Please excuse me if today’s blog post sounds like a rant. But… There is a feature article entitled “The Patchable Internet of Things” in this month’s IEEE Spectrum magazine. Citing success of recent distributed denial-of-service (DDNS) attacks that were mounted from malware in consumer electronics (CE), the authors advocate “patchable hardware”. I quote:
“We submit that engineers must permit not just the software but also the hardware to be patchable on devices intended to become part of the Internet of Things.”
In a nutshell the article advocates substitution of FPGA chips (field programmable gate array) for the more common ASIC chips (Application-specific integrated circuits) in today’s products. Where to begin unwrapping the many issues with this technologically sound but marketing-mad idea? Here is my list of its biggest problems:
First, the authors are addressing the “Internet of Things” as one market; including both consumer and industrial electronics. The big problem with doing this is the HUGE difference in customer expectations. Starting with expected operating life and supplier support levels, the industrial marketplace demands a far longer-term perspective from suppliers than does the consumer market. If you want to find 10-year-old baby monitors, then look in a recycling bin. But 10-year old I/O modules and PLCs are operating at their mid-life on factory floors all over the world.
Second, bill-of-material cost. For electronic products, the BOM cost is never static. BOM cost initially goes down as expensive electronic components (microprocessors and memories) get cheaper. The cost rises years later, only when the same electronic components are so obsolete that they become difficult and expensive to source. Consumer products are in recycling bins long before that occurs. Industrial products are not. So developers of consumer electronics products pay far more attention to BOM costs than industrial suppliers (and industrials still give BOM cost too much weight). This is important because FPGAs have a significant cost disadvantage vs. ASICs. It’s wildly unrealistic to expect CE suppliers to use them. Sorry. That simply will not happen.
Third, maybe let’s do first things first? First let’s make ALL the embedded software upgradeable? That seems more sensible to me, and giving products this capability also requires more hardware resources (chiefly memory). Except for cellular handsets the CE market lags badly here. Industrial suppliers, who deal with 20-40 year operating life have already done some great thinking. I’m most impressed by this work from Siemens research and this work by venture-stage firm Resin.io.
Summing up, there is no single IoT market. Any IoT analysis that lumps the industrial and consumer IoT segments is not worth doing. The markets are far too different.