3 Key IoT Benchmarks

The  has been billed as the next huge opportunity for semiconductors—tens of billions of things connected to a centralized infrastructure, to people, and in many cases to each other. After conducting hundreds of interviews with executives, engineers and university professors from all facets of the global supply chain over the past year and reviewing dozens of technical papers, Semiconductor Engineering has found three critical benchmarks that will define the IoT’s future.

The biggest problem that many companies find with the IoT is that the vision is too vast and complex. That has led to conflicting approaches at times and far too many standards to be useful. The bottom line, though is that no company can it all, no matter how grand their ambitions. And no one person can describe it all in enough detail to be useful. But there are some common critical benchmarks that cross enough sectors to be useful in understanding where are the bottlenecks, how progress can be measured, and what needs to be done in the future for the IoT to live up to its promise.

1. Relevance. The killer apps for the IoT are still being defined and refined. This doesn’t mean that smart devices and pervasive connectivity aren’t important. In many markets, being connected and context-aware is a major selling point. But so far, the advancements being made in many markets such as automotive, industrial or consumer are market-specific, and frequently location-specific, and none of them is satisfying enough to say there isn’t room for improvement.

Much attention is focused on wearable devices, in large part because they are the touch points for how the greatest number of people will interact with the IoT. A connected watch or portable fitness monitor may be fun or cool, but are these killer apps? Maybe. And if not, what will make them killer apps? Those questions need to be answered by markets, and they need to be proven as sustainable over time.

Moreover, the underpinnings of the technology need to advance with the ideas. For example, there has been research on monitoring heart irregularities and body chemistry that could lead to either a stroke or a heart attack. So far, those are still in the exploratory stages. But until a wearable device can go for days without recharging, because some of these monitors will require readings while people sleep, it’s not clear how useful or attractive they will be. Technology needs to move in lock step with the concept, and so far there is a huge gap to close on power management, energy scavenging, and possibly semiconductor architectures and microarchitectures.

Automotive sensors are another potential killer app. Cars that can communicate with other cars, or at least sense their presence, potentially can slash the number of serious accidents caused by blind spots, inattentive driving and bad road conditions. Whether this ultimately leads to autonomous vehicles, or vehicles that are just safer, will depend on country- or region-specific regulations, but there seems to be almost universal opinion this is a good first step. How far it goes from here—whether insurance companies will receive real-time information about driving habits or whether people simply sit back and let the car do the work—is all speculative at this point. But none of this can progress without a buildup of a communications infrastructure that can handle massive amounts of data with reliable uptime, throughput and always-on connectivity.

Industrial sensors are a third area for possible killer apps. What’s interesting about the Industrial IoT (IIoT) is that it’s already up and running, and saving untold amounts of money for companies. But whether this is really the IIoT, or simply a collection of industrial sensors deployed within companies, is a matter of debate. The IIoT is ill-defined in part because companies don’t want to talk about what they’re using for competitive reasons, and in part because these are unique applications of technology that frequently don’t connect to anything outside of a company’s infrastructure. Many of these organizations use private clouds, install sensors inside their industrial equipment for uses such as monitoring the flow of liquids through a valve or vibration in a pipe, and analyze their data internally or even remotely. But while those things use a common communications infrastructure, they’re otherwise company-specific.

Many analyses by companies trying to assess the IoT have concluded that it will become, at least initially, a set of isolated markets where devices are connected within those markets. Whether that makes it relevant, or whether it limits its relevance, is uncertain at this point. But one of the key considerations for the IoT will be relevance to something that people value enough to pay for, and therefore warrant continued investment. So far it’s uncertain what can scale, what technology will be needed to make that happen, and who will benefit most from these killer apps.

2. Security. Being connected has enormous benefits, but it also raises the risk of security breaches to unprecedented levels.

While security won’t limit development of the IoT, it could limit the rate of adoption in consumer markets, what gets connected inside of automobiles, how much data is shared through medical devices, and the connectivity options for companies developing the IIoT. Most security experts believe that breaches will need to occur before the entire supply chain embraces security.

Security is an end-to-end design challenge, from the initial design of IP and I/O in an SoC through the manufacturing process to the supply chain, and from the software to the network and cloud infrastructure. This is a system-wide problem, and the system includes everything that is connected within it and across it. Every point in a complex interconnected Web of communication needs to be impregnable, and new information about breaches needs to be shared as breaches become more sophisticated.

Cyber crime, cyber warfare and cyber terrorism all will leverage weaknesses in system design. The more things are interconnected, the more a vulnerability in a seemingly insignificant area could affect something else of much greater importance. The old observation about the weakest link in a chain is magnified in cyberspace, because breaches can start from anywhere on the planet. And some of those breaches will have the full support and resources of governments, sophisticated criminal organizations, and even companies looking to gain an edge on rivals.

One of the scenarios that has been frequently cited in Silicon Valley over the past year is that government agencies are at least moderately prepared for someone hacking into a power plant, but what about causing 30,000 smart toasters in a crowded metropolitan area that can be made to overheat and start fires? Car electronic control units can be “buzzed” to take control externally. Chips can be ground down and probed to figure out how they work using side-channel attacks. And memory impressions can be detected to figure out what was a one or a zero.

Software is even more of an issue, largely because complexity can hide vulnerabilities and breaches. And what gets manufactured in a small fab somewhere could easily contain extra circuits that were not part of the original design.

There is no such thing as complete security. Credit cards for years were considered more secure than carrying around cash, and banks were considered more secure than hiding money in your mattress until many of them failed. But it will require diligence in many sectors that never worried about it in the past, and it will cost money to address the problem more effectively than it is being addressed today. The big question is who will bear that cost, how quickly it can happen, and how difficult and lucrative it is for organizations to tap into a much more connected world.

3. Interoperability. A good way to view the interoperability of IoT devices is from the number of remote controls people use to manage them. In a fully connected IoT world, there should be one remote control. Even the universal remotes that are available today are difficult to program because there are so many protocols to support. Some are vendor-specific, some are outdated, and some are so new that not all of the advertised features will work.

This extends well beyond just the home, too. Automobiles are built to adhere to rigid standards because they contain safety-critical systems. But as the owners of high-end automobiles found out recently, when an automotive infotainment system is driven by a smart phone and that smart phone operating system is updated, the automotive infotainment system can fall out of sync to the point where it no longer works.

One of the biggest complaints among new car buyers is that the technology doesn’t work as promised. J.D. Power released a report on Feb. 25 that found Bluetooth connectivity and voice recognition were among the most frequently reported problems after three years of car ownership—issues annoying enough to impact brand loyalty.

Nor is it limited to these markets. With so many new devices coming on the market in every arena, dependability is now measured by connectivity. That connectivity can be impacted by other devices that could affect communications, security issues, design failures of individual components, software glitches, sensor fatigue, electromigration, or even environmental extremes such as blazing-hot deserts, arctic cold, or from the growth of tin whiskers in outer space.

As more devices interoperate with other devices, reliability becomes an issue that is defined by all of the components in a communications scheme rather than one component. A car driver doesn’t know why a particular function isn’t working. It may be the sensor, the car’s internal network, or a base station two miles away that is required for communication to a cloud. But until that kind of information does become available, and until companies begin working on much larger systems rather than their own slice of it, interoperability will remain a problem in need of a solution.