Real-time fault-tolerant communication in computer networks
Qin Zheng · Deep Blue (University of Michigan) · 1993
A network is required to provide the users with a convenient means of guaranteeing delay bounds in message transmission and making message transmission tolerant of network component failures. Solutions to this problem will greatly improve the quality of service of the contemporary computer networks and expand their application domains to such areas as distributed real-time controls and digital continuous-media (motion video, audio) transmissions. Our solution to the problem is to use a new transfer mode called a real-time channel which guarantees the timely delivery of messages like the circuit-switched transmission while preserving the high transmission efficiency of the packet-switched mode. We give a comprehensive coverage of this new transfer mode, from the fundamental deadline scheduling theory and real-time channel protocols, to the detailed hardware implementation. Using the spatial redundancy of a network topology, real-time fault-tolerant communication is achieved by enhancing the basic real-time channels to be Single Failure Immune (SFI) or Isolated Failure Immune (IFI). Backup channels can also be used to increase the reliability of real-time channels. The issue of establishing real-time channels over shared-medium Local Area Networks (LANs) is then discussed, which is of practical importance since most end systems are connected to a LAN first and then to a point-to-point Wide Area Network (WAN). We also present a technique which can significantly improve the FDDI's capacity of supporting real-time traffic with a few simple modifications to its Medium Access Control (MAC) protocols. Finally, the application of our results for multimedia networking is illustrated through simulations.