The Next Bigger Things

The Internet of Things is turning out to be a lot more complex than initially predictions, and there doesn’t seem to be any end in sight to just how complex this will become.

popularity

When the Internet of Things really started making headlines several years ago—the concept had been around since at least the early 1990s—the assumption was that most of the semiconductors involved in sensing and communicating would be simple, highly limited, and developed using older technology. As the concept evolves and grows, however, it’s beginning to take on a whole new texture.

Rather than dumb chips, the new incarnation involves highly advanced communication devices that have built in security, advanced power management, and highly complex networking capabilities. And while some of the sensors may indeed be based on older technologies, they likely will be paired with logic, memory and IP developed at advanced process geometries.

“The basic premise is that the IoT involves a little bit of RF with simple devices to connect to the Internet,” said Michal Siwinski, vice president of marketing for the System And Verification Group at Cadence. “But that’s changing with consumerization of the marketplace. You can see that with Nest. It’s a mobile system with software and an operating system. It’s not just a simple sensor. People want the full gadget with all the tricks. If you’re shopping for a smart phone and you get to choose between a dual processor and a quad processor, you’ll take the quad processor even if you don’t know what it does. There is an insatiable demand for technology, and with the Internet of Things we’re going to have very complex devices that are quite smart, not just simple sensors.”

More connected stuff
Ideas start out simple enough, but they quickly grow as connections are conceived or made possible. So rather than just things, they evolve into complete systems talking to other systems.

“The connected ‘babygrow’ is a cool idea, and you can buy one today, e-mailing you the body temperature, breathing, or other factor for the baby,” said Hayden Povey, director of marketing for security at ARM. “However, if we can pair this with the fire alarm and lights to wake you up in the middle of the night if the baby stops breathing, you suddenly have an power system that delivers fundamentally more value. The issue is that none of these systems, the babygrow, the alarm, the lights, are designed to work together. We have silos, and we need to move to the full IoT.”

This is one facet of why the IoT is becoming so complex. Not everything is designed to work with other things or networks—essentially a series of unanticipated use cases—which means that “things” need to support multiple connection options, from wearables to smart homes to medical devices.

“All of these applications have unique requirements from a power, form-factor, user-interface and performance standpoint,” said Eran Briman, vice president of marketing at CEVA. “Such a diversity also means further difficulties in getting all the sensors to talk to each other, to a central hub (such as a smartphone) or to the cloud. There are standards missing here with regard to unified local (on-device) data-fusion APIs, and a preferred communication channel—BLE, various WiFi flavors, 802.15.4—or we may see the proliferation of multiple air interfaces, sometimes co-existing on the same ‘thing,’ depending on the application, its required bit rate, range, or even geographic location. IoT-specific flavors of the OS and the sensor-specific, always-on APIs are other domains where we’ll see a lot of evolution and a race of different vendors while the ecosystem standardizes and settle down. For example, the recently announced Samsung Galaxy Gear 2 runs Tizen and not Android, while Google is rumored to quickly join the game with its own wearable version of Android.”

Complex security requirements
Connecting everything also means that the outside world is connected back, though, which creates a security risk of unprecedented proportions. In a blog published last November, Semico Research analyst Adrienne Downey highlighted a panel discussion about whether the IoT is driving us toward a security Armageddon.

That security risk runs from servers in the cloud to home networks to the devices themselves. ARM’s Povey sees six major challenges associated with security:

  • Identification. There could be as many as 6 million devices per cubic kilometer—and the number needs to be cubed because of multi-story buildings—in large cities. That means unique identifiers and key management, and possibly a public key infrastructure.
  • Authentication. New devices, such as a smart lightbulb, need to register automatically.
  • Device lifecycle management. How do you protect against viruses over the lifespan of a 10-year product, such as a washing machine?
  • Communications. How do you ensure performance with tight power constraints over transport layer security?
  • Privacy. A public key infrastructure will play an important role here, but complexity in managing data streams that need to be private is a huge challenge.
  • Protection. Things need to be tamper-resistant or anti-malware. Anti-virus solutions won’t be sufficient.

Security comes with a price, though—both in real dollars and in terms of performance and power.

“No matter if it’s done in hardware or/and software, security certainly adds more work and should increase total power or energy over exactly the other one with the exactly same conditions except security,” said Charlie Su, chief technical officer at Andes Technology. “But, it can get to the same or even lower power by, say, using a more advanced low power design or a more advanced process. Either way, it’s fair for an additional feature to cost a little more.”

Security measures are dependent, to some extent, on what kinds of attacks are likely to be used against a device. “For network attacks, devices would start with authentication and send only encrypted data,” said Su. “For physical attacks, such as devices that are stolen or placed in a lab for analysis, the devices must protect both program code and data, and secure CPU cores must be used. Security solutions must be implemented as a combination of hardware and software. With more hardware, it usually costs more, but it runs faster and burns more power. With more software, it usually costs less but runs slower and uses more energy.”

Security also can involve separate systems for the IoT, according to CEVA’s Briman. “One part could do the sensing, user-interface (voice/gesture/face activation) and always-on functions, and the second part the communication function. Both must be extremely low-power in order to fit the power envelope of most IoT devices, hence application-specific DSPs are a good match. On the cloud side, data coming from billions of things necessitates heavy real-time processing. A strategy to offload some of the tasks to a dedicated, high-performance DSP from a CPU or apps processor makes sense to deal with processing requirements.”

Bigger business
What’s clear is that the IoT is truly beginning to pick up steam, and complexity is both a result of that growth as well as a source of new opportunities.

“The vision of the opportunity is becoming clearer, but the underlying technology is not defined,” said Jerome Nadel, chief marketing officer at Rambus. “This is about capture, secure and move. It’s that underlying technology where the computing is happening, and how you secure and move data is material to that.”

Rambus looks at the IoT from three levels—the sensor level, the gateway level and the cloud level. “If you just have great sensors you’ll be commoditized quickly,” Nadel said. “You need to have two of the three pillars to be successful in this, and if you have all three that’s even better.”

At the SoC level—which intersects with all of those—the top concerns are low power, area and cost. A wearable device, for example, needs to take into account electromagnetic radiation, heat, as well as the ability to communicate sizable amounts of data to the cloud.

“The good thing is we’ve solved all of these problems at the chip level,” said Aveek Sarkar, vice president of product engineering and support at Ansys-Apache. “What’s lacking at this point is the platform to pull it all together. Will it be Android or some other open platform? Once we establish that, you’re going to see a whole wave of innovation.”