I have seen it. I know you have probably seen it too. It’s that dreaded “Convergence of IT and OT”. It’s the buzz-phrase that just won’t quit.
Another buzz-phrase that I keep seeing is Industry 4.0. Most people who toss these terms around are probably not aware of the starkly different philosophies behind them.
Convergence of IT/OT?
IT/OT Convergence is –well, I’m not certain what it is. It is full of a lot of fluffy marketing babble, and conceptual clouds but not much specificity. The term appears to have been created by marketing people to sell lots of products that enable centralized management of OT.
Many see industrial control systems as a source of data to fill a corporate data lake from which people or artificial intelligence (AI) will divine incredible cost savings that nobody would have imagined had it not been there. This is the sort of thing that is being marketed. There are no explicit notions of who these people are or how an AI might convey what it has discovered.
Nevertheless, marketing people would like you to think that their products are positioned in this way. There is a persistent myth that somewhere, away from the plant floor, is an extremely smart analyst equipped with amazing tools and training. They’re supposed to be people who can discover incredible efficiencies and savings.
I am not sure who those people are, or that they even exist. The most intelligent and aware people I have seen in my career are the people that are grown from inside the company. These are people brought in as operators and technicians, and then encouraged with training and education to become the engineers and superintendents of today. They are the people who know the process best, and can identify and document the cost savings.
My point is that companies do not need to broadcast production data hither and yon in the hope that someone somewhere else will solve problems in ways that the plant staff will not. So what data are we “converging” and why are we doing things the hard way?
In the old days we used to hand-write reports from clipboards of daily data and send carbon copies of the results in a neatly typed piece of paper to summarize what our plants did. Then, as control systems developed in to the digital realm, there were report generators. They could do things like totalize gauges with inhuman accuracy and regularity. They could even perform extensive mathematical summaries to show how well the plant equipment performed compared to theoretical limits. The paper used to circulate through the company.
And then we started doing all this online. In a few cases, people would ask for better data, so historians were created and they were eventually dual-homed on to the company network so that people could look at specific data whenever they wanted to see it. Well, they were dual homed until the discovery that it exposed the company to needless security risks.
We also discovered by then that the people who spend most of their time in the office rarely understood the point naming conventions, the instrumentation, or the details of exactly how the process worked. They would often ask questions because even if they knew about the process, they didn’t have the specifics such as where the instruments actually were, and exactly what they were measuring.
Finally we developed protocols to convey this data and meta-data to a company data lake through a data diode or a firewalled server. And that’s where we are today. Even today, there is never enough meta-data to explain what the production information actually means.
Now we have this talk of “Convergence” of IT and OT –and I really can’t figure out what they’re talking about. Convergence of what?
Is it the data? We already do ship most of the process data to the office. In fact, plant staff spend considerable time explaining to the people in the office building what that data actually represents. If anything, the office has too much data and what they really should do is make a request, so that someone can summarize the things they want in a regular report. The notion that the office staff can generate their own ad-hoc reports from plant data lakes in their networks is pretty much dreamy eyed marketing. Yeah, it is possible, but that level of familiarity and understanding of the data is usually lacking.
Is it the people? If so, who is going to do the training to keep these new people safe when they wander out in to the plant? Who is going to explain the weird and wonderful architecture and the different design parameters? Who is going to enforce the different change management protocols with these IT people? How are they going to manage the updates and coordinate with the plant? Isn’t that why we have OT in the first place?
I think “Convergence” advocates believe that the data needs to be centrally managed, centrally monitored, and that policies should be made centrally. So they talk about convergence, but from my experience the people making policies and managing things centrally have very little idea what they’re dealing with or looking at.
Industry 4.0
Conceptually, Industry 4.0 is really the machine to machine, process to process, and factory to factory connectivity that removes human beings from the picture except in unusual circumstances. It is supposed to decentralize things in control systems.
The notion of Industry 4.0 is that the people closest to the field will have more information and more capacity to optimize. This presumes that the Operations staff will save enough labor through automation so that they will have time to review and react all this new interconnected data.
One area of Industry 4.0 that I disagree with, is the notion that this will all be connected with IoT. IoT brings nothing new to the field that doesn’t already exist in other standard protocols. In fact, when the IoT protocols were being considered, the people behind it had a golden opportunity to leapfrog over some of the less desirable features of industrial Control system protocols. But instead they reinvented the wheel and acted like it was something brand new. Regardless of what protocol we use to move the information, the fact is that the data will move between distributed systems with even less human intervention than what we have now.
Modern technologies are making it possible to push more and more processing power further to the field than ever before. Decentralization of the computing is more and more practical with each new generation of automation. And in fact, distributed automation with decentralized management is where Engineers have been pushing this technology for decades. It is more resilient, less expensive, and because it limits the amount of data it collects, it scales up better.
Comparison of Concepts
Now let’s circle back to the larger picture. The IT/OT Convergence talk is the effort to centralize manufacturing management. The Industry 4.0 discussion is the effort to decentralize manufacturing. These concepts are pretty much at odds with each other in many ways.
Of the two, the effort to decentralize is the best defined, and in my opinion, the one most likely to bear a financially desirable result. Industry 4.0 is easiest to work with from a security perspective because with decentralized security, the plant can break apart in to multiple functional islands of automation. When everything is centralized, if the central node fails, there had better be at least one backup operations plan and preferably more.
The IT/OT Convergence crowd believes that they understand enough and can bring enough financial and operational goodness to the table to justify their involvement and the security risks that go with it. But so far, nobody has identified exactly what that value proposition is.
Nevertheless, if you want your big air-conditioned Valhalla of computing to evaluate everything going on at the plants, go for it. It’s technically feasible –but I’m pretty sure that most of the people involved are wasting their time.
To make the IT/OT Convergence concept work staff need to learn to ask for better reports. If they can’t get those reports, it could be for two reasons: First, they might not understand the detailed process and the contingencies in daily plant operations. Second, they might not understand the limitations of the data they have. For example, there might be a deadband, or a turn-down ratio limitation that may limit the accuracy of the data.
Regardless of which philosophy you subscribe to, it is important that you recognize what you’re buying in to, and what features you should be looking for. And above all, ask what is being “Converged.” If the advocates of “Convergence” cannot describe some return on the investment that the convergence brings to the table, I recommend waiting until they can.
