An extension of goal-question-metric paradigm for software reliability

Abstract

The driving need in software reliability is to mature the "physics of failure" and design aspects related to software reliability. This type of focus would then enhance one's ability to effect reliable software in a predictable form. A major challenge is that software reliability, in essence, requires one to measure compliance to customer/user requirements. Customer/user requirements can range over a wide spectrum of software product attributes that relate directly or indirectly to software performance. Identifying and measuring these attributes in a structured way to minimize risk and allow pro-active preventive action during software development is no easy task. The goal-question-metric paradigm, discussed by Dr. Vic Basili (1992), is one popular and effective approach to measurement identification. However, in practice, additional challenges in using this approach have been encountered. Some of these challenges, though, seem to be alleviated with use of a reliability technique called success/fault tree analysis. Experience has shown that the goal-question-metric paradigm is conducive to the building of G-Q-M trees which may be analyzed using reliability success/fault tree logic.