RELIABILITY ANALYSIS USING SOFT COMPUTER TECHNIQUES

Abstract

With the growing advances in the digital world, software development demand from industries is growing at an exponential rate. Due to enormous demand and lack of time and budget, software companies are not able to develop fault-free software. Latest tools and techniques have been applied for the development of defect-free software, but still, it is impossible for the software developers to develop defect-free software practically. Software must go through exhaustive testing and debugging, which requires time and money to enhance the reliability. The occurrence of fault is inevitable in the current demand of software. There should have some means to avoid software failures so that devastating losses whether related to life or any other field could be evaded. According to IEEE standard 729 [1] , reliability is the most significant quality aspect of the software. If we could measure the reliability of software under development, better we can predict whether the software would be operational in the future or not. Reliability estimation process must be precise to provide information to the manager like what should be the release time of the software and amount of man-hour consumption etc. while developing any software. Software reliability models are only the ways in order to simulate software reliability estimation curve to predict the reliability of the system under study. Numerous reliability estimation models for software have been developed, and all are working on specific applications, specific environments, datasets and assumptions made by them. Initially, a systematic survey of software reliability models is done that shows how a new model is evolved from other models on the basis of their assumptions and attributes. Among the available research in software reliability model development [2] [3], It is found that, all developed models are basically made for reliability estimation of software‟s developed under traditional waterfall software development life cycle process. As estimation of software reliability is primarily reliant on the process of software development. Software‟s developed using latest software development life cycle approach can capture and implement all the user requirements within time and budget. This work focuses on development of new software reliability estimation model that incorporates iterative software development life cycle process by replacing earlier waterfall based development process. It assumes imperfect debugging during each of iteration. All the latest iterative software development life cycle processes may be used to predict software reliability by applying proposed software reliability estimation model. Existing failure rate models cannot be applied to the current software development methodology. The proposed model takes care of complexity and paradigm shift of iterative based software development process by introducing modulation factor. viii | P a g e Further, Keeping in mind multi-release policies of open source software, this thesis work also proposed another model that introduces a new testing effort factor based fault content function. This factor is showing the change in fault content function with the amount of testing effort in each version of open source software development. Altogether it is reflecting complete testing effort functionality added or upgraded in each version of the software. Effort factor changes its value according to the effort coefficient which takes its value from zero to one by assuming that complexity value added in each version increases from lower to higher. Effort factor has a change in its value, depending on whether it is a minor or a major release. For precise estimation of open source software system reliability, there is a need to have a parameter estimation method that could provide optimum parameter values of models. Classical methods [4] of parameter estimation are based on number of constraints. An alternative to these classical mathematical optimization methods is nature-inspired optimization algorithms for solution of the non-differential, non-linear and multi-modal problems [5]–[7]. The research work move ahead and also focuses on the development of new hybrid swarm evolutionary algorithm for software reliability model parameter estimation. A new algorithm based on the concept of ecological space [8] , method of differential evolution and intelligent behaviour of artificial bee colony for optimizing the parameter values [9][10]. The exploration capability in ABC algorithm has been improved by introducing the concept of ecological space. Ecological space is one of the important factors for evolution and reflects the expansion of individual bee in search space. DE technique provides the diversity of bee's population and faster convergence. The proposed algorithm has been tested with four standard failure datasets. Proficiency of proposed algorithm is compared with other well-known algorithms. Proposed algorithm is very much effective in a field of software reliability estimation and would be a competitive one among meta-heuristic optimization algorithms. Finally the thesis is concluded with the perspective of future work.