Research And Application Of Component-based Software Reliability Assessment And Allocation Model

With the development of modularization idea in the design of modern software systems,component-based software is one of mainstream software forms,widely used in various important systems.Reliability of those systems considered significant attention from academics and industry.Work of reliability assessment and allocation is part of early software reliability design and is throughout all the software development.It is of great significance for improving the reliability of software systems.The content of the component-based software reliability prediction model proposed in this thesis mainly includes the following parts:Firstly,aiming at the problem that most current reliability assessment methods analyze the system reliability without consideration of component impacts,a component impact calculation method with three parameters based on the idea of graph theory and Page Rank algorithm is proposed.The model focuses on analyzing the impacts of different components based on three parameters i.e.self-influence,failure influence and fault propagation influence,making the defined component failure behavior more consistent with actual software characteristics.The impact of each component is used for the system reliability assessment and optimization to obtain more accuracy evaluation results.The experimental results show that the proposed model is able to obtain a reasonable evaluation results that could be used for reliability optimization.The experimental results also show that the proposed model is able to to identify relatively critical components and help software designers allocate resources reasonably.Secondly,to deal with the diversity problem of software architectural styles in the software systems,a component-based software reliability assessment model based on component impact and architecture analysis is presented.Analyzing the characteristics of component interactions under different architectures,a state mapping mechanism is proposed to provide flexible transitions from components to system states.Then,a global state transition model is developed and the discrete-time Markov chain is used to predict the component-based software reliability.The component impact is discussed in the reliability calculation to better reflect the influence of different components on system reliability.The experimental results show that the new model is significantly accurate in reliability assessment compared with similar methods.Thirdly,a component-based software reliability allocation model based on the improved bird swarm algorithm is proposed.The current software reliability optimal allocation models lack the analysis of the impact of component importance on allocation results,and the original bird swarm algorithm is easily trapped in local optimization for handling with complex nonlinear optimization problems such as reliability allocation.Firstly,a three-parameter exponential cost function is used to transform the reliability allocation problem into a constraint optimization problem.Later,an optimal allocation model with the consideration of component impact is established.Furthermore,the original bird swarm algorithm is optimized by improving the initialization,introducing decision factor and enhancing boundary handling.The improved algorithm is applied to solve the new model.The experimental results show the significant in saving development costs of the new model and the improved algorithm is more feasible than similar algorithms with lower software costs and better robustness.Finally,a software reliability prediction reamework based on component analysis is proposed according to the above fundamental researches.Moreover,a reliability prediction system of component-based software is further designed and implemented,including requirements analysis,overall design,functional design and system implementation.The reliability of a component software can be obtained according to this system during the software development process.