A framework for the estimation of OO software reliability using design complexity metrics

Abstract

Reliability of a software product essentially denotes its trustworthiness or dependability. The stakeholder's requirement is normally a robust and highly reliable software product. Reliability is one of the dynamic characteristics of software, which can be quantitatively measured by analyzing the failure data. In order to estimate as well as to predict the reliability of software systems, failure data need to be properly measured by various means during software development. Most of the software failures happen due to poor design quality. Controlled and well-monitored design with respect to reliability can be achieved by quantitative analysis and appropriate mapping of design metrics. This analysis leads to the prediction of reliability at the design stage of Software Development Life Cycle (SDLC) of Object Oriented (OO) software, which will provide early information for software architect about the quality of the design with respect to reliability. Chidamber and Kemerer (CK) design metrics play a dynamic role to understand the design aspects of object-oriented software. In this context, we have proposed a framework for predicting reliability at the design phase of software development life cycle. Through empirical analysis, a valid hypothesis is established relating the design complexity metrics and reliability. A functional relationship is established through polynomial regression technique for selected design metrics. With these equations, an estimation model called R-MODEL is formulated through weighted linear combination of multifunctional parameters, which predicts reliability for any module of the OO software project. The R-MODEL is evaluated and validated with commercial software, which is found to be satisfactory. This quantitative empirical analysis provides an idea for the percentage influence of design metrics on reliability at the design level. The approach is to make, measurement as an integral part of the software design process to assist software designer to predict the software reliability at early stage and to improve the product quality.