Abstract
Over the past 30 years, many software reliability growth models (SRGMs) have been proposed for estimation of software reliability growth. One common assumption of conventional SRGMs is that detected faults are removed immediately. In reality, this assumption is unreasonable. Therefore, we should not ignore the fault correction time. In this thesis, we will take the fault correction time into consideration and use a queuing theory in order to model the fault correction process during test and operational phases. In a test phase, a project may be found to be unstable and therefore may need many resources to find and fix the bugs. We propose an infinite server queuing (ISQ) model in order to predict reliability. In addition, the fault correction rate may change at some moments in time which are called change-points (CPs). Thus, we incorporate CPs into the ISQ model. During the operational phase, because the resources are limited and controlled, we use a finite server queuing (FSQ) model. This model can help developers measure the project reliability and thus estimate the number of debuggers needed for the project. We can also estimate the average number of faults in the queue and the system, as well as the response time and waiting time of faults. Experimental results show that the proposed models can predict the behavior of software development more accurately.