Expectations about what constitutes a successful product and who’s going to win in this market need to be reset.
The IoT means many things to many people. While the numbers and projections are all very large, the views of what it actually means are vastly different because it can include anything from a sensor in a car or an ingestible medical device to a data center full of servers and the connectivity in between.
1. Digitization is disruptive. It’s also potentially destructive. Prior to the introduction of the Internet, the average lifespan of corporations was 63 years. It’s now 10 to 15 years, according to Aglaia Kong, CTO for the IoE solution at Cisco, and that could be reduced even further with the introduction of the IoT/IoE.
That works in reverse, as well. “About 50% of IoT devices in three to four years will be made by companies that don’t exist today,” said Gowri Chindalore, head of technology and business strategy for the microcontrollers business group at Freescale.
Behind these fundamental shifts in business, and the devices driving those businesses, is an acceleration of digitization. That digitization now encompasses everything from videos to personal data, and while sensors may record the physical world the processing of that collected data is almost entirely digital.
“The biggest disruption is digitization,” said Kong. “When you put sensors everywhere, and there are how many smart devices, that’s digitization.”
She pointed to a recent McKinsey study citing the enormous value from digitization. General Electric, for one, now leases jet engine propulsion, continuously monitoring basically everything that spins as a service.
GE is a often-cited reference point for this shift, but it’s far from the only one. Dean Freeman, research vice president at Gartner, said one of the big shifts that will occur with the IoT involves many more things as a service. “You may see where a city no longer owns street lights. They’ll split the energy savings with the city and keep the system up-to-date and running. You may see this with mall parking, too, where they direct you to a space that may be paid for by advertisers pushing coupons and specials.”
Digitization allows many more things to be doled out as services. So there may be robots as a service or fully equipped offices as a service. Uber is another successful example of this kind of approach.
2. Efficiency really matters now. Moore’s Law has always been about efficiency, but it was largely efficiency in one direction—reducing the cost of transistors every two years or so. In contrast, the IoT’s definition of efficiency comes in many flavors. At the most extreme is the block chain concept, a distributed database supported by peer-to-peer transaction processing that completely bypasses any centralized infrastructure.
Paul Brody, Americas strategy leader for Ernst & Young, points to bitcoins as an example of a block chain. He contends they’re a proven approach to circumventing expensive infrastructure and leveraging the compute and storage power already out there. And so far, at least, there has never been a successful attack on a block chain because the data is so distributed.
“You can safely, scalably deploy transactions across devices,” said Brody. “Those kinds of systems cost 99% less to maintain and manage over their lifespan.”
He said the cost of building and managing the network is zero. “There are 890,000 petabytes of storage in PCs, Macs and iPhone equivalents. And the most amazing feature is you don’t have to trust anyone. The block chain works when the majority of devices work.”
Brody noted that one of the biggest problems we deal with is that we live in a world of “low asset utilization.” The average lawnmower is used 1% of the time, a car 4% of the time, and offices less than 20% of the time. Even airplanes are only used 60% of the time.”
Whether the IoT increases utilization rates remains to be seen, but those numbers are certainly going to be closely monitored by companies that measure efficiency.
3. Smarter is cheaper. It may sound counterintuitive, but the SoC is a less expensive alternative than a custom embedded chip with fewer features. There is so much pre-developed IP available, and increasingly entire platforms and subsystems, that SoCs are replacing much simpler ICs.
“If you look at the Apple Watch, which uses a 32-bit A7, that can cost as little as 50 cents,” said Gartner’s Freeman. “If you use a quad-core Snapdragon, it would be $20 to $30. So if you have a total of $10 silicon content, you can gain market share because you can sell the end device at a lower price.”
This is true even in the microcontroller world, he said, where the 16-bit and 8-bit microcontrollers were the early winners, but 32-bit MCUs designed specifically for the IoT will really take off. “There will be a lot of volume, but at a very low price.”
That doesn’t mean there won’t be a place for less-sophisticated MCUs, though. Greg Robinson, 8-bit MCU marketing director at Microchip, said the real value of 8-bit is power consumption. “8-bit will always use less power,” he said. “You can get down to the sub-nano amp range. You can still do work while the core is asleep.”
How these worlds come together is unknown at this point. At least initially, there’s likely to be an evolutionary approach to combining different capabilities, until the economic tradeoffs of each is fully worked out.
4. Power remains the No. 1 issue. But reliability is a close second, and the two are frequently related.
The biggest problems are battery life for mobile things, and utility costs for racks of servers, and both are exacerbated as the volume of data and the number of interactions with other data increases.
“We’re talking about everything that’s electrical and connected,” said Steve Carlson, group marketing director at Cadence. “As you shrink, there is an increase in problems.”
Those problems show up in everything from online games that overheat to videos that overload the processors because they run at resolutions that are too high for systems.
“This is not a contained problem,” said Carlson. “And it’s not even just about hardware. It’s the use of a core on the hardware. It’s a really hard problem to solve. Heat density is increasing. There is a renaissance in multicore. Single-threaded frequency scaling has leveled off and the number of cores is rising.”
5. Security is an ecosystem problem. One of the biggest challenges for security in the IoE is that just because one piece is secure doesn’t mean another piece won’t compromise the whole system. And if companies developing IoT devices are still on the drawing board, planning a multi-year rollout is almost impossible.
“Security needs to be enabled from the ground up,” said Freescale’s Chindalore, who noted this is being hampered by a lack of standards involving security. “There is not a single standards body out there today.”
What’s needed is end-to-end security, from the edge devices all the way through to the cloud, and at every point in between. “You can expect to have 8 to 10 edge nodes per home,” he said. “The typical end node is an MCU. So you have a suite of microcontrollers, with RF and sensors. Within each group everything needs to be secure, and data transmission needs to happen in a secure fashion.”
This is a tall order in a widely distributed ecosystem filled with young companies. While most people recognize the need for security, one of the big problems is providing metrics so that a value can be placed on security the same way it has been added to battery life or processor speed. Security may indeed be a selling point, but not unless there is a reference point, and not unless the cost can be distributed so it doesn’t impact any particular vendor in the supply chain.