Software Engineering Institute

As will be shown in this paper, there is a fair amount of uncertainty among system engineers about how to determine the impact of software on overall system reliability, and this uncertainty is especially clear when attempting to evaluate the impact of software on system of systems (SoS) reliability.

This paper discusses the uncertainty that is evident today, based on presentations given at a reliability, availability, maintainability, and testability (RAM-T) summit for a large system of systems. Clearly, new approaches (or at least, better guides) are needed to deal adequately with software aspects of system and SoS reliability.

A few suggestions are provided in this paper (the need for giving software failures consideration when doing system-level FMECAs, the need for specifying failure definitions and scoring criteria at the SoS level (not just at the constituent system, or platform, level), and the need for Software Reliability Improvement Programs undertaken during system design), but the main point is that it is not enough to simply formulate software reliability goals or to collect statistics on detected defects.