mist over a forest
PHOTO: Dustin Scarpitti

Buzzwords, even when they hold some grain of truth, are typically met with suspicion. Consider that Oracle founder and CTO Larry Ellison once dismissed "cloud computing" as a bunch of marketing hype. Here is what he said in 2008:

“A cloud is water vapor …. What do you think Google runs on ... water vapor? The cloud is databases and operating systems and memory and microprocessors and the Internet! I don't understand what we'll do differently in light of cloud computing, other than change the wording on some of our ads. We'll make cloud computing announcements because if orange is the new pink, we'll make orange blouses. I'm not going to fight this thing.”

We all know how that turned out.

What Is With All the Water Vapor?

Here we are almost 10 years later with some experts predicting edge computing will eclipse the cloud. Still others insist fog computing is (already) bigger than edge computing. Still other terms like “mist computing” and “dew computing”  have been taken seriously enough that the Institute of Electrical and Electronics Engineers has published papers about it. And we would be remiss not to note cloudlets are also apparently a thing.

If all of this is makes you want to turn your windshield wipers on, or take cover in your data center (if you still have access to one), relax. We've talked to, or heard from, thought leaders from universities with the best computer science programs as well as Constellation Research, Dell Computing, Microsoft, OpenFog Consortium, Vapor and VMware among others to find out at, a high level, what some of these terms mean and why they do — or don't — matter.

Defining Edge Matters

"The planet as a whole believes that Edge is real," said Mahadev "Satya" Satyanarayanan, Carnegie Group Professor of Computer Science at Carnegie Mellon University in Pittsburgh. "The emergence of edge is a term widely used to describe the movement of compute away from the cloud provider and closer to where the data is being created."

It is important to note that "edge device" and "edge computing" and "edge cloud" are not interchangeable terms as they are not the same thing.

An edge device according to Matt Trifiro, CMO, Vapor IO is "on the downstream side of the last mile network. It is literally a 'device' in the field, whether it is a car, a gateway, a sensor, a smartphone, a computer." 

"It has none of the characteristics of cloud: i.e., it's not elastic, scalable and the end user tends to have to own it and manage it. Again, think of your smartphone or your Nest thermostat," he said.

Device to cloud communications are too slow for newer applications, according to Trifiro. Bad things can happen, or opportunities can be missed in the time that it takes for information to travel from the edge device to the cloud and back again.

Constellation Research vice president and principal analyst Holger Mueller cites a windmill as an example. "When local software shuts down a windmill because it risks breaking due to excessive wind speeds, that is edge computing. Edge is often motivated by moving the code to the data because transferring the data is too slow or too expensive," he said.

Edge cloud, on the other hand, extends the (private and public) cloud to the edge of the network on the upstream or cloud side of the last mile network, according to Trifiro. "This 'edge cloud' behaves like a centralized cloud (e.g., it's scalable, on-demand, not managed by you, separate from the device, etc.)," he said.

Edge cloud is a "must have" in cases like autonomous driving, which, according to Peter Levine, general partner at the venture capital firm Andreessen Horowitz, is like a "data center on wheels."

Unlike a windmill, an autonomous car needs to take in, correlate and analyze data from a large number of sources, some of which are sensors inside the vehicle (think brakes), some of which come from satellites  (GPS, Lydar, weather data), some from cameras that see other vehicles and pedestrians and more. All of this data needs to be brought together and processed so that decisions can be made instantly. Traveling to the cloud is too slow and, in a case like a small child wandering into traffic, could cost lives.

So, companies are increasingly going to edge computing for efforts like these. Mimi Spier, vice president, IoT at VMware explained it this way, “’Edge Computing’ is a flexible cost-effective architecture that allows for near device analytics and decision making. (At VMware) we believe that Edge Computing has a Micro Data Center aspect for some customers. This aggregates 'edges,' like gateways and embedded edge systems, in things and brings additional opportunity for decision making, security, storage and resiliency."

Spier noted that the cloud (Azure, AWS, Google Cloud) will continue to be relevant but will typically assist, "in less time sensitive activities and provide access to more resources in decision making."

What Is a Cloudlet?

Satyanarayanan and his colleagues at Carnegie Mellon introduced the concept of "cloudlet" or "data center in a box" to handle applications where speed, trust and survivability are requirements. Cloudlets sit between the cloud and the device. Because of their proximity to the device, cloudlets can process data more quickly. Regardless of what AWS, Google or Microsoft might promise, cloud speed is only so fast, so proximity matters, according to the argument.

But that's only one benefit. In the cloudlet model, data emanating from things like high bandwidth sensors is analyzed locally on cloudlets and only metadata and a smaller amount of information is transferred to the cloud. This may improve privacy policy enforcement because it can handle data locally and only compliant information moves on to the cloud. Finally, cloud outages can be masked because in cases that the cloud becomes unavailable, cloudlets can temporarily mask the problem. Satyanarayanan wrote more about cloudlets in "The Emergence of Edge Computing" (pdf).

What Is Fog Computing?

Depending on whom you ask, fog computing is either the next big thing or a "clever" marketing pitch created by commercial IT vendors (ARM, Cisco, Dell, Intel, Microsoft) to promote a "commercial way to use distributed resources as one system." 

"It's a reason/ excuse / sales argument to buy more on premises gear, as you can use it all together," according to Mueller.

Thought leaders from Carnegie Mellon, Stanford and MIT, who asked to remain anonymous, all independently said edge computing and fog computing are the same thing when asked to explain the difference between the two. 

"There is no fundamental difference," one of them said. Another said, "Fog is an attempt by Cisco (which coined the term fog computing) to grab mindshare and be cute."

Needless to say, the members of the OpenFog Consortium disagree. The OpenFog Consortium's goal is to evangelize fog and to help "to drive reference architecture(s) to serve as a foundation for fog computing standards,” according to Brent Hodges, head of IoT planning and product strategy at Dell and an OpenFog Consortium officer. He noted that this includes working on the complex interoperability requirements that will lead to these standards.

Matt Vasey, director of IoT business development, Microsoft and an OpenFog Consortium officer, told CMSWire that fog computing is a framework that encompasses everything between an endpoint (edge) and the cloud.

"Edge is part of fog," he explained. "Fog starts at the 'thing' (from IoT) and moves up to the cloud."

Vasey also noted that edge does not compete with fog. "Edge is to fog as apple is to fruit," he said, quoting Helder Antunes, chairman of the OpenFog Consortium.

"To truly be transformative, the Internet of Things (IoT) needs to be able to make use of computing architectures wherever analytics are needed from smart things to the cloud continuum," said Hodges, adding that, “Dell Technologies sees fog as the next evolution of computing where billions of smart things generate data to connect to analytics at the edge, to core IT systems and to the cloud, which will radically transform our customers businesses. Our customers expect us to provide leadership in this new computing environment which helps them to solve problems they see in the real world every day."

Marketing ploy or the next big thing, one thing can be said for certain: both edge and fog computing enthusiasts want to solve the same problem, namely getting the right information to the right place in the right formats and context so that decisions can be made, and actions can be taken in real time.

What’s in a Name?

What you call the framework, architecture or process does not really matter, according to Lynne Canavan, vice president of Marketing for the OpenFog Consortium. 

"The reality is that customers don't care what it's called, they care what it does. What matters is the need to solve the technical challenges that will enable these massive data-intense applications to operate in complex environments as designed — securely and with sub-millisecond latency," she said. "OpenFog members are working on the complex interoperability requirements that will become industry standards to make these applications work."

Whether OpenFog does that, or if it is somebody else, “someone needs to figure out how to automatically distribute workloads across the gradient,” Trifiro told CMSWire.

So, whether you call it edge computing, fog computing, or both, it is what's next.