Software fault tolerance

Software design faults are a cause of major concern, and their relative importance is growing as techniques for tolerating hardware faults gain wider acceptance. The application of fault tolerance to design faults is both increasing, in particular in some life-critical applications, and controversial, due to the imperfect state of knowledge about it. This paper surveys the existing applications and research results, to help the reader form an initial picture of the existing possibilities, and discusses in an orderly fashion the design options available for using software fault-tolerance in a design. The decision to employ software fault tolerance, and in which form, cannot be based on a rigourous reliability evaluation, since both experimental data and proven models are lacking (this situation is common to other popular engineering techniques). On the other hand, some software fault tolerance techniques are intuitively attractive as a means for improving reliability, and experimental evidence, albeit limited, supports this idea. The specific combination of techniques and design options in any individual software system must be dictated by their suitability to the characteristics and requirements of the different parts of the system. The discussion provided here offers a general guide for these choices, with extensive references to the appropriate literature.