End-Host Middleware Approach of Fault-Tolerant, Real-Time Ethernet for Distributed Process Control Networks

摘要

We present an efficient middleware-based fault-tolerant, real-time Ethernet (FTRTE) developed for process control networks. Our approach is unique and practical in the sense that it requires no change to commercial off-the-shelf hardware (switch, hub, Ethernet physical link, and network interface card) and software (commercial Ethernet NIC card driver and standard protocol such as TCP/IP), yet it is transparent to IP-based applications. The FTRTE performs failure detection and recovery for handling multiple points of network faults, eliminates packet collision on the network using TDMA scheduling, and supports an efficient load balancing. Our experimentation shows that FTRTE performs efficiently, achieving less than 1-ms end-to-end swap time and less than 2-secfailover time, regardless of the concurrent application and system loads. In this paper, we describe the FTRTE architecture, the challenging technical issues addressed, our performance evaluation results, and the lessons learned in design and development of such an open-network-based fault-tolerant, real-time network.