Is Edge Computing Key to the Internet of Things?

Scientists say moving processing power to the edge of networks solves tough problems. Practitioners say it’s too early to tell.

by / July 14, 2015
Mesa, Ariz., is enlisting the Internet of Things to get a detailed view of traffic in the city, said Manager of Technology and Innovation Alex Deshuk. Mark Lipczynski

Some say edge computing will be the brains behind the Internet of Things. Others aren’t so sure.

Like its name implies, edge computing pushes computing power to the edges of a network, so instead of devices like drones or smart traffic lights needing to call home for instructions or data analysis, they can perform analytics themselves on streaming data and communicate with other devices to accomplish tasks. Researchers contend edge computing will allow systems to degrade gracefully, work autonomously and deliver information to decision-makers faster and more efficiently. Practitioners say they’ll wait and see.

Ryan LaMothe, research scientist at the Pacific Northwest National Laboratory, has worked on edge computing for the past four years. No one talks much about it, but it’s coming, LaMothe said. The transition to edge computing is subtle. But slow shifts in technology are frequently the most pervasive and have the greatest impact on society.

Edge computing would solve many of the most difficult problems facing robotics and computing infrastructure. A swarm of air- and land-based drones examining a remote forest fire, a collapsed building or a vast tract of farmland is today challenged by an inability to connect and transmit large quantities of data over wireless networks or to receive instructions from a central controller in a timely fashion. These problems are exacerbated by the confusing terrain of disaster environments, but edge computing circumvents these obstacles.

“You don’t necessarily have the bandwidth to send the compute or the data back to a cloud system, so the compute needs to happen on the devices,” LaMothe said. “They need to be able to figure out who is in their area and they need to understand the context of their mission, and then take that data and send it to the human in the field who needs that data right at that time. It’s ultimately to make the human emergency response significantly more efficient.”

Hiding in Plain Sight?

As a trend, edge computing tends to fly under the radar because it’s not necessarily a discrete technology. Instead, mobile devices, the Internet, the Internet of Things (IoT), health care, public safety, personal vehicles and public infrastructure are coalescing into a dervish that merely brings incidents of edge computing along for the ride.

Cities like Mesa, Ariz., are experimenting with new sensors that more intimately connect city infrastructure and citizens’ personal property. Alex Deshuk, Mesa’s manager of technology and innovation, detailed a new traffic pilot program that doesn’t just count cars, but also would show the city traffic patterns and where each car is going. By detecting the radio frequencies transmitted by a car’s OnStar system or satellite radio device, along with any smartphones or tablets in the vehicle, the city hopes to take a more granular view of what its traffic looks like and where individual cars are going.

Mesa also uses RFID tags to track assets, like expensive public safety radios. “As those become cheaper, more connectable and more powerful in edge computing, they can do more than just count something,” Deshuk said. “They become integral to the function of the devices.”

Palo Alto, Calif., is another city with early IoT projects that may gradually push more computation to the edge of the network. A home energy smart meter program allows citizens to access their utility information online. A smart grid pilot in one neighborhood intends to identify usage patterns and improve energy use. A parking space sensor pilot could reduce time spent driving the car while looking for a space. Most recently, the city launched a $3 million smart traffic signal project that will enable traffic lights to integrate with connected vehicles, potentially enabling a future in which people don’t need to sit waiting in their cars at a desolate intersection at 2 a.m. for no reason.

But while new IoT applications are sure to continue sprouting like weeds, the future of edge computing is less certain. “This is so early,” said Jonathan Reichental, CIO of Palo Alto. “We don’t even know where the value is going to be gleaned.”

Ability to Infer

Scientists at the Pacific Northwest National Laboratory are working on two main challenges surrounding edge computing, LaMothe said. The first is around a field of study called inferential controls, which is one of the core components needed for successful edge computing. As the name suggests, inferential controls are the capacity of a device to infer things about its environment and communicate with infrastructure controlled by other entities, a task that humans accomplish every day without much thought.

IoT projects are predicted to move more computing to the network's edge in Palo Alto, Calif.

Anyone who sees a stampede of people running down the street, for instance, would immediately know that something unusual is happening. If the people were wearing paper numbers and running shorts, the observer would conclude a marathon was under way, whereas if people were screaming and there were helicopters overhead, the person might infer that there was a fire or a crazed shooter, depending on the presence or absence of other observable phenomena like smoke or gunfire.

The trouble today is that most devices are designed for one purpose only. “If you’re asking a traffic system — because Fukushima happened — where the heck all the cars are going, that system’s not worried about Fukushima,” LaMothe explained. “It’s worried about how traffic works.”

Bringing inferential capability to traffic lights or other pieces of infrastructure is no small task. “We’re talking about a highly chaotic system,” he said. “We don’t have all the answers to those problems.”

Another big area of research is figuring out how to remove humans from the loop while ensuring that people can leverage the Internet of Things in a controlled and constructive fashion. Much like the Internet, the IoT is expected to take on a life of its own, controlled by no single entity, but because of its immense power will be in great need of oversight.

The National Security Telecommunications Advisory Committee recognized the security implications of rushed IoT implementations in a report to the president last year, which concluded that “there is a small and rapidly closing window to ensure that IoT is adopted in a way that maximizes security and minimizes risk. If the country fails to do so, it will be coping with the consequences for generations.”

LaMothe predicts edge computing will take hold in the next five to 10 years and that its proliferation will be shaped by government policy. Just as doctors have had computers for the past two decades but only recently started adopting electronic medical records, edge computing may depend on the privacy and security framework formed around devices as they join today’s next-generation networks and unleash powerful new functionality.

More Value

For cities, edge computing could unlock the true value of instrumented infrastructure, said Pete Beckman, senior computer scientist at Argonne National Laboratory. The need for operational intelligence is driving city leaders to deploy sensors on roads, bridges and other physical assets. But in 20 years, today’s “dumb loggers” — sensors that merely log metrics like temperature or weight — will seem as antiquated as an abacus, he said. Adding processing capability to those devices lets them act in real time based on the information they collect.

So instead of gathering data for analysis later on, traffic light cameras could analyze the data as they collect it and make immediate decisions to improve the flow of vehicles. Furthermore, traffic lights that can make their own decisions aren’t hamstrung by network outages or limited connectivity.

“We’re moving the algorithm to the data, not the data to the algorithm,” Beckman said. “And that’s because we now have really cheap, low-power processing to make that possible.”

There’s also new software to support computing on the edge. Beckman and his team recognized a need for an edge computing platform and developed one — it’s called Waggle and it’s used by an IoT pilot project in Chicago called the Array of Things, a 500-node network to explore how such device networks can make cities smarter.

As infrastructure becomes more intelligent, cities will become better equipped to handle incoming generations of self-driving cars, which will employ edge computing by necessity, lest the streets be filled with the twisted metal of vehicles that momentarily lost their Internet connections and didn’t know what to do.

Edge computing also might reduce challenges posed by the growing use of police dashboard and body cameras, which are poised to generate gigantic volumes of audio and video recordings that will tax storage infrastructure and bandwidth. Edge computing cameras could analyze video feeds on the fly and only send home relevant data when needed.

Pedestrian counters and bicycle-sharing sensors are another popular city application where edge computing can transform data held by city IT managers. Instead of a few figures that show how many times the service was used, these devices could create real-time maps that show how fast people are traveling, what routes they take and where they stop.

IoT Takes Shape

Edge computing doesn’t make sense for every use case, but sometimes it can bring an IoT implementation to life, said Miguel Gamiño, San Francisco’s CIO.

Transportation will drive edge computing in cities, Gamiño predicted. Dynamic parking meter rating, traffic routing and public transit optimization, along with public safety, are obvious projects that cities will pursue first.

Los Angeles proved that dynamic traffic management works, improving traffic efficiency by 14 percent as it monitors its nodes in real time and communicates with municipal buses to ensure mass transit stays on schedule, said Peter Marx, chief technology officer for the city. “L.A. probably has the largest municipal management traffic system in the country,” he said. “We have 4,500 different intersections with stoplights wired to a central computer, and in addition we have some 56,000-odd loop detectors in the streets, which provide real-time traffic conditions across the entire city.”

Boston is another city pursuing traffic and parking implementations in its IoT infrastructure. The city has pilots around smart parking space sensors, smart trash compactors and parking meters with connectivity to other sensing platforms. The city held a hackathon in April in which one team examined the relationship between the smart trash compactors and nearby restaurants.

“That’s the kind of question that when you start to combine multiple data sources together you can not only react more efficiently, but you can plan,” said Boston CIO Jascha Franklin-Hodge. “There’s no point of having a smart trash can just to say the trash can’s smart. Who cares? But if it can make the city work better, then that’s a worthwhile investment.”

Boston is expanding its sensor network in other ways too. Through data-sharing agreements with Uber and Waze, and a growing network of sensors on city vehicles, it’s learning more about traffic patterns and other logistical data that can enrich life for residents. Just as anyone can search Google for an address, other types of data, like where people can park easily or the location of litter problems, can be added to the tome of public knowledge.

But Franklin-Hodge doesn’t see edge computing as an essential piece of the IoT. In fact, he said, as IoT applications explode in number, edge computing will be marginalized. “I think edge computing is highly overrated,” Franklin-Hodge said. “There are some very specific use cases where edge computing is the antidote to not enough bandwidth and not enough connectivity. What cloud infrastructure has taught us over the last 10 years is that centralized, high-efficiency computing infrastructure in most use cases is going to outperform distributed, lower-efficiency systems in price, performance, scalability, resiliency and all the other things we value. I think a lot of the use cases of edge are going to fall off as we build more robust networks.”

And regardless of how pervasive edge computing actually becomes, reliable networks remain a necessity as cities build out their infrastructure and develop the beginnings of an Internet of Things. An infrastructure to support all these devices and sensors is essential, said Reichental of Palo Alto, which currently is expanding its fiber network. Although edge computing may offload some tasks, there will still be a need for broader data aggregation and analysis that’s beyond the scope of local computing.

Whatever combination of edge and cloud computing ends up enabling new generations of infrastructure, everyone seems to agree that new ideas and new ways of living enabled by technology will continue being revealed for decades and centuries to come.

“The Internet of Things is a new way of doing business more than it is a product or anything specific,” Franklin-Hodge said. “It’s a way of thinking that starts with this deep and rich set of data about the objects in our world and poses the question of: With all this knowledge, what do we do differently? How do we build our governments, how do we run our operations, how do we provide services in a way that takes advantage of this knowledge?”

Colin Wood former staff writer

Colin wrote for Government Technology from 2010 through most of 2016.