Experts At The Table: The Internet Of Everything

Last of three parts: Who’s responsible when something goes wrong; security issues; local vs. cloud; re-usability of IP; what will speed up or slow down adoption.


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: Considering much of the Internet of Things will involve older process geometries, what will differentiate one chipmaker from another? What will constitute a winning strategy?
Guedj: In any industry where things are still being defined, you need to move fast. You need to have an ability to grab things and put products together very quickly, adapt them and change them very quickly. You can do all the studies you want, but at some point things will change and you need to adapt. That affects the ecosystem, whether it’s software or hardware, because not everything will be programmable. You need to make changes according to how people are using the products and the market. Time to market and rapid evolution are key.

SLD: But this also will be a very long-term play, right? The Internet of Things is a unifying force for lots of markets, from cars to medical devices.
Heinlein: If you think about your connection to the Internet, you have a very low tolerance for pain on that. You will have an even lower tolerance here. So you’ll probably see a multiple-layered approach. First, you have more fixed infrastructure and commercial installations. That includes things like smart streetlights for energy savings that auto dim when there are no cars on the road. As we get good at that, and get local-area networks and personal-area networks solved, it will go more and more into the home and the personal space. You’ll get innovation at all these levels. Networking, including self-assembling networks, will be critical to this.
Izadi: Networking is going to be the platform. Customers are not interested in buying a box and paying for that box up front. They want to turn on services as they go. The semiconductor hardware players have to get used to that model. You don’t get paid for that piece of silicon right away. That’s going to be the challenge. How do you manage that business model and still make it successful.
Logvinov: We’ve been talking about the smart home for six or seven decades. It still isn’t successful yet. In today’s environment, it’s being talked about again. Why? Because we need these services. We need new ways to manage and interact with newly appearing electric cars and many other things that enter our home. If you talk about success in that environment, it relies on the ability to provide a pre-integrated platform for the ecosystem. Whether this platform is at the silicon level, software level, or a hardware box, it all needs to come together. The timelines and development cycles are very compressed. You need very short time to market enabled by platforms.

SLD: Will the Internet of Things be a driver for a platform strategy?
Heinlein: There are a couple different issues. Different applications will require different metrologies. Maybe you want partitioning, motion or protection from humidity. You want plug-and-play solutions so you don’t have to do custom chips for everything. And if this becomes pervasive, there is an issue about how much data gets created. So you need to do more and more local processing and filtering. ST has a sensor that does multiple different things together. That sensor could do a lot of filtering so the traffic that gets sent up to the network is much more efficient. The sensors adapt to their application so they don’t overload the network.
Guedj: We were talking about the smart home, but we also have to think about how smart we want our home to be. There is a question of how the consumer will feel about their body being controlled by the Internet. There may be some models that work and some models where consumers push back. And there may be some levels of assurance for the consumer that we will not stop their pacemaker remotely or force them to drive into a tree. You can’t afford to have everything programmable. You need a mix. And you need to re-use as much as possible into a bigger subsystem. That will be the challenge.
Heinlein: This goes into packaging, and you’re going to see heterogeneity in terms of the sensors. They could very well be in different process technologies.

SLD: You’re talking about 2.5D and 3D stacked die?
Heinlein: Yes. That’s pretty expensive now, but conceptually that kind of technology is valuable. And there’s also the question of whether it’s battery-powered or energy harvesting. It’s all of the above. But when you can buy a thing at the store, put it on the wall, and five years later the sticker turns red and you pull it off and buy another sticker—that’s the level of integration we need to get to.
Izadi: One of the very significant drivers of architectures and how they will be implemented is security. And it’s not just security in what we think about today. It’s security in the sense of unified platforms that can cater information to multiple consumers. Now, the question is how we separate those domains. How do we make sure a pacemaker is not affected—or a car is not affected—by someone trying to hack into your music library? We need to make sure that while we have a single source of information, there are silos—and those silos are separated so there is no danger of intrusion.
Logvinov: It’s not just intrusion. The system can fail.
Izadi: Yes. The system has to be resilient enough to self-recover.
Logvinov: Reliability, security, the integration of all these pieces—those are critical. What’s also required is some sort of a push toward standardization so these pieces can work together seamlessly. How that standardization will come together has yet to be seen, but I could envision an open-source type of development where people could develop software or pieces of IP, and they could share it. That could evolve into an open-source development structure where a lot more people are involved versus a close ecosystem of companies.

SLD: What becomes the basis of that standardization? Is it the hardware, the software?
Logvinov: It’s everywhere. If you look at what’s been standardized, everything that’s been standardized today is under scrutiny for whether it should be standardized. Open source provides a great platform and it’s a wonderful way to create rapid development, but at the same time what do we do with security concerns?
Heinlein: We have an initiative called Linaro that focuses on middleware. How do we take the middleware that you need and get it done more quickly. That’s a case where open source has been successful. When it comes to security, there may need to be an initiative to keep things safe. There’s going to have to be a balance. But for standardization, networking is going to be a key challenge. There are a few standards. IPv6 is going to be an issue, rather than only focusing on IPv4.

SLD: One of the challenges of standards also is that you don’t want to implement them too early. As you get standards things progress faster, but standards also can create problems as witnessed by the CPF-UPF issues.
Guedj: Standards will evolve, but there is a challenge because of the diversity. People solving one problem are not going to standardize with a totally different application. We need some intelligence to buffer that. The standards will evolve, and hopefully you won’t need to re-do everything.
Izadi: That’s why the programmability piece is important in the devices, so as the standards mature you can program them in. We’ve seen that in cable. The standards can be programmed into the field remotely without having to change the hardware infrastructure.
Logvinov: The ability to evolve standards is very important. But as an industry, we also have to be cognizant of a different problem. There are more and more standards bodies. If you’d like to implement something, here’s a standard from Standards Body A, another from Standards Body B, and another from Standards Body C. This is something we need to deal with as an industry.

SLD: We also see other standards that don’t go anywhere and which aren’t fully baked. How do we bridge those worlds?
Heinlein: There’s nothing like a killer app to cause standardization. I recently bought a Nest thermostat. Being able to detect when you’re there and when you’re not there, and then turn the heat down, is pretty interesting. You can imagine having those kinds of sensors in every room so it knows you’re not just behind a closed door. An app like that may seed other standards. Bluetooth Low Energy, ZigBee—those sorts of standards have been around and are beginning to take hold. Once one or two of those begin to take hold and get critical mass, the industry will rally around them.
Logvinov: I see multiple standards in places where we have a push for technology, and far fewer in places where we have a pull for the technology. If you look at standards driven by service providers willing to adopt something, the convergence happened very quickly. Where you see silicon vendors or equipment vendors trying to push standards forward, there is always a flurry of multiple directions.