Implementing fault tolerance with an attribute and functional basedmodel

摘要

Programs constructed using techniques that allow software oroperational faults to be tolerated are typically written using animperative computational model. Here, an alternative is described inwhich such programs are written using an attribute and functional basedmodel called FTAG. The basic model is introduced, followed by adescription of mechanisms that allow a variety of standard faulttolerance techniques to be realized in a straightforward way. Techniquesthat can be accommodated include replication and checkpointing totolerate operational faults, and recovery blocks and N-versionprogramming to tolerate software faults. Several examples are given toillustrate these techniques, including a replicated name server and afault-tolerant sort that uses recovery blocks. These examples also serveto highlight the advantages of FTAG when compared to traditionalimperative approaches, such as a declarative style, separation ofsemantic and syntactic definitions and the simplicity of a functionalfoundation. Finally an outline of how this model can be implemented in acomputer system containing multiple processors is given