A Fault-tolerance Layer for Distributed Fault-tolerant Hard Real-time Systems

This paper describes the conceptual model for, and the implementation of, a software fault-tolerance layer (FT-layer) for distributed fault-tolerant hard realtime systems. This FT-layer provides error detection capabilities, fault-tolerance mechanisms based on active replication, and the interface between the application software running on a node of the distributed system and the communication services. Communication is based on the fault-tolerant time-triggered protocol TTP/C. The FT-layer handles all necessary information transfer across the TTP/C bus transparently. The conceptual model for the FT-layer is based on the DFR meta object model. This model is based on a separation of the three domains: value domain, time domain, and distribution domain. The DFR model supports a exible choice of the degree of replication of ne-grained software components according to the application-speci c dependability requirements. As the DFR model captures all the relevant design information explicitly, it allows the construction of powerful tools supporting the software development process. One such tool, called xOLT, analyzes the application software and generates the FT-layer automatically and without user intervention. Due to a novel treatment of domain separation and system factorization, the FT-layer generated by the xOLT meets the stringent performance constraints of application areas extremely sensitive to cost such as automotive electronics.