New Tools Enabling The Internet of Things

Last week I attended CDNLive Boston as a speaker and was really looking forward to the keynote given by Samuel H. Fuller, CTO and VP of R&D at Analog Devices, called “The Third Exponential Wave and the Challenges Ahead”. It was great to see, re-affirmed by Dr. Fuller, a lot of my thoughts about the Internet of Things and how it requires new tools in EDA. This, by the way, conveniently translates into me being safer by the day to be able to bring my 9-year-old daughter to school, assuming EDA will keep me around in it for a bit longer.

Dr. Fuller opened the presentation with the statement that, in his view, we are at an important inflection point in time, offering the semiconductor and EDA industries the potential to make a real difference and deliver unique value to customers. This could provide the spark to allow the industry to break out of the +/- 3% growth rates in semiconductors and lead to the creation of new products and EDA tool flows.

A brief recap of the past characterized the last two waves along the axes of computing, networking and applications, leading to the picture in this post below showing some key devices and their reach as “installed base” over time.

The first wave was characterized by centralized computing combined with centralized data and signal processing, lasting really from the 1960s to the 1990s. On the networking side it used circuits switching, and, yep, there it was, on the screen, my old acoustic coupler with which I connected my Commodore 64. Those were the days! Applications were written using programming languages like Cobol and Fortran.

The second wave, from the 1990s to the last decade leading up to 2010, was characterized by personal computing, combined with the invention of packet switching networks and personal productivity apps like Visicalc, combined with Web access.

In the third wave, the one upon which we will embark, computing becomes focused on pervasive sensing, creating massive data. Wireless networks become ubiquitous, enabling data to be shuffled in from every point, enabling “big data” analytics.

Dr. Fuller then proceeded to outline four critical enablers for the industry to ride the third exponential wave – low power, security, multi-die products, and embedded analytics.

Low power is a pervasive issue from the edge-node devices that may be remote from power sources, may have limited ability for battery replacement, or may even rely on energy harvesting through the networks transmitting the data to the servers in the cloud that today already consume 2% of energy in the US. The big data analytics we are looking at going forward will require 10X to 100X of today’s compute power, but can hardly consume 20% or more than 100% of energy in the US. Key focal areas for low power will be edge devices designed in the microwatt age with sub-threshold circuits, aggressive power and clock gating or even asynchronous circuitry, network protocols that reduce power consumption with low duty cycles, ultra-low standby and efficient packets, as well as radical reductions of the required joules per computation.

Security is equally pervasive across all areas of the . Dr. Fuller pointed out that encrypted communication for a secure network will be necessary but not sufficient. The combination of sensor and actuator network and the cloud need to be secure, and just like there are certification processes today for robustness, there will likely be certification processes for security. This confirmed some of my previous thoughts, as I discussed in “Are Value and Security Needs Misaligned in the IoT?” and “IoT – Is Security Impeding Development of the IoT?”, as well as in “Securing the Internet of Things”.

Multi-die products are one area in which Dr. Fuller pointed to the need for new EDA tools and flows. Digital signal processing, analog processing, multi-voltage power management blocks, and MEMs transducers will have to be combined into single packaged semiconductor products to enable minimized interconnect parasitics and size, delivering a full solution to customers.

The fourth critical driver, embedded analytics, is based on the realization that the analytics itself will also be distributed across the different areas of the IoT. Local, embedded processing analytics will be required at the edge nodes because of low latency requirements, local optimizations, and safety. Mid-level cloud and server processing and analytics will enable sensor fusion, machine-to-machine communication, and site optimizations. Finally, enterprise cloud processing and analytics will be required for supply chain optimizations, and will allow detection of and adjustments to shifts in markets and the environment.

Lots of challenges lie ahead for the semiconductor and EDA industries to sink their teeth into, and as I had previously outlined in “Internet of Things (IoT) and EDA” and “The Importance of Ecosystems in the Internet of Things Era”, quite a lot of innovation will be required across several types of ecosystems. This will keep us busy for years to come, which is good news for my daughters school …