Abstract Techniques for achieving high levels of system reliability by means of fault tolerance usually involve some explicit form of error recovery. Errors, ie defective parts of a system state (eg inconsistent values in working storage), are caused by faults, and unless successfully recovered from will give rise to system failures, and hence to degraded system reliability. Faults can be of many kinds (hardware component failures, mistakes by users and operators, design inadequacies in hardware or software) and can cause many errors before one is detected and error recovery is initiated.