Recently, a friend described to me a great installation of one of the low level protocols. A few years back, when they added an IP binding to the protocol, they had been able to construct an integrated control system between two factory production lines. One was at a final assembly factory in a northern city and the other at a supplier in a rural area down south.
Each of these factories used the same low level control protocol. My friend described to me how they were able to integrate the inventory control and JIT manufacturing at the two plats by linking their control systems. He was quite proud of it, and I’m sure it was a technical tour de force.
I was horrified.
There is no useful control level information exchanged between the two factories. Unless they do no quality control, there are no real exact inventory linkages between the systems. Certainly, there is no real-time linkage – they are a day’s drive by truck apart. There are only two reasons to link the systems: (1) they use the same protocol or (2) the interface is so contrary to what the corporation expects that they do not want to do it twice.
The first reason is a horrible reason to link systems. It is an example of the “Small Boy with a Hammer” school of analysis. More importantly, it skips all interesting and useful service integration by plunging quickly to process integration. Process integration is the process of paving the cow-paths. Process integration makes for brittle systems. Process integration prevents agile decision-making in the future. More importantly, process integration forgets the services that each system performs and replaces it with a mechanistic approach that belies the services.
I like to refer to system services as an overlay on system mission. Each system has a mission, and its first job is to defend the integrity of that mission. HVAC systems have a mission to provide healthful air conditions and long term maintainability of a building. They should defend this mission, including preventing outbreaks of Legionnaire’s Disease blocking large growths of mildew in the walls while providing the services of responding to requests to shed load or to the requests by occupants who push the thermostat up. Fume Hood systems have a mission of life safety that they must not skip. This system must defend this mission, as flooding the lab with poisonous gas is not an acceptable response to a load shedding request. Two systems. Nearly identical components. Perhaps the same protocols. Radically different missions.
The transition to a service orientation is going to be a difficult one for controls engineers. This sort of over-integration was understandable when low level protocols were all that was available. It was appropriate when there was no higher level interface to the control systems. It was appropriate when hand-crafting the enterprise interface, a process requiring someone skilled in translation between the worlds of production and controls, was the most difficult step.
But it is a bad architecture, one forced upon the practitioner, as the lack of structural steel forced heavy stone massing on early high-rise buildings.
We need to stop oner-integrating control systems. We have interfaces now.