Uncertainty Explicit Assessment of Off-the-Shelf Software: Selection of an Optimal Diverse Pair

Abstract

Assessment of software COTS components is an essential part of component-based software development. Sub-optimal selection of components may lead to solutions with low quality. The assessment is based on incomplete knowledge about the COTS components themselves and other aspects, which may affect the choice such as the vendor's credentials, etc. We argue in favor of assessment methods in which uncertainty is explicitly represented (`uncertainty explicit' methods) using probability distributions. We have adapted a model (developed elsewhere by Littlewood, B. et al. (2000)) for assessment of a pair of COTS components to take account of the fault (bug) logs that might be available for the COTS components being assessed. We also provide empirical data from a study we have conducted with off-the-shelf database servers, which illustrate the use of the method