Invited Talk II

The widespread use of software systems and their ever increasing size and complexity induce many challenges to software developers and quality assurance practitioners. A fully integrated approach, based on qualitative and quantitative aspects, is needed to ensure that software dependability is correctly handled and the expected goals are reached for the final product. Dependability assessment, based on measurement, plays a vital role in software dependability improvement. Measurement encompasses both the observation of the software behavior during its development or operational life (i.e., field measurement) and controlled experimentation (i.e., experimental measurement). Field measurement requires the collection of data related to failures, maintenance, and usage environment, in order to evaluate measures such the overall software failure rate, the failure rates according to some specific (critical) failure modes, the components failure rates, and system availability. Controlled experimentation complements very well field measurement, particularly when considering Off-The-Shelf software for which, most of the time, no information is available from the development phase. The presentation will focus on dependability assessment, based on measurements. It will 1) outline current approaches to measurement-based dependability assessment, with examples from real-life systems, and 2) identify some research gaps.