Edge computing is rapidly becoming the most critical component of industrial IoT, replacing pure cloud-based systems in large manufacturing environments. Over 72% of American manufacturers are already investing in Edge computing solutions, according to Futurum Research:
The reasons shouldn’t surprise you.
We all started hearing about the Internet of Things (IoT) early in the last decade. The promise of IoT was that almost everything could be connected to the Internet, each of which could send data to the Cloud for processing and analysis with a whole set of applications that could take that data, aggregate it, and make decisions using advanced computing techniques like Artificial Intelligence. The Nest thermostat was born, along with hundreds of other appliances that were going to "change our lives." It all sounded so, uh, promising.
But for industrial IoT, a number of issues arose. As you can imagine, even a medium size factory could end up with literally thousands of devices, all connected to the internet, and all sending data to the cloud for this computing to take place. This is a lot of potential data, potentially much more than the bandwidth capacity of many factories without significant upgrades and expense.
Security is also a huge concern. According to Department of Homeland Security incident response data, since 2015 U.S. manufacturers considered "critical" to the economy and to normal modern life, like makers of autos and aviation parts, have been the main targets of cyberattacks—outstripping energy, communications and other critical infrastructure.
Case in point: In August of 2017, AW of North Carolina, a large manufacturer of automotive parts for Toyota, was attacked by "ransomware." The malware entered the North Carolina transmission plant’s computer network via email last August, just as the criminals wanted, spreading like a virus and threatening to lock up the production line until the company paid (an undisclosed) ransom amount.
AW North Carolina stood to lose $270,000 in revenue, plus wages for idled employees, for every hour the factory wasn’t shipping its crucial auto parts to nine Toyota car and truck plants across North America, said John Peterson, the plant’s information technology manager. As a result, many large manufacturers actually have rules preventing ANY of their devices from being exposed to an internet connection.
Latency is also a potential problem. Of course, it is ‘cool’ if your Nest thermostat is a few seconds late turning down your temperature due to a hiccup in the network, but not so ‘hot’ if a sensor on your production line is late reporting a failure and your robots weld a bunch of bad parts together.
All of these problems begged for a smarter architecture, one where:
The data could stay in the factory
Banking-level security and zero inbound connections
Zero latency was possible
The idea of edge computing could solve all these issues, but the processing power had to go up to handle all these devices and the cost had to go down so you could deploy these edge computing devices across the factory at scale. Then along came the Raspberry Pi and other similar devices that exponentially improved the power to cost ratio for computing on the edge. And they could be fitted with as much memory as needed for the job at hand.
Finally, we have a new generation of solutions for Industrial manufacturing that deliver edge-computing applications. The best ones use edge computing for instant processing of local data and ban inbound connections from the internet, but still aggregate the results into reports and alerts that can be acted upon and used to improve the machine-learning models. Tend is one of those solutions. We check all those boxes and we also have some of the largest manufacturers in the world installing our solution.
It is unfortunate for some IoT solutions vendors that they built their entire system before edge-computing was ready. Most of the robot vendors themselves fall into this category. Fanuc’s ZDT and Kuka’s SmartProduction both require all their data to be sent to their massive data centers for processing. Yikes! But it is understandable as they started developing these products way before more modern edge computing devices were available.
But you don’t have to be sucked into a bad architecture. With modern edge computing providing local processing complimented by the cloud for improvements to machine learning models, reporting and alerting, you can have the best of both worlds. If you’d like to see how Tend is doing this for our customers, please contact us.