Description

Software Reliability Engineering is an emerging discipline whose importance cannot be undermined. It is receiving unprecedented attention from researchers. In this thesis, it was endeavored to develop more practical resource allocation approach catering to different fault removal models under dynamic environment. A systematic technique for the software project managers to allocate the resources for detection and correction, when a company has limited budget has been developed. It is observed that initially the path of detection effort increases but after a certain point of time, it attains a saturation level. Also, when most of the simple faults have been debugged, there will remain some hard and complex faults though detection of these faults will no longer remain difficult due to learning curve phenomenon. Due to this, the path of correction effort initially decreases and then attains a saturation point.
GA is applied successfully to obtain optimum value of detection and correction effort when the company has limited budget. In addition to this, a trade-off between reliability and release time of a software project is examined. Using sensitivity analysis, we have observed the pattern of correction cost and variation in optimal allocation with respect to the parameters. We observed that, in a cut throat market, managers are better off underestimating the cost of correction left in the system, resulting in a choice of project duration that is below the desired level. In section 2, optimal policy has been developed for proposed resource allocation model for multiple releases under dynamic environment. Using GA, optimum value of testing efforts for each release is obtained for the particular case considered. From the comparison analysis, it is concluded that testing efforts and faults detected for each release of our proposed model & real data set are closely related to each other. Graphical illustrations are also presented displaying the above relationships.