An industrial survey on the influence of process and product attributes on software product reliability

Abstract

Software reliability is a major attribute for software product and can be considered as one of the major performance parameters. Software Reliability unlike hardware reliability cannot be considered merely as function of time, although researchers have come up with models relating the two. In literature, numerous models on software reliability have been proposed but they seem to have limitations in being used to predict reliability in software industry. As on today, we do not have good quantitative methods developed to represent Software Reliability without excessive limitations. Various approaches can be used to improve the reliability of software, however, it is hard to balance the development time and the budget with required levels of software reliability. This paper quantitatively and qualitatively analyzes the data from an industrial survey, with focus on current practices and their dependency on the estimation of reliability in software product development environment. Analysis focuses on perception of predicting software reliability in different categories of respondents. Categorization of respondent is based on phase of development, their profile (as user, developer etc.), and application domain. While confirming some commonly known facts, our findings also show discrepancies between our perception and actual facts in some cases. The objective of this is exercise is to find out input parameters making impact on reliability. Though it is field perception, we have identified them from practitioners and taken as reference for conducting future experiments and literature survey. We believe that continued research efforts are essential to provide guidelines for reliability estimation process to take care of important but hitherto ignored factors, thus improving relevance and accuracy of reliability predictions.