Cloud Computing – The Inconvenient Truth

The Internet of Things (IoT). If only I could have a penny for every time those words have been written or uttered in the last three years.

I’ve always contended two things about the IoT. One, we’ve always been doing that in manufacturing automation. We’ve always been sending data to the business systems one way or another. I remember in my first job – this is really going to make me sound old – where we had key punch operators. That was long before DeviceNet, Profibus, and even Modbus. The manufacturing people would write down product counts, rejects, quality inspection data, and everything else on log sheets for the shift. The log sheets would go to the key punch operators who were located just outside the “IBM Room.” They would encode the data on cards, and the cards would be given to the computer operators. They would start the application program and load the deck of cards into the card reader, and the application program would process the data and print a daily product report. Actually, it would be for yesterday because this whole process took 24 hours. They would also load a mag tape (magnetic tape) on the huge tape drive and the data would be archived. At the end of the week, that tape would be loaded and weekly and monthly production reports would be printed. Yes, I can hear you saying “Was the computer operator named Flintstone?”

The second point I always make about IoT is that it’s not about the technology – it’s about the business model. Technically, there are a multitude of ways to move data from an automation system to the cloud: MQTT, OPC UA, HTTPS. That’s not hard. What is hard is to monetize it – to develop a business model of monetizing the data in a way that makes sense to the customer. There are few companies in the manufacturing world that have found a profitable way to do that. Many are sending data. Few are making money at it.

What is true of the IoT and that people don’t really understand is that there will be trillions of devices connected to the Internet. Yes, trillions, from everything in a plant to my running shoes. Everything in a plant will have a current sensor, a vibration sensor, a temperature sensor, and more. My shoes will log how far (for me – in feet), how long I ran (probably in seconds), my stride, the temperature outside, my heart rate, and who knows what. When you think about everyone’s shoes connected to the Internet, you see that trillions of IoT devices is conservative.

There are a multitude of implications from this. One, do we have the bandwidth to move all this data to the cloud? We are talking about gazillions of terabytes every second. Our minds can’t really conceive of this kind of data. And second, where is it all going? Are we going to cover every inch of the earth with servers? Yes, we can invent new storage mechanisms, but enough to record data from trillions and trillions of devices?

What this means is that the era of Big Data and Cloud computing is over. We can’t move all the data to the cloud, let alone process it, make decisions and send something back to the process. Can’t be done. If you have 50,000 devices in a plant sending data to the cloud, there just isn’t time, bandwidth, or enough processing power to compute it all and develop new operating instructions from all that data. It just won’t work.

Our biological systems don’t work that way. When your liver encounters an operating problem, it doesn’t send all sorts of data to the brain and wait for guidance. There is intelligence right there in that organ. It processes the data and makes the needed corrections. It will get your conscious brain involved when the problem is overwhelming.

Edge processing like this is the future. Curate the data at the edge. Send only the important data or summary data to the cloud for archival and post done processing. The analytics, the data analysis, is all going to have to be done locally, right there where that data was created.

This is not going to be popular. This flies right in the face of where Google, Microsoft, and Amazon are going. They want all that data sent to the cloud. They want you to process it there and send instructions back to your plant, but I don’t think that’s going to happen.

This is where I see the future of PLCs and of companies like Siemens, Rockwell Automation, and Schneider lies. They need to become edge processing devices that do the analytics of the data on the manufacturing floor and make the decisions to affect the process. Instead of control – they might still do control – they need to evolve into data processors. The end of Cloud computing and Big Data could make Rockwell Automation and others very successful.

Frankly, I hope that happens.