By Steve Whalley, Chief Strategy Officer, MEMS Industry Group
At Sensors Expo June 9-11, it was encouraging to see many first timers at the pre-conference MEMS track hosted by MEMS Industry Group as well as at the two day main event. Perhaps it was the Long Beach, CA location for this year’s event or just the very healthy double-digit growth rates in MEMS and sensors that brought upwards of 75% of new attendees to see what the excitement was all about. Either way it was encouraging to feel the buzz in the sessions and exhibit hall for all three days.
While the event superbly covered a multitude of topics across the MEMS and sensor supply chain, applications, and adjacent technologies and ecosystems, I wanted to highlight just one important topic that a panel, expertly hosted by Mike Feibus, addressed in the MEMS pre-conference track. The topic was the technical side of Big Data and its use. It’s clear that as billions of sensors are now at the edge of constantly feeding the upstream data pipeline, there is much more to come. With the era of a trillion sensors not that far away, we could be soon dealing with brontobytes (A brontobyte is 1024 yottabytes which is about 1.24 * 10^27 bytes) of data.
Some notable quotes from the Big Data panel reveal some challenges and opportunities our industry will need to deal with. To highlight just three, they are:
1. “A data Tsunami carries a lot of garbage with it .”– Ian Chen, Freescale
Can we keep sending this deluge of data and its associated garbage up to the cloud? According to some, the garbage could be as much as 99% of the content generated. While it’s useful to do all the aggregation and processing of data in one place, it’s unlikely we can always afford the computation, storage and more importantly the bandwidth required and the round trip latency to keep pushing all this data to the cloud. Clearly we need to get smarter at where the processing and aggregation takes place. The Cloud, the Fog, the gateway, the application processor, the sensor hub and the sensor itself are all options along the way in a typical IOT environment. The question becomes; do you know when, where and how you process the data and remove the garbage from further upstream travel with constraints such as cost, power, hardware and software footprint, security, latency, and processing power to name a few? Will we need new system architectures to handle these complex questions in the rapidly approaching era? Or is it more expert data scientists to program in advance languages like Hadoop to facilitate the digital data exhaust?
2. “Sharing data across the ecosystem will help accelerate best practices and solutions.” – Eduardo Pinheiro, Muzzley
The panel discussion around this comment was how do we foster growth and accelerated innovation in the industry by sharing data-sets, perhaps in an open source type environment? As we currently collect data, both valuable and garbage, besides the issues noted in item 1 above, we also have questions around who owns the data and who owns the security and privacy of it. And how do we respect the privacy of the people the data is related to while still delivering a simple user experience? While many are willing to share their date freely today, others are not so forthcoming and this needs to be respected. These questions make it challenging to just liberally share data from user to user to company to government institution for example. Assuming there are applications and systems where answers are crystal clear on these topics, what opportunities would exist if the owner was willing to share that data-set with other clear owners of other synergistic data-sets too? Do we need explicit permission from users to do so? Could we avoid redundancy in collecting similar data-sets? Could we improve device performance, power and accuracy by sharing certain data-sets from a world-wide perspective versus just one region within a country? Could we ultimately reduce time to market and cost significantly? Also, could we increase the accuracy of evaluation of security and terrorism threats by using Big Data? And at what cost to ‘individual’ freedom whatever that may mean to you specifically?
3. “A venue is needed where we agree on how to tag data and share it.” – Ian Chen, Freescale
This comment is somewhat related to item 2 above in that if the data-sets are all characterized by a standardized tagging scheme, the sharing process should be technically much easier. The question becomes is there an existing form of data tagging that could be readily adopted today and if not, what is the right forum to drive this topic to an industry agreed solution? Perhaps we should also ask, does the industry agree that this is a relevant topic to even address? Maybe as a smaller but important step, we could agree on a basic protocol to help make Big Data a little more manageable at the edge and sensor node?
Next Steps – over to you
I have purposely posed questions arising from these comments on the panel versus express my opinions on how to address them. The MEMS Industry Group and the Accelerated Innovation Community we currently use for sharing open source sensor algorithms could be a venue to begin addressing them but it’s not the only one. I would like to hear from you on the relevance of these topics and your suggestions on how we should begin to tackle them. Your feedback would be much appreciated. Please provide your feedback here: MEMS Industry Group LinkedIn community; http://memsindustrygroup.org/linkedin