Using fault injection to assess software engineering standards

Abstract

Standards for quality software are increasingly important, especially for critical systems. Development standards and practices must be subjected to quantitative analyses; it is no longer adequate to encourage practices because they "make sense" or "seem reasonable." Process improvement must be demonstrated by a history of improved products. Fault-injection methods can be used to assess the quality of software itself and to demonstrate the effectiveness of software processes. Fault-injection techniques can help developers move beyond the practical limitations of testing. Fault-injection techniques focus on software behavior, not structure; process-oriented techniques cannot measure behavior as precisely. Fault-injection methods are dynamic, empirical, and tractable; as such, they belie the notion that measuring the reliability of critical software is futile. Before focusing too narrowly on the assessment of software development processes, we should further explore the measurement of software behaviors.