Smart Cities - Part 1

By Silicon Labs


Smart Cities. With thousands of urban centers around the globe, which ones can call themselves “smart”? What are the technologies, the policies, the philosophies that are making this technological revolution a reality? And, how do we define a smart city?

In Part 1 of our inaugural, two-part episode, we explore some key technologies associated with smart cities. Daniel Cooley, Vice President and General Manager of IoT Products at Silicon Labs, joins host David Finch to discuss the role of the semiconductor industry in the evolution of the smart city.

“The engineering concepts that work in a cloud infrastructure don’t necessarily work at the far edge node.”

All that to come, but first, the technology.

[DAVID FINCH] I'm joined now by Daniel Cooley, senior vice president and general manager of IoT products at Silicon Labs. Daniel holds an MSEE from Stanford University, a BSEE from the University of Texas at Austin, and he holds four patents in radio and low-power technologies. Daniel, thank you for joining me.

[DANIEL COOLEY] It's great to be here. Thanks for having me.

[DF] So you started your career in IC design and now you're helping to, essentially, shape the landscape of what has become the Internet of Things which is an enormous concept, means a lot of different things to a lot of different people. And I was wondering if you could just tell me a little bit more about the role of the IoT in smart cities.

[DC] So the IoT, in a lot of ways-- actually, it started around the city concepts and smart cities. One of the applications that a lot of people may realize has gone very mainstream is smart metering. And that market has really done a transition over the past 20 years where people are trying to eliminate maintenance crews rolling around in trucks, carrying clipboards, coming out and reading meters into a more automated process.

[DF] At people's homes, right?

[DC] At their homes, at businesses and buildings, but all throughout the city because these cities have to manage that entire network of meters that are out there. And it was safer, it lowered costs, it was more reliable. There's no human element to introduce errors to these systems. And it drove a big wave of innovation around how to connect products and big networks that needed to be reliable.

Billing systems were maintained on these things, so there's money that changes hands. There was opportunity for theft on the meters for electricity, for water, for gas, for other utilities. In some applications, these meters weren't powered. A water meter does not have electrical mains lines to it, so you have to do all this wireless technology on a battery that may have to last 10 to 20 years. So it drove a tremendous amount of innovation around the end-node devices, the meters themselves, the network to connect those devices to infrastructure, and then the maintenance reliability to secure it all, to protect the consumers.

[DF] And so this kind of leads me to something else I've been curious about just because the new buzzwords and the new trending articles are suddenly hitting us all over the place. That doesn't necessarily reflect the age of some of these technologies or, certainly, any of these ideas. And so, from your perspective, what are some of the more established technologies, things that we should be looking a when we're considering smart city or smart home design?

[DC] That's a great question because you're exactly right. A lot of these technologies are now hitting the mainstream. I see it in the news. I see it in places where I wouldn't have seen it five years ago, but they were already around. There's not been that much that's actually been fundamentally invented in the last five years say. And so that's good in one respect because these technologies are somewhat hardened already and they're not going to go through the first phases of deployment and the first phases of working out the bugs so you can put this into critical infrastructure like you might in a city.

At the same, they may not have been invented for the same applications that they're trying to be used for now. And so you have to understand where these technologies bump into their limitation and how the standards will have to evolve to keep up with where the deployments are coming through right now.

So some of the more mainstream technologies that are out there that you might see in the news revolved around the idea of mesh networking. And that's become more popular over the last year or couple of years as you've seen standards organizations like the Bluetooth SIG launch a Bluetooth mesh; or Wi-Fi which just-- the Wi-Fi meshes that you might see in your house from certain products have been proprietary. The Wi-Fi Alliance just launched a standard based on that.

And so mesh is really going mainstream but has been around for a long, long time. And mesh is the technology that allows a lot of networks to really scale into many, many many devices instead of kind of a hub-and-spoke way to interface with those devices. So it's a very different kind of networking technology, but it's one that's coming out much more to the forefront these days.

[DF] And something that we need to see in these types of applications because they are more autonomous, they're self-healing. To have them sort of managing themselves and each other is really just a design imperative at this point.

[DC] Yes. You get a lot of benefits from a mesh. Once you hit a certain critical number of devices that have to connect into a network, the benefits of a mesh really shine through. It is more complex to build a mesh than to build a point-to-point communication interface. And it only makes sense to implement that mesh because of that complexity once you need it. But those networks are all starting to get to that point right now.

And we're seeing the technologies really sweep through. Even some of the more established standards like around metering which is where we started this conversation, we're seeing a new predominance of Wi-SUNs start to come out, whereas a lot of the networks were managed via proprietary serial links in the past. And so you're starting to see a lot more meshing technology come into that landscape, whereas 10 years ago it wouldn't have been there. And those metering companies were trying to transform their businesses away from reading a meter into providing services to municipalities.

The big issue hitting municipalities right now, especially out in California, is the peak loads compared to the average loads of these networks. And so these cities are grappling with how to build capacity for the peak demand and yet leave most of that infrastructure and machinery off most of the time. it's really wasteful from their perspective.

[DF] How secure are these existing mesh network technologies, and is there an evolution that you're seeing towards more native security in these applications?

[DC] That's a great question because I think security, second only to IoT itself, is probably the dominant headline out there. One of the benefits of mesh is the fact that most of those edge node devices in a mesh have a little bit higher compute capabilities than a lot of the point-to-point links. And so from the beginning, they've had the ability to implement more sophisticated security in these standards. And so you start--

[DF] In the nodes?

[DC] In the edge nodes themselves. All the way at the very, very, very end. And so you've seen the mesh networking technologies pioneer more of the AES or elliptic curve key exchanges ahead of where the point-to-point links were so that they could start out from kind of a leg up on the security there.

You know security, as a concept, is really never, 'Is the network secure or not?" It's just, "How expensive is it going to be to hack into it?" So you really try to make it uneconomical for somebody to break into that network. If a nation state chooses to target Daniel Cooley and hack me, they will find a way, right. My hope is just that the hacker sitting at home for fun, there's just no way.

But when we start talking, especially, about people's lives and the data that matters to them and secure infrastructure like in a city, we have to take security pretty seriously. And you're starting to see every standard that's out there adopt, at almost [anomaly?] rate, very exponentially increasing rate. The most advanced security in hardware, in the software, so coding standards. How much third-party software are we using? Is it open-source? Do you really have access to everything that's in your product? All the way to the supply chain. So we're starting to see customer demand for guaranteeing the authenticity and the tamper resistance against these products, all the way through till the time they're installed on a bridge or on a home, and anywhere else.

And the semiconductor providers are all putting this into the hardware. It's not a software-only solution. And they're all racing to just, basically, enable these products with as much security as possible and try to train our customers to use it because a connected device is a more exposed and hackable device. And some of the security concepts are pretty sophisticated. So we all have to work hard to make it simple to use otherwise people and organizations will take too long to actually implement it. And that's one thing we've been worried about with the roll-out of the IoT is just, will our customers who may be smokestack, big iron, metal-bending organizations, be able to pick up very sophisticated embedded technology and actually use it?

[DF] Have you seen real-world examples of being able to participate in energy conservation or energy management initiatives?

[DC] Actually, I have. It's a good example. So I'm based and I live in Austin, Texas. And Texas, it's a deregulated grid. Austin Energy offered to pay me money if they could control my thermostat at certain times of the day. And so, essentially, what they were trying to do in the hot, Texas summers is pre-cool homes.

While I'm not there, maybe around 2 or 3 o'clock in the afternoon and then leave the fans running but the compressor is off in these air conditioning systems so that by the time I get home, my house is already cool or my apartment's already cool and I'm not ratcheting up the demand on their grid because there's a big spike around 4 or 5, 6 o'clock in the evening as people are getting home.

And so it was cheaper for them to use that energy earlier in the day, and they were willing to pay me money. They paid me $85 to do this. And generally, utilities aren't in the business of paying their customers money [laughter]. So it was well worth it to incentivize people so much that they were willing to pay them money to do it. And I think the program's been a big success.

[DF] So one of the considerations, I think, for semiconductors-- and correct me if I'm wrong, but semiconductor manufacturers have to consider that their hardware solution or even their platform solution, hardware plus software, is going to have to play nicely with cloud services. They have to take into consideration where the data's going to live. Can you share a little bit with me about, maybe, why it's important for semiconductor manufacturers such as Silicon Labs and some of your peers, right, to be even considering this larger ecosystem?

[DC] Yeah, that's a great question. I mean, so one of the beauties of the IoT is that there's all these physical things that are out there and they all needs chips. And that's where the semiconductor companies step in. The difference these days is that these chips are then creating data that's ultimately going to flow through and sit on a datacenter server rack somewhere, whether it's in-house, whether it's at Amazon at AWS, or Alibaba in China, or Baidu, or wherever.

And so those cloud companies are very aware that this is happening and that all this stuff is being connected. They're trying to work down-stack to enable this as much as possible. They want to get all that market share, as much as they can right now because they know it's going to get really, really big in the future. And the semiconductor companies are working up-stack to do everything we can to enable that process because it allows us to sell more chips. They get more data, we sell more chips. This is a win-win scenario. It's not zero-sum.

That being said, the engineering concepts that work in a cloud infrastructure don't necessarily work at the far edge node in a door-window sensor. And so we have to kind of clarify the landscape. Where does our product start and their product begin? But I can promise you, every major cloud infrastructure company right now is racing like mad, just like the chip companies, to enable the IoT. And I think that it's great that we're working across this ecosystem, across the entire technology stack to make this possible because that wasn't happening three or four years ago.

[DF] Do you think you might provide some examples of ways that semiconductor manufacturers and these cloud services can work better together or have worked or will have to work better together to make things come together smoothly for everyone involved?

[DC] So one program that was announced recently is Microsoft's Sphere. And so Microsoft - and this is built to support their Azure business - has provided a operating system down to run on Cortex-A class, Linux-based end equipment. Now, this is targeted initially at white goods and they're hoping to spread it out further. But that's kind of where rubber hits the road is it has to be a chip in that system enabling that product to be what it is. Microsoft's not making that chip.

So we have to interface in that operating system layer. And if we do it in a smart way, it's secure, it's cloud-connected with Azure by default. The Microsoft Sphere program is, basically, guaranteeing the security and updatability of the device, but we have to work hard with them to make that possible.

Another example is something Amazon is doing with Green Grass which is an operating system specifically targeted at gateways because, in a lot of wide-scale IoT deployments, the gateway is the most complicated part. And so they're working hard to make that possible. And they took it one step further also by bringing free RTOS which is real-time operating system for embedded devices under the Amazon umbrella to target even lower-class edge node devices below Linux where a lot of the volume in the IoT lives. So those are just couple of examples of how we're interfacing together with these cloud organizations.

And if you want to know where the next, big, marquee problem to be solved really is, it's in managing these networks. So Amazon does not have the product offering to manage the network. They want to make sure the data can get there and live there and be analyzed and all that. Same thing with Azure and the other cloud companies. But right now, everybody's also racing to provide the services to actually manage a network and keep it alive, so to speak, with software updates, with authenticity of the devices, with telemetry information about what's going on down there. And that whole market's up for grabs right now.

[DF] When you start looking at the volume of data that gets generated where you've got public safety in the sense that you've got traffic camera or you have public safety cameras. You have microphones located throughout the city, especially in some of the higher risk areas listening for things like gunshots or glass-break detection.

You're shuttling around massive amounts of video data, monitoring traffic conditions, commuter conditions, that sort of thing, all the different areas where you can incorporate sophisticated technology to, essentially, make our lives easier and give first responders better, more clear information about where they may be needed or what might be needed. That's a ridiculous amount of data to manage. Are municipalities looking at these same cloud services that you're describing, or are they going to be, ultimately, in the business of developing their own proprietary clouds just because of the volume of traffic? Or am I looking at this all wrong?

[DC] It's a really good question. I have my theories on this. I don't have as much first-hand experience. What I foresee, actually, is not necessarily shuttling around vast amounts of data on these networks. And it will be on-demand, is the way to think about it, just in time, for a manufacturing kind of lexicon. And the reason for that is the vast, vast, vast, vast amounts of data is noise, actually, in these systems. And housing that noise and analyzing that noise and being liable for having that noise on your servers and everything are real concerns, actually. And every time you pass that data through somebody else's equipment or through a gateway of some kind is just a chance to see it.

I believe that in the IoT, the model we're going to migrate to is much of the cloud infrastructure descending out of the cloud and kind of on-premises and as close to the edge node or even in the edge node as possible is going to be the way it goes. The data will be there. It'll be distributed.

Just like we've kind of entered the era of distributed computing, I think you're going to enter an era of distributed data where the data is housed, you know where it's housed, and if you need it you can get it. But it doesn't necessarily mean it's going to be stored in your servers, in your cloud. And that way, you can get critical information through the network, analyzed information through a network, and that compute may be done in a local gateway or at the edge node itself. You might see video recognition software into the video camera itself. And it's only telling you what's going on. It's not passing all the raw footage all the way through that network.

That's ultimately where I think IoT is going to go. The smartness is going to descend as far to the edge as possible. It's more energy efficient to compute down there. It optimizes the critical infrastructure of network bandwidth. It's more secure. It's more private. And in many ways, it's more scalable in all of this. So that is just my theory and I see more Moore's Law marching. I see what the fabs are doing. And we're all trying to enable a smarter IoT where not just the computing is distributed the way they're being distributed now, but also the data being distributed.

[DF] Very insightful. I've really enjoyed this conversation. Thank you so much, Daniel.

[DC] Thanks for having me. I really appreciate it.

[DF] And thank you for joining us for our inaugural episode of EE Times On Air. And be sure to check out part two where we look beyond the technology to find a broader narrative around smart cities, one that impacts all of us. EE Times On Air. This program is produced by ASPENCORE. Thanks for listening.

Supplier
Related Products