Software fault tolerance

Abstract

Methodologies for analysing the reliability of complex systems and techniques for making such systems tolerant of faults, thus increasing the reliability, are well established. In software, the emphasis has been on improved software construction techniques or “software engineering” to reduce latent errors; there has also been work on techniques to introduce redundancy into software systems. The majority of work on fault tolerance has concentrated on “anticipated faults”, i.e. faults which the design can anticipate and hence “design” in tolerance. A much more difficult and insidious problem is that of faults in the design of the system. These are by definition “unanticipated” (and unanticipatable) faults. Design faults can occur both in complex hardware and software, but are more common in software, and much of the effort of software engineering has been directed towards reducing design faults, i.e. unanticipated faults.