If you are interested in mathematical models of the dependability of hardware and software systems, you should certainly read this paper. The authors’ conclusion that “it is theoretically possible to perform reliability and availability evaluations of systems accounting for hardware and software” appears to be valid. Nonetheless, utilization of these theoretically possible results awaits the development of techniques to estimate parameters such as failure rates, which are especially hard to obtain for software systems.
Early in the paper is an insightful contrast between the places of hardware and software evaluations of reliability in the design process. A goal of the paper is to help incorporate the evaluation of software reliability into the design process. Classical hardware reliability is extended so it can be interpreted from both hardware and software points of view. This extension is called “X-ware.” The use of the models and formulas that the authors derive requires estimates of failure rates. Carrying out experiments that estimate these rates is relatively easy for hardware, because it is relatively easy to duplicate the kinds of input that cause hardware failures in practice. This is not true of software failures.
The authors also consider maintenance policies using the same framework they used for design considerations. Once again, the difficulty is in estimating the parameters that appear in the formulas in order to arrive at appropriate policies and to estimate the consequences of using them.