Hardware- and Software-Fault Tolerance

The paper is devoted to the methods and architectures aimed at tolerating hardware faults and software faults. The paper is composed of three parts. The first part presents in a unified way the methods for software-fault tolerance by design diversity; the faults to be tolerated are discussed from two viewpoints: their independent and their persistence. Hardware- and software-fault tolerant architectures are defined and analyzed in the second section; implementation issues involved in design diversity as well as structuring principles are discussed; several architectures are proposed, aimed at tolerating single faults and consecutive faults. These architectures are analyzed from both the reliability and the cost viewpoints in the third part.