SOFTWARE RELIABILITY GROWTH MODELS BASED ON NHPP FOR MULTIPLE TYPES OF FAILURES

Abstract

With the progression of software innovation, it has been observed that one of the most important parameters which controls the software reliability enhancement is the testing Effort function (TEF). In this paper, effects of testing efforts on NHPP based software reliability model has been studied. The sensitivity analysis has been provided to examine the effect of system parameters of the developed model on the reliability of the software, mean value function and on cost function. The maximum likelihood method has been used for estimating the parameters of the model. A warranty cost model is also framed for evaluating the optimal release policy. The analytical results obtained are supported by numerical illustration. Innovative advancements are quick in an industry, to stay aware of the bit by bit expanding race and to get upper hand, the product advancement associations need to acquire the right information about software, use it competently and pass it to the next generation. Since software reliability is a key component of good programming, it is possible to decouple quantitative measures for the reliability of programming frameworks from models of software reliability. Software reliability is commonly understood to be the likelihood that a software system will successfully fulfil the task that has been assigned to it in a specific environment for a predetermined number of inputs, assuming that the hardware and the input are both error-free. One of the primary factors contributing to the software quality is its reliability. The three-phase software reliability growth model discussed in this research includes test coverage in an adequately debugged environment. We construct two models for the growth of software reliability that take complete debugging and testing coverage into account. The findings of the sensitivity analysis show that the parameters of the formulated models affect both mean value function (MVF) and the reliability of the software. GA, and PSO methods has been used to optimize the cost of the software. The stability or life of a software system with different capacities is referred to as software reliability. The software quality is the most important consideration while designing a software system. Software quality is determined by a variety of criteria, including software reliability, efficiency, testing abilities and cost considerations. In this paper, we deal with two different SRGMs (software reliability growth models) based on NHPP (Non-homogeneous Poisson process). We develop the models for inflection S-shaped context by considering time independent fault content factor for perfect debugging environment whereas another model v differs in an imperfect debugging environment by involving time dependent total fault content factor. We suppose that once a software problem is identified (removed), immediate debugging begins, and that either the total number of faults is detected (removed) with probability p1(p2) or the total number of faults remains constant with probability (1−p1)(1−p2), i.e., p1 % of the faults can be identified successfully and p2 % of the faults may be removed as well during the software testing phase. The intended SRGMs' results are used to assess the software's reliability. We further compare the performance of both the software reliability growth models. The findings show that the model can perform better in terms of fitting and prediction. In the current era of modern technology, a human cannot think to survive without software as such key area of attention of software manufacturer is to produce bug-free software and maintain the reliability and compatibility with human activities dependent upon software embedded devices. The manner in which the software will perform in a random field environment is a very major issue to study. Taking effects of random field environment into account, in this paper we develop a generalized software reliability growth model (G-SRGM) with generalized fault coverage function. For demonstrating the better performance of the proposed model, two data sets are taken and computational results of proposed models are compared with the existing models using Least Square Estimation (LSE) technique in MATLAB software. The three goodness-of-fit criteria such as the sum of square error, R-square, and root mean square error are also used for comparison. Chapter 1 is an introductory in nature and provides the basic definitions, background and motivation of the work presented in this thesis, along with basic definitions and preliminary concepts of software reliability growth models. Chapter 2 provides the literature survey on the application of Software reliability growth models (SRGMs) incorporating various factors for assessment of the software’s reliability. Chapter 3 explores a generalized software reliability prediction model for module-based software incorporating testing effort with cost model. Chapter 4 proposes a generalized two and three phase software reliability growth model incorporating testing coverage with cost model in a perfect debugging environment. vi vii

Chapter 5 presents a software reliability growth model in perfect and imperfect debugging environment.

Chapter 6 explores a generalized software reliability growth model for Software System Operating in Random Environment.

Chapter 7 summarizes the work, present important conclusions, and discusses the future research endeavors. The major contributions of this thesis along with concluding remarks in theoretical and application facets are presented in this chapter. The results obtained in all the chapters indicate that the performance of SRGMs can be improved significantly by integrating it with either random field environment, test-coverage function or imperfect debugging process. The maintenance cost of the software can be minimized by using the soft computing techniques like- nature inspired algorithms for optimization.