Abstract
Fault detection and removal play major roles in software testing as well as follow-up maintenance after software release. Over the past three decades, researchers have been studying fault detection removal using mainly the software reliability growth models (SRGMs) that are assumed to be a non-homogeneous Poisson process. In addition, many of the SRGMs simplify the models by ignoring the fault correction time in the software debugging process. However, the fault correction time should be considered in practical software development because when failures occur in the test phase, it takes time for the developers to identify the failures and remove the corresponding root causes. This thesis applied the queueing theory and introduced a finite server queueing (FSQ) model that took the fault correction time into consideration. The model was adopted to predict the possible fault detection and removal behavior as a quantitative method. We verified that the FSQ model helps developers to anticipate in the possible response time of each detected fault and the fault correcting efficiency. This assists project managers to set up reasonable schedules and allocate proper personnel resources. The obtained result showed that the FSQ model which incorporated the detection and removal processes exhibited an accurate prediction ability and high adaptability.