Fault tolerant distributed process-control systems

Abstract

Distributed processing is a cost-effective way of enabling critical process-control systems to react to emergencies in real-time. We discuss methods of making the system robust based on the use of self-stabilizing fault tolerance techniques. The control entities are organized in a hierarchical control structure. Upper level nodes deal with more abstract information and decompose the task of the system into a series of intermediate subtasks. These subtasks are further decomposed into a set of decentralized subtasks which are assigned to lower level nodes which operate without excessive coordination. Decentralized decomposition also makes the system inherently fault tolerant and simplifies the code. Other nodes are made fault tolerant using telescopic replication which provides economical redundancy without compromising system reliability.