A fault-tolerant protocol for railway control systems

Abstract

Railway control systems are largely based on data communication and network technologies. With the adoption of Ethernet-IP as the main technology for building end-to-end real-time networks on railway control systems, the requirement to deliver high availability, quality and secure services over Ethernet has become strategic. Critical real-time traffic is generally penalized and the maximum restoration time of 50 msec sometimes is exceeded because of real-time applications hangings, so passengers' safety could be committed. It occurs on more than twenty percent of critical fail tests performed. Our main goal is to minimize restoration time from the application point of view. This article describes a protocol to improve critical real-time railway control systems. The algorithm designed gives us fast recoveries when railway computers fail down. The protocol permits to manage the railway control system from every computer in the network mixing unicast and multicast messages. Simulations obtained for a real railway line are shown. We have reached excellent results limiting critical failures recoveries to less than 50 msec.