Terminology and paradigms for fault tolerance

To familiarize the reader with the field of fault tolerance, this report discusses the most important terms and paradigms used in that field. After establishing a basic terminology, the fundamental techniques to achieve fault tolerance, i.e. the basic ways to employ redundancy, are identified. In particular, the role that fault hypotheses play in the design of a fault tolerant system is illustrated. To enable the development of formal methods for fault tolerance, the interaction between fault hypotheses and design decisions is analyzed in detail for two fault tolerant systems. The first is a stable virtual disk which is implemented using a number of unreliable physical disks. The second concerns a reliable broadcast protocol.