of the three areas.
Considering the second point, the Fog cannot be represented as a layer in between Edge and Cloud. The Fog extends into the terminals, so in this case the Edge will be in between the Fog and the Cloud. Today this is not the case but in the future my bet in that the largest Fog will be residing in terminals. Most probably the Fog will end up being a virtual layer embedding both Edge and Cloud and using these as physical resources whilst many services will interact, virtually, across the Fog.
Also, I feel that a good portion of the future 6G architecture will be Fog-based.
Looking forward to your comments!