Capability analysis of distributed switching systems in interprocessor communications

Abstract

Recently developed switching systems all use distributed control to some extent. Among the advantages claimed for distributed control are higher reliability, flexibility, and increased traffic capability. An analysis is made of the traffic capability of a fully distributed switching system, taking into account the overhead of interprocessor communication. The author considers a clocked event interprocessor messaging scheme (e.g. polling) in a two layer architecture, and shows that there is a given configuration for which the capability reaches a maximum; as more processors are added, the increased overhead actually results in a decrease in traffic capability. An analysis is also made of an improved interprocessor communication scheme in which the capability of the distributed architecture can in principle be increased by appropriately adding processors. Extensions to this basic model are presented along with suggestions for future work.<>