Reliability Design Of Complex Software Systems Under A Bounded Resource Environment

Reliability design has become a highly active and important area of software engineering. With the wide application of large-scale complex software systems in electrical power systems, railway transportation, aerospace, national defense security, etc., people have become increasingly demanding on the reliability.By introducing the related domain knowledge and constraints, the dissertation mainly focuses on how to make the reliability design of complex software systems (CSS) more effective. Specifically, these addressed problems can be divided into three aspects:reliability optimization, testing resource allocation and quality evaluation.The main contributions in the dissertation can be summarized as follows:(1) The dissertation elaborates the signification of the studies on the reliability design of complex software systems, points out the relationships among the reliability optimization, testing resource allocation and quality evaluation, and gives the main objects of those studies. Moreover, the related researches are discussed and the existing problems are noted simultaneously.(2) Reliability optimization plays an important part during the design phase of software, and has attracted increasing attention in recent years. The dissertation deals with a reliability optimization problem for multiple softwares with budget constraints. To tackle such problem, first, the dissertation formulates the architecture-based multi-software budget-constrained reliability-maximization model, in order to allocate the reliability to different modules to maximize the reliability of all softwares with a limited budget. Then, considering the uncertainty and incompleteness in practical engineering, Dempster-Shafer Theory (DST) is adopted to determine the global relative weights of each element in every level in the hierarchy of multi-software reliability. In addition, an algorithm based on differential evolution and encoding repair is developed to solve the proposed model. Finally, comparative results show that the proposed model and algorithm are efficient and may promote the reliability design level and efficiency of the development of complex software systems to a certain extent.(3) Software testing is a stage with the most time and resource consuming in software development. In complex software systems, since the system reliability may change over time, if the testing resource is allocated according to the initial established scheme, a big waste of testing resource may take place. To solve such problem, this dissertation is addressed in a testing resource dynamic allocation problem with limited testing resource in complex software systems. For this purpose, first, the definitions of testing resource, reliability of CSS and testing cost are given. Then, a testing resource multi-objective dynamic allocation model is presented to distribute the testing resource to different modules in order to maximize the reliability and minimize the testing cost with limited testing resource. Besides, an algorithm based on the DE and one-dimensional integer vector coding as well as improved population initialization is developed to solve the dynamic model. Finally, comparative results show that dynamic allocation of test resources can greatly save the consumption of system test resource and improve the efficiency of the development of complex software systems.(4) The evaluation of software quality is another key issue in software reliability engineering. The dissertation is devoted to an uncertain and comprehensive evaluation of software quality based on the fuzzy soft sets theory. In the process of evaluation, experts’ preference is considered, and each expert is allowed to have different sets of personal evaluation index, then the fuzzy soft sets theory is used to fuzzify each expert’s evaluation information. Thereinto, what the most important is how to use the fuzzy soft sets to deal with the experts’ different preference, fuse different evaluation information and obtain the results of the comprehensive evaluation of software quality. Finally, a verifiable example is given to show that the proposed method can not only express the subjective judgment of experts flexibly, but also deals with the uncertainty of information effectively and reasonably, which accords with the thinking and judgment process of people.