Applicability of decentralized/centralized control procedures in distributed processing system development and operation

It was inevitable that the performance attained by a single processor configuration would not be sufficient to accommodate the requirements emanating from a rapidly expanding discipline such as computer science. Distributed processing configurations have a distinct advantage over a centralized approach in their adaptability to a constantly changing environment. There are many other desirable attributes of distributed processing, including enhanced survivability (with the consequent increase in system availability), reconfigurability, and recoverability. Many views have also been expressed on the desirability of a distributed philosophy to the preclusion of any centralized approaches. These enhanced capabilities necessitate effective production and operational pistol procedures in an environment which is by definition intricate and complex. It is therefore considered essential that all aspects associated with distributed processing system design, development, production, operation, and maintenance are subjected to an overall philosophy applicable to all its phases. To this extent, two models are proposed for use in the demonstration of the two philosophies which can be adopted, i.e., decentralized and centralized. Here the aim is to highlight the need, at the higher echelons of the control hierarchy, for a centralized philosophy for the effective coordination of locally identifiable subordinate control functions.