March 2018: Fog or Cloud? The Answer Is Yes.

A blend of fog and cloud computing architectures helps solve today’s business problems.

It’s no surprise that during the next several years, the number of IoT (Internet of Things) devices and the amount of data these devices will generate will explode. According to BI (Business Insider) Intelligence,, for instance, there will be more than 24 billion IoT devices on Earth by 2020—approximately four devices for every human on the planet. How will companies leveraging IoT manage all of this data? Where will they process it, store it, and analyze it? Is there too much latency wrapped up in cloud computing? Is the cloud too risky? As the industry experienced last year during a rare outage of Amazon’s Web servers, a lot goes down when cloud servers go down.

These are all the questions that companies must answer. In cloud computing, a network of remote servers hosted on the Internet is used to store, manage, and process data.

The cloud offers a vast, elastic resource, but cloud datacenters are many tens and hundreds of milliseconds away from the things they need to interact with and/or control. More simply, cloud computing has been credited with providing low-cost computational power with the use of large banks of centralized servers from companies such companies as Amazon,, and Microsoft,

These companies typically provide cloud-based services of centralized computer maintenance and management at reduced costs.

Interoperability: Seeing through the Fog

How can we improve interoperability? Perhaps the answer lies in the cloud—or, better yet, the fog. Chuck Byers, principal engineer and platform architect, Cisco Corporate Strategic Innovation Group, suggests fog computing, which was coined by Cisco, is having a significant impact on smart cities and it is changing interoperability.

Interoperability, he says, is the ability of two or more systems or applications to exchange information and to mutually use the information that’s been exchanged. He identifies the challenge with interoperability today, the solution, hurdles, and opportunities for the future.

Challenge: There are somewhere between 350 and 900 different platforms. He asks: how are you going to achieve interoperability if there’s 900 different flavors of how you do IoT (Internet of Things) data abstraction, and interconnectivity. If there are 900 different variants, how are you going to pull that together?

Solution: We need to figure out how to come up with a common language, some mechanism where, for example, a smart-city IoT application can take a bunch of data from various sensors and run that through the smart city, he suggests.

Another answer is fog computing. Enter the OpenFog Consortium. “We’re particularly concerned about what it means to interconnect and sort of referee all of this Wild West of interconnected space. We’ve got a taxonomy that kind of blows out how things talk to the network, and the first split is wired versus wireless.” Wired includes things that are delivered on fiber or copper. Wireless could be dived into three different methods: unlicensed spectrum (Wi-Fi), licensed spectrum (cell network), and Li-Fi, {light fidelity) the free space optical communications world.

According to Ada Gavrilovska, associate professor at the College of Computing and the Center for Experimental Research in Computer Systems at Georgia Tech,, the value of cloud computing is undeniable, and it has had transformative impact on every aspect of our lives, especially in the last decade. In fact, Gartner’s,, research suggests that across all industries, companies spend an average of 20.4% of their IT budgets on cloud. However, due to physics or economics or both, along with cloud computing comes bottlenecks. For some IoT solutions, bottlenecks and even small amounts of latency are a big deal.And while the cloud will continue to provide the computational resource required for big data frameworks that build sophisticated models and turn data into actionable information, Gavrilovska believes the problems inherent in the cloud architecture can only be solved with adoption of fog computing technologies.

For the sake of this discussion, imagine if all the processing power noted in the cloud was pushed back to a local centralized end device or computer to manage a particular service or event. This allows for each computer to manage its own IoT data processing. When moving the power computing down to ground this is referred to a fog.

Similar to the first part of this definition, the close-to-the-ground part, fog computing refers to embedding computing infrastructure near the end devices, along the edges of a network. The edge is finite in capacity, its reach is localized, and its view into the data is myopic. “If cloud is about aggregation and consolidation, then fog or edge computing is about distribution and dispersal of computing,” Gavrilovska explains.

Analysts are also predicting significant fog computing growth in the future. For example, Grand View Research,, says fog computing will grow 61.3% through 2025. The software segment is expected to see the fastest growth, while the routers and switches segment is expected to account for more than 30% by 2025. A number of industries will benefit from the growth of fog computing including smart manufacturing, smart grids, and healthcare, to name a few, according to the research.

Conversely, is the edge all it’s cracked up to be? Converging trends may have made fog computing more attractive than ever, but there are clear limitations at the edge. When it comes to building solutions, the question of cloud or fog computing may have architects scratching their heads. But is it really a matter of one or the other, or is cloud the yin to edge’s yang? Perhaps the battle of fog and cloud isn’t really a battle after all. “Both (cloud and fog) provide absolutely essential ingredients for realizing connected world solutions,” Gavrilovska says. “Unlike the real fog, fog computing—or I prefer edge computing—doesn’t obstruct our view and our abilities to reach the clouds. It accelerates them.”

Caught up in the Fog

Gavrilovska says that on a basic level, the ideas behind fog computing—i.e., putting computational resources close to the data sources to accelerate time-to-insight and reduce costs—isn’t new. “What’s different now is that we have a confluence of recent trends which shift the cost of the pain point and the cost of the solution, and it becomes both critical and practical to start turning these ideas into reality,” she says.

Matt Vasey, director of IoT business development for Microsoft,, and an officer for the OpenFog Consortium,, also suggests a number of trends have converged to make fog one of today’s hot topics. “The huge increase in compute and sensing capability at the edge has resulted in an explosion of data,” he says. “The old edge model, where all the data flows north into the cloud, cannot keep up with the increase in data resulting in missed opportunities to extract value. Fog computing provides a new model to process this data, to train new predictive models and operate them near the edge without the requirements to move all of the data to the cloud. The result is AI (artificial intelligence) near the edge that operates with less latency and more reliability.”

“Both (cloud and fog) provide absolutely essential ingredients for realizing connected world solutions.” –Ada Gavrilovska, Georgia Tech

Chuck Byers, principal engineer and platform architect, Cisco Corporate Strategic Innovation Group, explains it this way, “It’s the ability of two or more systems or applications to exchange information and to mutually use the information that’s been exchanged. It’s really about the data model. It’s about the information as the currency of the IoT. We don’t want a bunch of isolated silos.”

Byers stresses, “We want the ability for that information to flow freely, vertically, between the cloud, and the fog, and the edge, and the smart IoT devices.” He goes to explain, “We want it to flow freely horizontally across all of the different vertical marketplaces, different suppliers of equipment, and different folks who might be working on the data and those that want to use that data. We want the data to flow seamlessly.

Cloud, Edge, and Fog: Focus on Business Outcomes

The cloud, edge, and fog are changing how businesses leverage the IoT (Internet of Things), but Michael Morton, chief technology officer and vice president of Dell Boomi, suggests taking a step back and identifying the business outcomes first.

“Everybody seems to want to launch into talking about devices, device data. They launch right into the technology. But the one thing we had to learn as well as talking to these customers is let’s stop talking about technology just for the moment. Everybody’s anxious to be the geek.”

Instead he suggests asking some targeted questions including:

  • What are you trying to do?
  • What is the business value?

Worldwide Public Cloud Services Revenue Forecast
(Billions of U.S. Dollars)

Source: Gartner (October 2017)

For the IIoT (industrial IoT), too, fog solves a real need. Because IIoT devices generate a tremendous amount of data and IIoT applications often require extremely fast response times, the edge offers what the cloud can’t. “Fog computing solves challenges such as latency and network bottlenecks that have slowed deployment of advanced use cases like computer vision, realtime controls, etc.,” adds Vasey. “The fog computing approach solves these issues and creates new capabilities that will enable more device autonomy.”

Paul DeBeasi, vice president and chief of research at Gartner,, says the cost of pushing compute to the edge is going down, and miniaturization, an increase in processing power, and better battery life are all advancing to the point where it is becoming more economically viable to push compute closer to the edge. DeBeasi sees three main benefits of fog computing: autonomous edge operation, reduced end-to-end latency, and compliance with data requirements.

“One of the reasons why you’d want fog computing is autonomous edge operation,” he says. “You’ve got a factory, and it’s got to run and it can’t depend on anyone else or any other kind of processing. Another advantage of fog or edge computing is reducing end-to-end latency. The amount of time to sense something (at the edge), send it all the way to the cloud, process it, send it all the way back, that’s latency. That could take a few seconds. But what if your requirement is you need to deal with events in milliseconds? The latency is too long to go to the cloud. If you’re doing fog computing, you can be much faster, much more responsive.”

Peggy’s Blog

Cloud, Edge to IIoT

I have spent a lot of space reviewing fog, edge, and cloud computing and now for the column I would like to consider one of my favorite vertical markets: manufacturing. Manufacturing is such a great indicator, at least in my opinion, of where we are now with the IoT (Internet of Things) and where we are headed.

Read More

Turning to the Edge, Fog or Cloud?

This column has devoted a lot of ink to fog, edge, and cloud computing. I spent a considerable amount of time last week digging into the basics of fog and edge for those who still needed a better understanding of how these technologies are different.

Read More

Cloud Isn’t Perfect, Is 5G?

It’s time to take a deeper dive into the future of the cloud versus fog computing and how 5G is going to play a significant role. Last month I opened my column discussing cloud versus edge computing.

Read More

Cloud, Fog, and Edge: We Need Them All

For the past few years vendors campaigned heavily telling us all the importance of moving data to the cloud. It was such a hard-hitting effort that even some corporate pitches declared that the best CEOs just might find themselves on the chopping block if they didn’t take the necessary steps to move their data to the cloud.

Read More

As an example, imagine a quality-control solution in a factory. As products move along a conveyor belt, machine-learning algorithms and cameras work together to inspect each product and detect defects in near-realtime. Within a few milliseconds, the system takes pictures and compares them against its algorithm to determine which products can go through and which ones can’t. In this context and others, such as a heart device detecting an arrhythmia or a vehicle operating autonomously, even a small amount of latency is too much.

DeBeasi’s third advantage of fog computing may be less obvious: compliance. “Some countries, like Canada, for instance, have laws where data that’s gathered in country can’t leave the country,” he explains. “And so a factory may be collecting analytical information about the operation of their system, and, by law, they must keep that information in country. In that case, you will probably want to use fog computing to do your analysis and storage right there in the factory as opposed to sending it to a cloud data center that might be in Virginia or California.”

Fog has its disadvantages, though. By keeping compute close to the edge, processing power, processing speed, and storage capability are all limited. “If your edge is a heart device, you can’t put a big processor there,” DeBeasi says. “Or if it’s a drone, you’re limited by weight and battery life. Or even if it’s a vehicle, you’re limited by the weight, the space, the cost—all of those things constrain what you do at the edge in fog computing. So the disadvantages would be those processing constraints, the storage constraints; the more you push to the edge for cost-sensitive things like a drone, it increases the cost, so there’s a cost factor there as well.”

The cloud, on the other hand, offers almost unlimited scalability and processing power. Cloud computing offers inherent simplicity and tremendous flexibility, as well as the ability to spin up storage and compute on demand. There is also a tremendous focus on tools for developers, which has not only created a rapid pace of adoption but also disrupted the way software is built, deployed, and maintained. The two architectures—cloud and fog—are really just two different ways of solving a problem, and, in many cases, leveraging both is the best way forward.

Head out of the Clouds

Increasingly, fog and cloud are complementing each other in the real world. In a factory environment, for example, the realtime aspects of control systems can be pushed close to the edge, while monitoring and long-term analysis could leverage cloud computing. Whether a business adopts fog, cloud, or both depends upon the use case. It comes down to how architects blend the best of both worlds in order to solve their business problems.

Vinay Nathan, CEO of Altizon,, an industrial IoT platform provider, says fog computing has emerged as a technology trend that is complementary to cloud computing, and for the IoT, both fog and cloud are necessary. “Machines generate tremendous amounts of data and (the) vast majority of this data is extremely ephemeral, which can be examined and dropped,” Nathan says. “The fog is necessary as the first layer of quick response to very fast-moving information. The cloud is the repository to which the fog layer pushes data for post-facto analysis.”

Source: OpenFog Consortium

Founding Father of Li-Fi

Harald Haas, professor, chair of mobile communications, University of Edinburgh, joins Peggy to talk about Li-Fi. He says we are moving into the fourth industrial revolution, and we will see environments that are enriched with hundreds of thousands of sensors, and we need a transmission system to send the data to a central processor and into a brain, which would be the cloud that would get all the information.

As an example of how the two layers interoperate, Nathan describes building a machine-learning model that predicts the impending failure of a piece of expensive equipment. The cloud could be used to process, build, and train the machine-learning model before being deployed on the fog layer. “The fog layer runs this model against fast-flowing inbound data from that equipment and flags anomalies and problems,” he says. “It can even instruct the equipment to shut down if critical operational parameters are breached. The fog provides relevant feedback to the cloud about the performance of the machine-learning model and provides newer data sets. The model is continuously improved on the cloud based on feedback and pushed back to the fog for execution.”

Fog and cloud can work together, but it’s up to architects to put the pieces together. Thankfully, they have some flexibility in a model that embraces both fog and cloud. “A good architect will have some fluidity and the ability to move pieces around and not become too rigid in the design,” says Gartner’s DeBeasi.

At the end of the day, fog and cloud are not competitive, they’re complementary. The real challenge organizations face in building IoT solutions is to do the difficult work of integrating a multilayer, multitier system. “That’s why I get phone calls,” DeBeasi says. “(Organizations) want to know: What are all the pieces? Who’s supplying them? What are the interfaces? How do I integrate it? They don’t say, ‘Gee, should I use fog or cloud?’ The answer is yes—use both.”

Want to tweet about this article? Use hashtags #M2M #IoT #cloud #cloudcomputing #edge #edgecomputing #fog #fogcomputing #IIoT #blockchain #data #AI

Guest Contributors

No grid was found for: New Grid.