Separating Safety and Control Systems to Reduce Complexity

Abstract

Software applications run our modern world. Software is now used to control systems with definite safety implications; it is used to collect, store and disseminate information of incredible variety; it is a vital component of communication devices, and is indispensable in almost all current entertainment and art forms. We often bemoan the fact that software dependent systems are not more dependable, but that loses sight of the fact that software engineering is a relatively young discipline, has made incredible advances—and is often used because alternatives cannot cope as well with the complexity of modern systems. A common theme in Software Engineering, and in this book, is that high complexity adversely affects our ability to develop low defect systems. Computer Scientists and Software Engineers have evolved techniques to improve our ability to develop dependable and safe complex systems. However, there is an alternative to conquering complexity—and that is to avoid it wherever possible! As an example, Canadian nuclear regulation states that safety systems in nuclear power plants have to be completely separated from the control systems in that plant, and isolated as much as possible from each other. Similar regulation is actually common in other countries as well. This is a special case of the general principle of separation of concerns. We propose a new principle, called the conservation of complexity, as a basis for addressing separation of concerns. This principle states that there is an inherent minimum complexity in any system, and that we cannot reduce it no matter what techniques we use. We can, of course, increase the complexity in our solution through inappropriate design. Hence, to address issues of dependability in today’s critical systems, we must use separation of concerns as a way of reducing the complexity of the critical aspects of such systems.