SREPT: A Tool for Software Reliability Estimation and Prediction

Although several tools have been developed for the estimation of software reliability, they are highly specialized in the approaches they implement and the particular phase of the software life-cycle in which they are applicable. Also the conventional techniques for software reliability evaluation, which treat the software as a monolithic entity are inadequate to assess the reliability of heterogeneous systems. We present here, a tool called Software Reliability Estimation and Prediction Tool (SREPT) that seeks to address these limitations. Architecture of SREPT: The interested reader is referred to [1], [2] for a detailed description of the tool. The SREPT GUI has been implemented using the Java programming language. Figure 1 shows the high level architecture of SREPT. In the pre-testing phase, SREPT accepts software product/process metrics as input, and produces an estimate of the number of faults in each module using either the fault density approach or regression tree modeling technique. During the testing phase, SREPT offers the user an option of doing analysis based on the failure data using the enhanced non homogeneous Poisson process (ENHPP) model to predict the failure intensity, number of faults remaining, coverage and reliability. When using the failure data to obtain reliability predictions, the ENHPP model in SREPT currently uses four coverage functions, namely, exponential (Goel-Okumoto), Weibull (Generalized Goel-Okumoto), Sshaped, and log-logistic. These belong to the class of finite failure non homogeneous Poisson process (NHPP) models. The ENHPP model in SREPT also offers a mechanism to combine software metrics, failure data and coverage based approaches. Various optimization engines to compute release times of the software are also provided. Unlike most models that assume instantaneous and perfect debugging, SREPT allows the user to analyze the effect of non-zero debug times to reflect more realistic scenarios. SREPT is planned to accept the architecture of the application modeled as a discrete or continuous time Markov chain, directed acyclic graph, stochastic Petri net, product form or non-product form queueing network [3]. The architecture of an application will be combined with the failure behavior of its components to provide architecture-based software reliability and performance predictions.