基于FC光传总线的分布式容错系统组通信协议

分布式容错系统常用于构造高可靠的关键应用,其核心构件组通信协议(GCP)实现了带有可靠性语义的多对多的通信原语。本文主要讨论基于光纤通道(FC)总线用于分布式容错的组通信协议,探讨了使用冗余消息技术实现实时、可靠数据传输的方法,提出了将数据和控制消息分开处理来实现数据输入实时性和控制消息按因果顺序提交的算法,力求获得最大程度的并行性。提出并设计了基于领导者-跟随者的故障处理机制,利用两阶段事务处理协议实现故障处理中的视图同步和状态切换,给出了协议的设计原理和实现框架。最后给出了该协议在基于PowerPC单板计算机的三余度容错飞行控制计算机系统中的运行结果,并对该结果进行了简要分析。

Fiber Channel Fly-by-light Bus Based Group Communication Protocol for Distributed Fault-tolerant System

Distributed fault-tolerant system is always used to achieve high availability and reliability in safety-critical applications,whose core component is group communication system.Group communication intends to im-plement multi-party communication with reliability semantics.Group communication protocol(GCP) for fault-tolerant distributed systems is discussed based on fiber channel(FC).It combines the redundant message technique and time-out retransmits strategy to improve communication reliability.This article proposes to deliver separately the data and control messages to achieve trade-off between reliability and real-tameness.A leader-follower based failure handling mechanism is used to implement view synchronization by two-phase transaction protocol.The framework of protocol is presented.The final part is the result and analysis of the protocol running in a triple modular redundant(TMR) flight control system.