Building tomorrow’s telecommunications network today

The current 5G evolution in network connectivity is expected to drive unprecedented demands for bandwidth, reliability, and security. However, a network of this magnitude and robustness doesn’t pop up overnight and enterprises and consumers are just beginning to realize the myriad use cases a 5G network can support.

For example, consider the increased number of connected devices in a house like smart thermostats, security cameras, tablets, smartwatches, and mobile phones, of course. Raj Savoor, the vice president of network analytics and automation at AT&T Labs explains, “Currently we estimate the average consumer home footprint has about 13 connected devices, including mobile and other devices.” And although that sounds like a large number, he continues to explain the real scale, “That’s going to increase to 30 to 40 devices over the next five years, so a really big increase.” And the real challenge he continues to explain is that, “This growth needs advanced network architectures to support, manage and provide fast, secure, and reliable services.”

Bandwidth will also increase five times in the next five years, according to Savoor, as consumers adopt immersive interactive applications. Immersive experiences also require lower latency and jitter, and a lot more security and reliability. For a company like AT&T that supports a large existing network, building the next generation network requires an incremental approach. In fact, AT&T’s 5G network has been years in the making. “We look at it as a journey. There are a lot of steps that we’ve taken over the past few years to build on it, and we have prepared for the next step,” says Savoor.

And as businesses and consumers transition to a 5G world, AT&T keeps looking ahead. “We are thinking about the next 20 and 50 years. Network investments take a long time, and we want to make those investments with economics in mind, but also very much ensuring the most reliable network offering,” says Savoor.

Laurel Ruma: From MIT Technology Review, I’m Laurel Ruma and this is Business Lab, the show that helps business leaders make sense of new technologies coming out of the lab and into the marketplace.

Our topic today is architecting networks. From cell tower to device, 5G is helping connect people and spark innovation with a reliable, fast, and scalable network. This means big opportunities for consumers and enterprises.

Two words for you: next, next-generation.

My guest is Raj Savoor, the vice president of network analytics and automation at AT&T Labs. This podcast is produced in partnership with Infosys Cobalt.

Welcome Raj.

Raj: Thank you, Laurel. I’m delighted to be here with you today.

Laurel: AT&T Labs has a long history of innovation. How is it now building that next generation of networks in a cloud-driven, digital-driven world?

Raj: Great question, and it’s something we spend a lot of time thinking about. Every technology turn is different. We’re certainly leaning into this change to 5G networks and a lot of increased fiber penetration in our networks, leaning back on our culture of innovation, of our history and legacy, and particularly adapting to change as we’ve gone through so many generational changes. It is a unique period. We are investing both in other 5G wireless network and at the same time expanding our footprint of fiber optics further into the network, closer to our consumers and businesses.

We are seeing rapid adoption and unprecedented demand for bandwidth from our both consumer and enterprise customers. Usage for example, in the home, is increasing. It will increase five times over the next five years. We saw some of that during the pandemic. The number of devices in the home are rapidly increasing. Currently, we estimate the average consumer home footprint has about 13 connected devices, including mobile and other devices. That’s going to increase to 30 to 40 devices over the next five years, so a really big increase. This growth needs advanced network architectures to support, manage and provide fast, secure, and reliable services.

Our approach is evolutionary. That’s primarily because we are an existing, very large network that has gone through so many generational changes. When we use the next-generation architecture, we have to be cognizant of the existing infrastructure and work that incrementally, so it’s not like a brand-new, greenfield, overlay build. It is that incremental approach, and it’s never a single switch. We look at it as a journey. There are a lot of steps that we’ve taken over the past few years to build on it, and we have prepared for the next step.

Laurel: Some of those steps include making that network more stable and reliable and with great coverage to touch all those households, right?

Raj: Correct. I think the foundation comes with our connectivity. To make sure coverage and reach of the network, be it the radio frequency coverage or the fiber optics to the edge begins there, but it doesn’t stop there. There is a lot of additional elements to managing the reliability as well as attributes of that complete customer experience. We are shifting from just single dimensional views of connectivity to an experiential and secure and how we define reliability in a much broader context. That needs a lot more architectural forethought in how we design and deliver ultimately.

Laurel: What will that shift to 5G mean for innovation for consumers and enterprises?

Raj: Just to maybe step back, in our past generations of change and mobile talk wireless, in the first few generations it was all about mobility and basic connectivity for voice. Then we went to messaging and early data. Then when we went from 3G to LTE or the full generation, it was about speed.

As we look at where we are in terms of where the customers are driving us at 5G, that is all about immersive experiences. This requires lower latencies, which is the actual time it takes for a video to start or a download to begin of your next application, or an interactive session that may be in high definition or in AR or VR mode. It also needs lower jitter, which means lower variability, in that experience. You also need a lot more security and reliability. The security because a personal device is basically an extension of our persona. Many of our lives are entwined in that device. So, privacy information, protection of that device, protection of the data in the device, become equally important over the network channel.

Those are the requirements that our customers are demanding and that can really be met in the 5G network architecture. Besides the connectivity capabilities, it is bringing the cloud-native application platforms closer to the edge of the network. It’s an architecture that does require a use of a lot more automation and infusion of AI- and ML-driven approaches to provide that customer experience that is needed.

At the end of the day, we expect the consumers to be able to consume more of those immersive interactive applications rendered, say, in the cloud. This could be in a venue, it could be in their homes, it could be when they’re using a 5G-powered wearable device. It could be in a connected car running in a smart city.

When you look at all of these different use cases, there’s also a need for seamless transitions from 5G to WiFi, and so on. That is part of the overall network design and thinking that goes into our perspective. Again, security is also paramount. We constantly hear from our customers how important that is, and that is a cornerstone of how we are approaching it. AI- and ML-based approaches allow us to provide these threat analytics and security for that experience.

Laurel: Yeah, that’s a really good point. The network is now much bigger than it ever was. It reaches to the edge, which is every single device that uses a connection, pretty much. When you are on your cell phone at a football or soccer game, you expect to be able to look up scores or watch the latest play. You also tend to expect that real-time access to that information, to those video streams, to the data. That is one example, but we’re also talking about industrial uses as well. My favorite is reaching out to oil platforms or planes or ships. The edge could be so many different nodes that we actually have to broaden our definition of what network means, correct?

Raj: Absolutely. We certainly see the adoption of 5G and the edge technologies beginning at the enterprise. I think the enterprise and industrial adoption then is driving it for the consumer. Different generations of technologies have either begun at the consumer or at the enterprise level. Because of the adoption of edge cloud capabilities, what we refer to as multi-access edge compute, 5G is effectively bringing cloud compute, storage, and analytics capabilities and applications closer to those industrial applications. This is an area that we’ve really focused on for multiple years.

AT&T has a 5G innovation studio where we bring in our enterprise customers and their problem statements and use cases. We bring in startups and other partners to put together solutions to address those blind spots or problem statements in connectivity and applications. Those frequently make use of industrial internet of things. It’s where automation and industrial robots need hyper precision on location, and the network actually enables that.

Also, there’s a lot of video analytics to do assessment of safety issues within locations. The video analytics can be run very, very close to the industrial application and provide that real-time feedback we talked about in the consumer space. It allows quality of service and speed and low latency of 5G as well as security compared to unlicensed spectrum and other network technologies as an enabler for those industrial use cases.

Drones are another emerging area. There’s a need for autonomous control with low latency. Again, the network is an enabler natively for that. Yeah, we do see adoption across various verticals: healthcare, transportation, manufacturing, smart cities—a lot of sensor, network-driven opportunities.

Laurel: One of those examples is FirstNet, a way for first responders to connect during an emergency where other lines of communications may be down.

Raj: Yes, it’s an area we are very proud to support and be the network for our first responders. There’s a FirstNet authority that manages this network. It’s a nationwide dedicated platform purposefully built for the first responders and really the extended public safety community that includes our healthcare system. The mission is fairly unique, as you can imagine, relative to consumers or enterprises that have their needs. We were able to bring all those requirements into a common platform.

It does have an element that is different where, from a mission-critical perspective, there is no higher priority than the public safety mission. The first authority kind of enables this by enabling specific devices that are customized for the FirstNet experience, as well as applications. It actually has an application developer program as well.

One part of the mission that we are highly focused on is the resiliency of that network and the network resources needed not just on any normal day, but when you have that disaster impacting a lot of the infrastructure. In those cases, we have extended our network to take advantage of other resources. We have cell towers on light trucks that are mobile that are placed and then integrate seamlessly with the network. There’s also some early work with drones to provide coverage. We’re not just looking at tactically, we look at it strategically.

Laurel: Part of this need for first responder innovation is because of the changes of climate and the pressures with environmental challenges that are being seen, not just here in North America, but around the world.

Raj: Yeah, our network resiliency is one of those implicit goals for our network design and particularly for the public safety mission. We’ve been looking at a lot of historical data, natural disasters and the impacts, but also modeling for future and modeling in the future risks driven by climate change, where you can have events with a high wind, three-foot floods or higher. And what does that mean for the network? Where should we design and make design changes? Where would we build the next generation of cell towers? And how do you ensure an overall resiliency under those conditions?

So that is an important part of the mission and we’re thinking about network design and architectures. It is really not even for the next three years. We are thinking about the next 20 and 50 years. Network investments take a long time, and we want to make those investments with economics in mind, but also very much ensuring the most reliable network offering.

Laurel: You mentioned artificial intelligence and machine learning in a previous answer. What are some ways that AT&T is using AI and ML, or thinking about deploying artificial intelligence?

Raj: Great question and also a very timely one. As a company, we have had researchers working on AI for many years. With the advent of a lot more compute power and a lot more finer grain data, the opportunity has really opened up with the last, I would say, five years. It does play a very significant role at AT&T. Again, we have approached AI in an evolutionary way on how we infuse it.

First, we think about AI as the engine, and the fuel is the data. It begins with how we want to collect data and learn from it. That’s where a lot of the machine learning capabilities come in. We have been investing in a lot of big data management capabilities over the past few years, ensuring that those are well exposed to our AI engines. Our chief data officer in particular has worked very hard to establish a democratized ecosystem for both the data and AI capability. There’s a step function here in complexity as the amount of data increases, particularly with 5G, and we get kind of finer grain visibility, and we have a lot more intelligent controls to then apply decisions. So, we’re taking those steps in that evolutionary way.

Internally we have many use cases, including how we can use AI for planning, functions, AI for design decisions, but also in real time to help our customers, as well as the network, under various scenarios to provide better efficiency, better customer experiences, detect security threats, the threat analytics, as well as how to use feedback loops to constantly optimize the network. So, a lot of use cases across the life cycle.

Laurel: I’m speaking of that focus on security, which is top of mind for most executives these days. But not only security, AI and automation also are playing that really important role for 5G functionality.  What other ways is that coming into play right now with the capabilities of 5G?

Raj: Again, this is very timely and a very active work area. Let me give you some context on how we are structured. In thinking about 5G, we think about it as day zero, day one, day two. Day zero is the planning activities and forecasting. I can see some natural ways where AI and machine learning can help you through your forecasting. There’s your day one, which is actually building and designing your network. You want to do the greatest efficiency. Again, the feedback loops and reinforce learning kind of helps you do that as well as use of deep learning technology to analyze maps and geospatial data, to determine where you want to have buried fiber optics and where you want to place a small cell versus a macro cell. So, there’s a lot of the building engineering where we rely heavily on AI, deep learning, and neural networks.

Then there’s a lifecycle, which we call day two. In that, there are opportunities, things like energy savings where we are trying to optimize the energy footprint of our equipment. Again, both a corporate priority, but also a societal priority on the carbon footprint. We see great opportunities for economics but also helping the planet.

From a 5G technology perspective itself, there is an opportunity in what we refer to as beamforming. Beamforming is basically optimizing how the actual coverage for consumers is improved to mitigate some of the impacts of fading and path loss. The context-aware beamforming along with what we call MIMO [multiple-input and multiple-output], which is a very efficient way of transmission, requires us to understand where the demand is to determine where the customers who are consuming or using our service are located. We want to be hyper precise in that geolocation to optimize that beamforming. Is the consumer stationary? Is the consumer moving? Is he walking at three miles an hour or riding in a connected car? So that information to guide beamforming is a natural native 5G AI opportunity.

Laurel: That is certainly part of this complex web that companies need to really start thinking about. So, there’s an architectural challenge there to bring together cloud computing, edge computing, 5G, and then a focus on customer experience. As much as you have customers, you’re also quite concerned about your customers’ customers and how they’re experiencing these products.

Raj: Yes. While we have a direct relationship with consumers, in many cases, it is a B2B2C where we have a relationship with a connected car company, and then they’ll have relationship with the consumer. Or we would have the relationship with the transportation infrastructure provider, and they would have the relationship with similar other verticals. So that is inherently one of the opportunities that we are able to drive from this architecture.

One of those capabilities is what we refer to as network APIs. We derive intelligence from our network and then make it available by APIs from that cloud infrastructure to an application platform for them to further optimize for their unique applications and their consumers. It’s an emerging area, and it does require standardization. There’s going to be many steps that we’ve got to mature this. We are pretty excited, and the early results tell us that the overall ecosystem is very hungry for the type of analytics and data to optimize those end user experiences.

Laurel: The cloud has played such a pivotal role for numerous industries to build that kind of resiliency that we’ve been threading through this conversation, but then also drive innovation. When you think about, and you mentioned not just the next three to five years, for the next 30 to 50 years, in a smaller scope, what technological advancements excite you the most? What’s on that horizon?

Raj: First I think we’ve benefited from a few of the industry, I’ll call them, laws. Certainly, Moore’s Law, that we all enjoy increased computing power at lower costs, effectively where we are with growth in storage and cloud capabilities.

Then there’s the other part, which is the demand part—the consumer appetite for increased consumption. Some of it is behavioral. Some of it is autonomous as they adopt devices with higher resolution. Effectively that drives the consumption both downlink and uplink.

Trends that I’m following and I’m expecting to be the drivers for the future is first, like I mentioned earlier, we are expecting usage to increase five-fold in five years. While I don’t have a crystal ball quite beyond the next five years, I don’t see any reason why that wouldn’t continue, particularly bi-directional communication, especially with those immersive VR/AR [virtual reality/augmented reality] experiences. The number of connected devices and devices just around us, whether it’s variables in our automobile, in our home, is going to double or triple. We expect that trend to also drive effectively the quality of our lives and really operating in every phase of our day.

I see sensors increasing, and I think not just in the home, but also in smart cities and in the public domain. And a seamless opportunity between our home, work, transportation, and places that we visit. So it’s going to be increasingly driving that seamless experiences that I’m excited about. It is going to be an exciting future.

Laurel: It sure is. Thank you so much, Raj, for joining us on the Business Lab.

Raj: Thank you, Laurel.

Laurel: That was Raj Savoor, the vice president of network analytics and automation at AT&T Labs, who I spoke with from Cambridge, Massachusetts, the home of MIT and MIT Technology Review, overlooking the Charles River.

That’s it for this episode of Business Lab. I’m your host, Laurel Ruma. I’m the director of insights, the custom publishing division of MIT Technology Review. We were founded in 1899 at the Massachusetts Institute of Technology, and you can find us in print, on the web and at events each year around the world. For more information about us and the show, please check out our website at

This show is available wherever you get your podcasts. If you enjoyed this episode, we hope you’ll take a moment to rate and review us. Business Lab is a production of MIT Technology Review. This episode was produced by Collective Next. Thanks for listening.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.

Main Menu