How Do Design And Verification Change In The IoT Age?

Where is the Internet of Things (IoT) on the hype curve? Are expectations too high, or is it really the next big thing? My recent trip to the Design Automation and Test Conference (DATE) in Dresden, Germany, did not give all the answers, but it definitely did shed some light for me on this topic.

A very enthusiastic taxi driver took me back 25 years to the Nov. 9, 1989, the time when the Berlin Wall came down. She had used the term “Ja, Wahnsinn” when she heard that I had sweetened my business trip with a ticket for the renowned Semper Opera in Dresden. The term means “madness,” but can be used with both positive and negative connotations. It either describes something great or unexpected, or it can be used for something that goes completely wrong, like jumping off a cliff. Twenty-five years ago in Germany it represented the unexpected joy of the Berlin Wall finally falling and for me. Being born and raised in West Berlin still brings back memories of those days.

So will the hype around the IoT be “Wahnsinn” in a positive or negative sense? Where is IoT in the hype cycle described in my previous blog, The Hunt For The Next Application To Drive System-Level Design And Verification?

Synopsys’s Antun Domic gave one of the opening keynotes titled “Design Will Makes Everything Different!” He set the stage with the obligatory prediction of 50 billion connected devices in 2020, citing the CISCO VNI Forecast from 2014, which will be contributing to 160 million terabytes per month of IP traffic, with 90% being video, 90% coming from sensors. The very interesting, underlying semiconductor aspect that Antun outlined and validated with data was that as the breadth of active technology nodes widens, the adoption also becomes asymmetric.

Specifically, when measuring the percentage of semiconductor manufacturing in MSI (which I assume to be million square inch), then 23% comes from technology nodes ≤ 22/20nm, 44% from ≤ 32/28nm. In sharp contrast, when measuring design starts, only 3% of them comes from ≤ 22/20nm and only 8% from ≤ 32/28nm. This nicely confirms the bifurcation of markets into the (by numbers) comparatively few complex designs and the larger number of designs that can stay at less aggressive technology nodes. (See related article, Fallout From Scaling.) Bottom line, Antun concluded, manufacturing will be less and less a viable option for product differentiation, while design is both the enabler and the differentiator. All that is great news for electronic design automation (EDA).

Rudi Lauwereins from IMEC gave a great presentation in the track called “Enabling a Connected World via Internet of Things,” and cited some interesting research from McKinsey on business models – “Internet of Things: Opportunities and challenges for semiconductor companies.”Exhibit 2 of the McKinsey reference shows the value chain from silicon through system integration and applications. It nicely illustrates how “differentiation” is achieved by applications focused on end users — a key strategy of many players today — and how for that reason IMEC and iMinds just merged to “drive the digital economy.” It also shows that low-level software is expected to be “just provided” as part of the silicon sale, and how “full integration” and “monetization across the full value chain” is a market still too nascent to judge.

Being in Europe, and especially being in Germany, design for automotive applications was front and center, as well. According to the Huawei Global Connectivity Index 2015, of more than 100 billion “things” in 2026, 5% will be car-related. Given that in 2015, the number of new cars produced crossed above 90 million for the first time ever, this means that we are looking at about 50 “things” per car in 2026 … Ouch. How to design those, and what the changing requirements are, was discussed in a panel organized and moderated by Prof. Oliver Bringmann of the University of Tuebingen, entitled, “Which EDA Solutions can the Automotive Domain Reuse? Very Few or All?” The panel turned out to be fun and controversial, especially when the discussion came to how to make sure that open-source software, like one would find in GENIVI, is ISO 26262 compliant.

The introduction of a new gap, as outlined in the following graph, is relevant to this blog post.

As the leading-edge CMOS technology nodes shrink, the maturity time until they can be used in “shorter cycled” automotive domains like ADAS is shrinking dramatically, and has now all but vanished. In contrast, other semiconductor technologies like HV CMOS and Smart Power are much less aggressive. Combine this with the ever-increasing validation effort and the need to make sure that components are compliant with standards like ISO 26262, and EDA is challenged to keep up. When asked by the audience why things were so different in automotive than in mobile and servers, the main issue mentioned was “liability” in the context of human life being at risk in case of a malfunction.

So is the frenzy around IoT “Wahnsinn” positive or negative? It is too early to tell. Predictions are optimistic. There still are a number of areas that need to be worked through diligently:

• Monetization. Who actually makes money?
• Security. How much of our data are we willing to put out there?
• Control. How much are we giving up and how can we be sure that reactions taken by actuators based on sensor inputs are not putting human life at risk?

Bottom line: We will look back 5 to 10 years from now and can then assess what level of “Wahnsinn” the IoT really was.