The need for measurement based reliability evaluation

Abstract

Defensible quantitative assessments of the reliability and availability of computer systems including software are possible. This paper characterizes the need for quantitative empirically-based dependability assessment, describes some of the previous work in this area and identifies problems. While there is still ongoing research in measurement-based analysis of computer dependability, the techniques developed in this area have achieved significant experimental results. Measurement-based analysis can also provide verification of assumptions and parameters used in the design models. The results are useful for designing and maintaining highly-dependable computer systems intended for use in safety critical applications.