Software System Reliability Analysis

Establishing the reliability of embedded software-based systems is increasingly important as the DoD's dependency on software grows and the size and complexity of software systems increase. Currently, no theory is in general practice that adequately explains the behavior of software systems reliability in a consistent and repeatable way. Currently, we cannot point to a best practice (although John Musa and other experts would argue that operational profile testing is the SRE best practice). At the conclusion of Phase II of this STTR, the CC/UMD team is confident that we will not only have produced such a best practice, but that the practice will be publicized and widely known among DoD management and practitioners, and that these people will promote the best practice. The best practice will incorporate an advanced theory for explaining software systems reliability. The theory will be based on well understood principles of software operation. The software reliability theory is by nature fundamentally different from hardware reliability. Hardware reliability is characterized by physical changes in system parts which result in a breakdown in one or more components or sub-systems. Software does not fail in that manner. Software systems fail as a result of design flaws, design limitations, or simply unknown requirements.