This paper introduces the architecture of a large scale distributed computing system for the Area-wide Real-time h f i c Control system ARTC) [l, 2, 31. In each other through the communication network for real-time adaptive trafic control. For reliable service, the ARTCsystem is designed to be fault-tolerant at the 1/0 component level, processor level, and the network level. A modular system architecture is proposed for the A R T C signal controllers to support fault-tolerant design strategies. A jlezible, cost-effective networking strategy is presented to support different level network redundancy. In order t o estimate the expected service quality of the A R T C control system and to study the reliability impacts of signal controller failure rate, the repair eficiency, and the degree of system redundancy, an analytical model is developed.