Maximal Availability of Tree Structured Fault Tolerant Systems

Abstract

There is a field in telecommunications which sets very strong conditions in two aspects, capacity (speed) and reliability: this is telecommunications switching. It is stated in the literature that the largest software systems ever constructed for civil applications are the control programs for large switching exchanges [1]. About 70 to 90 percent of this software is related to maintenance and operation, a large part of this directly supporting fault tolerance. Fault tolerance and reliability is understood in this field in a somewhat different manner than e.g. in “traditional” application computer systems. There is a very strict request concerning continuous operation 24 hours/day (including maintenance and other operation related activities!), which cannot be performed by even the most ideally fault tolerant multiprocessor system. International recommendations refer always to reliability in the sense of partial availability of the complete system. This means an availability of all services guaranteed by the exchange at a specified (high) percentage of all serviced points (subscriber terminals, etc.)