First of three parts: What it is; what’s needed to make it work; why Moore’s Law doesn’t always apply; how it will be built; when it will ramp up; and who’s going to benefit and reap the profits.
By Ed Sperling
System-Level Design sat down to discuss the Internet of Things with Jack Guedj, president and CEO of Tensilica; John Heinlein, vice president of marketing for the physical IP division of ARM; Kamran Izadi, director of sourcing and supplier management at Cisco; and Oleg Logvinov, director of market development for STMicroelectronics’ Industrial and Power Conversion Division. What follows are excerpts of that conversation, which was held in front of a live audience at the recent GSA Silicon Summit.
SLD: What will the Internet of Things really going to be?
Heinlein: Everything that can be connected to the Internet will be connected. Just think about all the data that exists around you, whether it’s in your office, your home or your car. There’s tons of data that isn’t being collected, analyzed or delivered to you in a beneficial way. Because of the trends in integration, scale and innovation of deeply embedded products, we can add value and begin to connect devices that were never connected before.
Guedj: When people think about the Internet of things they think about connection to the Internet. We’re going to go from billions of connected devices to tens of billions of connected devices. Therein lies a great opportunity for all of us to capitalize on that. One topic we’re not discussing is that the only way this is going to work is if it’s seamless to the user. That involves a lot of adaptability, whether it’s in control or signal processing, and a lot of hiding of things—the software complexity is very high—while also keeping a small footprint. The only way we will get pervasiveness is if consumers can seamlessly access any kind of elements in their home or their bodies, anytime and anywhere.
Izadi: The superset of this is what we call the Internet of Everything. It’s about process, people, things and data. The things piece involves the objects that will get added to the Internet. There will be 37 billion new devices by 2020 on the Internet. But if you look at the other aspect, which is the people piece, the data piece and the process piece have to come together to make this a reality. What’s important is making sure there is seamless communication, intelligence and the handshaking that has to happen to make sure this is seamless and cohesive.
Logvinov: The Internet of Things is the singularity of us and machines. It’s us becoming smarter, as well as our environment becoming smarter about us. It’s about making better-informed decisions, as well as being more precise about what we need without us even understanding how it happens. Data analytics that will be digested in the embedded intelligence at every point of what we touch and what touches us is what the Internet of Things is about.
SLD: As an industry we seem to be focused on pushing down to the next process node, but the Internet of Things utilizes older processes. Where is the opportunity?
Heinlein: There will be a need for servers and infrastructure to connect all of this stuff, but as we get to the edge, which is what people think about—the things that touch us or our environment—nodes are irrelevant. It’s more about 10-year battery life on the coin cell. Those things are going to be done on 180nm ultra-low-leakage libraries or thick-gate oxide libraries that run at much lower speeds but where the leakage is in pico-amps. This will consume many different levels of the process technology family. We’ve been innovating at 180, 90 and 65nm quite actively.
SLD: This is sort of a rebound effect, isn’t it? We’re taking technology that was developed at the leading edge of Moore’s Law and rolling it backward.
Heinlein: That’s right. When 180nm came out, people didn’t do power gating or the kinds of power techniques that are now de rigueur in design. We’ve put power management capabilities, power gating and leakage control in EDA support, as well as in the IP itself, so people can do power management in these older nodes in a new way.
Guedj: It’s not so much about performance anymore. The analog component will play a bigger role. Can we move the technology on that end? Mixed signal will play a much more important role. It’s the integration of all of these pieces that are not all in the digital domain, and how do you standardize all of that IP so that there’s an abstraction layer you can build on top of. That’s going to be the challenge.
Logvinov: When we say performance that also means integration of multiple functions together. That’s more important than just the CPU inside of a device. One watt per year costs us about $2 these days, and that price will rise. It’s not just about how fast it is. It’s also about how energy-efficient it is, and whether we’re conscious about what we’re transmitting—and that we’re only transmitting when we need to.
Guedj: You do need fat pipes to the cloud, and those require extremely aggressive geometries and high performance. The consumer will not tolerate latency. We’re all used to clicking the channel and being able to change when we want to with the click of a button. At the same time, there is integration with the local devices, which will have analog and digital interfaces. There the challenge is low-power intelligence, which is not trivial. With some of those sensors you need to be able to aggregate the information. You may be able to send some information to the cloud and have the cloud get back to you with an intelligent response, assuming that you have a good connection. You need that level of intelligence and programmability, which is what we’ve been working on for the past year or two with optimization capabilities. Also, you need a somewhat seamless passage from the local device signal process to the cloud, and that boundary may move depending on what network connection you have and where you are, so you need to balance those factors.
SLD: Who’s going to reap the profits?
Guedj: Where there are opportunities and challenges, there is money to be made. Whether it’s in the ultra high performance or programmability, there are challenges in the IP, in the tools, you need the power savings, you need the physical layer IP—all of that has to come together.
Heinlein: If we do it right, everybody will gain. There are problems to be solved at every layer of the stack, from the high end in the cloud all the way down to the implementation and the software and services. There will be new application streams developed. We’re working on a platform called mbed, which allows rapid low-cost prototyping. You can buy a $12 board, plug it into your computer, go on the Internet and within an hour develop an application that you can begin to prototype. That means there’s a whole cottage industry with new services and concepts. There will be a potential revolution at every layer.
Guedj: Companies that nurture a strong ecosystem have a very high chance of success. On one side they can work with OEM partners to address customer needs. On the other side, they can address the technology. The challenges are different, but the opportunities are huge. Data I’ve seen suggest this is a $14 trillion opportunity over 10 years. That’s across the entire ecosystem, but there are a lot of players that can reap benefits.
Logvinov: Who will profit is a loaded question. If you look at semiconductors they’re enabling components. What is the cost of a small gyroscope inside a smart phone? It’s pretty low. But if you look at the dollar amount generated by services attached to it and the dollars spent on advertisers on getting location-based information and behavioral models of consumers. It’s reasonable to assume that at some time in the future we may see a shift in the business model. Conventional models still apply for silicon vendors and IP vendors making money. But if you fast forward into the future, analytics may become the basis for service and moneymaking models we haven’t thought about yet.
Heinlein: What’s interesting is when businesses recognize there’s a real win-win. If you could buy an Internet-connected, Internet-enabled dishwasher, that knows to start itself when there is a dip in the energy demand and PG&E gives you a 5% break on your kilowatt hours, that’s a win-win. Plus, you buy a better dishwasher. So maybe it’s four or five wins. That’s the kind of benefit that we’ll see coming out of this.
Leave a Reply