Fog Computing vs. Cloud Computing vs. Edge Computing

July 29. 2016. 3 mins read
Table of contents

The world of information technology is one where grandiose sounding names often mask just how simple the underlying technologies actually are. In a recent article, we demystified the term “cloud computing” by explaining it as a business model that leases applications on demand which are accessible via the internet. In the olden days this was called centralized computing. For those of us who were alive in the days of mainframes, these were essentially centralized computers that did all the number crunching and you connected to them via “dumb terminals” that had no computing power themselves but simply showed you what was taking place on the mainframe.

As Moore’s Law made computing prices drop exponentially over time, we entered the era of decentralized computing where everybody had a computer (or as it was called, a workstation) that could do small-medium sized tasks. These workstations used a local-area-network (LAN) to connect to mini-mainframes called servers that would do the heavy lifting. Anybody remember those times? Maybe the below diagram will help jog your memory:

Server Workstation
Source: IBM

Then came the era of ecommerce and we moved on to building huge server farms called data centers. From there we decided that the way forward was to build ginormous data centers and start calling them “the cloud”. We then decided that everything needs to “talk to the cloud” because that’s what the Internet of Things (IoT) is about, right? IoT means everything talks to the cloud and all that big data we accumulate makes society more efficient than ever. One problem that we ran into with that plan was a lack of bandwidth.

Solving the bandwidth problem that we’re encountering with “the cloud” is remarkably simple. We simply create something called “fog computing” and insert it between “the cloud” and all our “IoT devices” as seen below:

Fog Computing The Cloud IoT
Source: Cisco

It turns out that when you move a computer within close proximity of the device it’s communicating with, the bandwidth improves. Anyone that uses WiFi on their smartphones is familiar with this concept. “Fog computing” is simply moving some of the computing power closer to the devices so that “the cloud” doesn’t have to be consulted for every little minor detail. Many smaller time-sensitive computational decisions can be made by an intermediary device that then aggregates all the data it learns and sends that up to the cloud. It’s kind of like decentralized computing all over again. In this case however, we’re referring to the devices located in “the fog” (can we say that?) as nodes.

Fog Computing Nodes
Source: Cisco

We have “fog nodes”  that sit near all our IoT devices and then “fog aggregation nodes”, a concept that is pretty much equivalent to how workstations communicated with servers back in the day. See how simple fog computing is?

Now check this out. “Edge computing” is just another name for “fog computing”. Here’s a diagram that shows why:

Edge Computing

See how they just drew a circle with “the cloud” in the middle and then put all the IoT devices on the “edge of the circle”? The computers that connect all these devices to the cloud are referred to as either “edge computing” or “fog computing”. They’re exactly the same thing.

While “fog computing” and “edge computing” are overly simplified concepts that simply rehash ideas that we’ve had before, the real opportunity lies in configuring the “nodes” and optimizing their performance. The primary difference between your IoT device communicating with a node versus the cloud is that bi-directional communication with a node can take milliseconds while conversing in the same manner with the cloud can take minutes.


It’s only a matter of time before we see a slew of startups with value propositions on offer like the one below:

Hosting high performance processing, analytics, and heterogeneous applications closer to control systems and physical sensors, breakthrough solutions can enable “edge intelligence” for closed loop device optimization.

In other words, they’ll help you move portions of your cloud-based applications closer to the devices that use them. It’s not an easy task to figure out which software tasks to decouple from the cloud, but the rapid growth of IoT devices and the bandwidth that they consume demand that we take a different approach to cloud computing. That approach is what we call “fog computing” or “edge computing”. For retail investors, there aren’t any opportunities yet to invest in any pure-play companies in fog or edge computing. Cisco appears to be making a move, but there are also a number of promising startups that may look to IPO like Foghorn


Leave a Reply

Your email address will not be published.

  1. Here is a great comment from our reader Douglas Johnson on the difference between “edge computing” and “fog computing”. While we used research from Cisco, other vendors will define things differently,


    Interesting and nicely done article but Fog Computing has never meant and does not mean, “…take everything out of the cloud and move it “to the edge”.”. Not to be confused with Edge Computing, Fog Computing refers to the collection and processing of select IoT data within an IOT gateway or other sensor data collection point. This is most commonly done for automated decision making, controlling devices, performing analytics, streaming to control applications, or preprocessing data before forwarding or streaming the most important data to the cloud for additional processing or archive. This approach is most applicable in commercial and industrial applications such as factories, distribution centers, processing plants, automotive, department stores, and the like. The primary goals of this approach is to enable additional efficiencies in automation, cost savings, operations, performance, and in some cases to increase data security. I think it is equally important to note that different vendors will define fog computing in slightly different ways to better relate to their own specific offerings. On a related subject, Ryan Pierson at readwrite wrote an article focused on how fog computing differs from edge computing that is definitely worth a read.