A Comparative Study of Failure Data for Software Reliability Estimation
S. Charles Ilayaraja · 2013
Software reliability is one of the attributes of quality and its measurement is supported by Software Reliability Growth Models. Among many of the proposed SRGMs, some of the models have been widely used and few of them obsolete. The successful SRGMs are characterized by the accuracy of reliability estimation. In such models, the reliability estimation depends on the quality of failure data and its accuracy increases as the length of failure observation increases. In this paper, we have studied the role of failure data set in the reliability estimation by employing three different data sets on unified SRGM that has no distinction between failure observation and fault removal process. The selected SRGM is incorporated with two distribution functions one is exponential and another one is 2-staged erlang distribution. For easy implementation, we assume that the debugging process is perfect. That is the probability of error correction is one and no errors have been introduced during the debugging process. Three different cases of failure data are considered for our study such as database application software, web server, the interface of an operating system. The unknown parameters of each case are estimated by using SMERFS tool and the goodness-of-fit analysis is done in MATLAB environment.