IoT Growing Pains

The Internet of Things is having one of those defining moments — and not in a good way. Only a limited number of things will work together at first. Only in rare case will they actually cross boundaries between vertical markets such as consumer or home, automotive and medical. And in no cases will there be any guarantees, at least in the short term, that communications will be reliable, secure, or consistently fast enough.

There already are many devices that are connected to the Internet with a “smart” label on them. Like any mobile or connected device, a Tesla automatically downloads software at night from the cloud while sitting in the garage. The garage door can be controlled by a smart panel that locks out all remote control signals when you’re not home to keep your home safe. And moving inside the house, devices such as smart baby monitors can track everything from noise, motion, temperature and humidity over a phone, while a smart thermostat can regulate the temperature for rooms you’re actually using from that same device.

The promise is to move many devices much further along. The picture being painted is one of driverless cars that can avoid accidents and minimize delays; of watches that can detect heart irregularities and chemical changes in your skin to predict heart attacks a day or two before they happen; and of devices that can relay information across physical boundaries, shutting down electronics when you’re away from home, starting them up automatically when you’re at the end of your commute.

Nor is this confined just to home electronics. There is technology available today that can be embedded in cement to detect when a minor earthquake has damaged a bridge support deep inside the post and it needs to be replaced. And there are sensors that can be embedded into soil to determine the moisture levels, acidity, and whether certain parts of a field need fertilizer more than others.

These ideas are all interesting, and business cases can be made for many of them. But what happens when there’s a power outage and Internet access goes down? A completely autonomous car would stop. In the past, security breaches have been largely limited to credit cards and online sales, but in the future they might provide all the codes to your home. And what happens when one of the sensors inside a bridge malfunctions and doesn’t get or transmit the right message?

Defining and redefining the IoT
Getting this right is ultimately about technology, but the first step is finding a common lexicon to discuss it. The language and issues are changing, though, as companies dig deeper and begin to engage in the IoT.

“The first set of issues involves figuring out what we’re all talking about,” said Serge Leef, vice president of new ventures at Mentor Graphics. “Are we talking about end-to-end communications, or Web-based applications, or physical device to physical device? What are the elements involved? Is it an edge-node device that links to a WiFi access point, or is it a closed system? And if it is a closed system, is it simple or complex, because there may be a variety of ways of implementing it.”

There are some variations on this theme, but the basic premise is the same. Definitions are murky, at best, and not always useful. And sometimes what one company calls smart isn’t necessarily even part of the IoT, while for others, the association may be meaningless.

In fact, Chris Rowen, a Cadence fellow, said one of the hallmarks of a successful IoT application or technology is when people stop referring to it as part of the IoT.
“When you talk about wearables, medical devices or smart industry lighting, what you’re looking at is the real structure of the problem. There are some unifying themes, but it is abundantly clear that the more you drill down, the more there are unique characteristics and differences about what you care about.”

Following the “smart” money
Rowen said that for an SoC—which will probably be part of a smart phone, PC or a home gateway, the emphasis will be on low cost and low energy. But for other devices, such as a smart Band-Aid, the essential ingredients might be a combination of extremely low power, very low cost—perhaps as little as 10 cents total—and maybe an energy harvesting component.

And therein lies one of the fundamental problems of the IoT—squeezing enough cost out systems while preserving enough functionality to make them attractive to consumers. Part of this involves partitioning the solution—what goes into the end device, what goes into the edge device, and what goes into the cloud. A Nest smart thermostat, for example, communicates with a phone running its own app, while a Sonos wireless audio system supports multiple vendors’ devices and apps. These are two different approaches to basically the same concept.

Part of it also involves classic semiconductor considerations, namely how chips get architected and packaged at the lowest cost possible. “When you think about an IoT device, you’ve got a small die, which is either an ASIC or a microcontroller, a sensor, which is probably a MEMS device, and connectivity,” said Raj Pendse, vice president and chief marketing officer at STATS ChipPAC. “If you try to miniaturize all of this in a module, you’re probably looking at a fan-out package instead of putting everything on a board.”

A fan-out approach basically moves components on a PCB into an expanded package, but it also makes it simpler to route if it is mounted on a board. “We’re working with several companies using this approach, some involving wearables, some not,” said Pendse.

Rowen said many of these systems also will be developed at older nodes, where the cost of designs will be significantly lower. But that doesn’t necessarily mean less sophisticated systems. “There is a potential explosion in the number of designs, which means tools and IP have to evolve fast. They have to be reasonably priced, with fast turnaround, and little exploratory work. But there is also a global lack of sophistication about where does the energy go in chips, which is really an understanding of what needs to be on when. Almost all IoT devices will be off most of the time, but some parts can be off 99% of the time, some parts can be off 99.9% of the time, and others can be off 99.99% of the time. That’s a big difference in power consumption.”

Planning for the bad stuff
Knowing what you’re creating, how it’s supposed to work, and squeezing every last penny out of the design is only part of the challenge with the IoT, though. Getting everything to work, and to continue working, is potentially an even greater problem.

J.D. Power conducted a survey of new car owners in June. The newest cars experienced more complaints, primarily in the areas of voice recognition, Bluetooth pairing and audio systems, according to the company. In other words, the problems are showing up in the same infotainment systems that currently are connected to the Internet. And while it’s unlikely those systems will interface with the drivetrain and braking systems anytime soon, it shows there is still a long way to go in solving even basic communications issues.

There also is an enormous problem looming in security for these systems. “We’ve got the big brains in this area focused on securing large-scale systems like battlefield communications,” said Mentor’s Leef. “But this is going to be way bigger, with much easier access to firewalls. You’ve got machine-to-cloud, machine-to-app, and machine-to-machine. If everything was linked to the cloud it would be easier to secure, because then you could use a layer of obfuscation above the cloud. It’s not, and with the IoT the notion of hardware security is becoming scarier. Think about what happens if you compromise all the toasters in Orange County. You can burn down 400,000 homes.”

Outages are potentially even scarier. The more reliant systems are on connectivity, the greater the impact when there is a failure. And the longer the outage, the greater the impact. Think about a driverless car system that goes out at rush hour, for example. Or a medical device that administers medicine based on a centralized computer system. And then think about how long it has taken just to close up a security hole in credit card theft by using chip-based cards instead of magnetic strips, which won’t be standard in the United States until 2015—more than two years after massive breaches by some of the nation’s largest retailers.

The IoT, for all the hype, is still in a formative stage. It may be years before it enters the next phase.