TESTING AND OPTIMIZATION OF SOFTWARE RELIABILITY GROWTH MODELS

Abstract

With the advancement of technology, software has become significant both for industrial and scientific purposes. There is probably no other human-made material which is more omnipresent than software in our modern society. In particular, science and technology demands high-quality softwares for making improvement and breakthrough in various spheres of life and development process of the nation. A primary objective of the software developers in the software testing is to minimize the number of errors and increase the quality of the software. Software reliability represents a customer-oriented view of software quality and relates the practical operations rather than static of real time software embedded systems. The main aim of testing phase of the software development life cycle is to detect/isolate and remove the faults and hence increase the reliability. The key concern of this process for the testing team is when to stop testing and release the software for the operational use. The importance of software testing and reliability is due to the continuing growth in the software industry, higher expectations and demand for new and updated versions of the softwares. The main objective of our research investigation is the development of the software reliability growth models (SRGMs) based on non-homogeneous Poisson process (NHPP) and Markov process. The software system has been studied to determine the optimal release policies of the concerned software models. We have employed some effective techniques to improve the efficiency of the software testing and prediction of various reliability characteristics. The whole work of the thesis is divided into various chapters dealing with some realistic features related to SRGMs. The investigation on the SRGMs includes some worth-noting features, viz. error detection, imperfect debugging, testingefforts, discrete time software reliability predictions, fault reduction factor, modular software reliability, rejuvenation, etc. While analyzing the generic models, various performance measures are evaluated. The worth-mentioning indices evaluated are testing and operational reliability of the system, availability, mean time to failure, cost function, optimum release policy, expected number of faults in the system, etc. In our study, various methods of reliability theory and optimization have been employed to find the solution of the concerned problems. The coding of the computer programs to validate the analytical software reliability models has been done using iv software ‘MATLAB’. The numerical results computed based on analytical formulae are presented graphically and in tabular form to explore the sensitivity of the system parameters. Some models of NHPP have been examined by using soft computing approaches, viz. neuro-fuzzy method and genetic algorithm. We have organised the present thesis in eight chapters as follows: Chapter 1 is introductory and describes the preliminary basic concepts of the software reliability growth models. An overview of the continuous SRGMs, discrete SRGMs, Markov software reliability models and some methodological aspects related to the software testing process are presented. The optimal release policies of some existing models are briefly reviewed. A brief literature review on the topics on SRGMs having some additional features, namely imperfect debugging, testing effort functions, change point, cost optimization, discrete models, Markov models and rejuvenation models, etc. has been facilitated. The outline of the thesis and noble features of the investigation done are also mentioned. In Chapter 2, a new scheme for constructing the software reliability growth model (SRGM) based on non-homogeneous Poisson process (NHPP) with imperfect debugging, time-variable fault reduction factor (FRF) and multiple change points is proposed. We have established mean value function for three types of FRFs, which is further employed to determine the optimal testing time. The maximum likelihood estimation is suggested to calculate the unknown parameters of the proposed model. Numerical illustration has been provided to explore the effect of various parameters on the software reliability and total expected cost. In Chapter 3, we study the software reliability growth models (SRGMs) with imperfect debugging, testing effort function (TEF) and time variable fault reduction factor (FRF). The concept of Weibull type testing effort function along with multiple change points has been incorporated. The mean value function is established to explore the software reliability indices. The optimal release policy based on cost and reliability are also suggested. In Chapter 4, we are concerned with the warranty cost analysis of the software reliability growth model (SRGM) by incorporating the imperfect debugging, fault reduction factor (FRF) and multiple change points. The optimal release policies based on cost and reliability criteria are established for determining the optimal release time of the software. v In Chapter 5, we study the SRGM by incorporating the testing effort function and imperfect debugging concepts. The operational reliability and testing reliability are taken into account to suggest the optimal testing policies. The sensitivity of different parameters has been numerically examined and compared with the neuro fuzzy approach. With the help of genetic algorithm, we find the optimal release time of the software. In Chapter 6, we investigate a module based SRGM by incorporating the concepts of imperfect debugging and generalized modified Weibull (GMW) testing effort function. We suggest the optimal release policy to determine the software release time subject to cost reliability criteria. The sensitivity of different parameters has been examined numerically. In Chapter 7, the discrete software reliability growth models with different types of faults have been explored by considering imperfect debugging. The multi-modular software systems with and without testing effort functions have also been studied. The optimal release policies are proposed for determining the optimal test runs of the software testing based on cost and reliability criteria. In Chapter 8, two types of Markov models for the software reliability prediction are discussed. In the first model, we incorporate the concepts of error generation, imperfect debugging and reboot delay. The matrix method has been suggested to determine the queue length of the number of faults in the software. Various performance measures such as reliability, mean time to remove all faults, etc. are obtained. Further the sensitivity and relative sensitivity analysis of some reliability indices are presented. In the second Markov model, the four levels of software rejuvenation policy are proposed. The down time cost and other performance metrics are also calculated. The wide applicability of the software reliability modelling has motivated us to analyse the reliability and optimization aspects of SRGM in different frame works. We have developed continuous and discrete SRGMs and discussed optimal release policies. We have also studied Markov models to suggest the software reliability indices and rejuvenation policy. The performance measures such as mean value function, failure intensity function, testing and operational reliability, mean time to failures and total maintenance cost etc. obtained may be helpful to the system designers and decision makers for improving the reliability growth of their software systems. The research work done provides the valuable insight to the software developers and decision makers in predicting the performance of the software systems