Software Reliability Analysis of Multistage Projects

A Multistage Software Reliability analysis is performed. We propose to use an increasing failure rate model at the very first stage of development and testing and a Software Reliability Growth model at the last stage before release, the middle stages usually presents jumps generally due to the adding of new code. We remark then the importance of the increasing failure rate model applied at first in order to predict when the reliability growth stage starts. An excessive fast increase in the failure rate could alarm Engineers as to adjust the development and testing processes. Two real projects with similar metrics are analyzed. We found that the logistic model fits the best the first stage, in agreement with a previous study. We show that clustering over the failure per day points results a useful tool to separate the stages. Despite the analyzed projects are from quite different sources, we find similar behaviors, that shows as the wide ranging of our analysis as its applicability.