Design of a distributed fault-tolerant computer architecture applied to the traffic control system IVMS

This paper presents the design of a fault-tolerant computer architecture for the traffic control system IVMS (Intelligent Variable Message Sign). In this project, IVMS stations with point-to-point communication links are to be installed along highways so as to control the traffic flow, yielding homogeneity. The principal design goal is to achieve high system availability at foul cost; the availability is calculated by Markov models. Comparing several candidate architectures, we have found that a distributed IVMS system based on simplex computers without static redundancy, is the most interesting architecture, because it allows degradation. We discuss fault tolerance measures for transient failures, which occur much more frequently than permanent failures, and propose solutions for the stable storage concept and for network partitioning