Research On QoS-aware Service Composition Methods In Various Complicated Environments

Servitization has become the mainstream of building complex software,as synergistic needs of globalization have gained prominence and service computing technology has undergone in-depth practices.Service composition,the core technology of building serviceoriented systems,is subject to satisfy the functional requirements from users and optimize the quality of service(QoS),that is,to achieve QoS-aware service composition.However,faced with development of service technologies as well as complexity and diversity of user needs,the building of service-oriented systems is increasingly featured by dynamic trend,large scale and diversified preferences,hereby posing a series of new challenges to QoSaware service composition with respect to optimized configuration,scale,and objective.Thus,it is shown that the design of QoS-aware service composition method for complex and diverse environments has become a new challenge in the field of building service-oriented systems currently.Certain advances have been made in the extant researches around QoS-aware service composition in the dynamic environment,QoS-aware service composition in the large-scale environment and QoS-aware service composition in the multi-preference environment,but disadvantages also remain.(1)In the dynamic environment,it is hard for existing methods to adapt to the application requirements for optimized configuration of service composition,due to insufficient consideration on domain feature information of service composition and absent consideration on holy optimization.(2)In the large-scale environment,the optimization mode and means of the existing methods are armed with limitations,which make it difficult to search service composition effectively.(3)In the multi-preference environment,the convergence and diversity of the Pareto-optimal solutions have not been well-balanced,as well as the rapid growth of the Pareto-optimal solutions has been ignored,making it difficult to quickly obtain a set of Pareto-optimal solutions that simultaneously approximate the Pareto-optimal frontier from convergence and diversity.To address the above-mentioned questions,research on QoS-aware service composition methods in various complicated environments has been conducted in this thesis.Firstly,in the dynamic environment,the restricted Boltzmann machine-based estimation of distribution algorithm has been proposed to support the adaptive service composition.The performance of service-oriented systems is guaranteed by means of mining the domain feature information of service composition and configuring globally optimized component services to respond to dynamic changes.Secondly,in the large-scale environment,the multi-clusters brain storm optimization algorithm has been put forward to support an expandable service composition.The algorithm speeds up the optimization efficiency by taking advantage of the core strengths of dividing solution space and curtailing solution space scope,as well as improves the optimization through the discussion mechanism and competitive strategy.Thirdly,in the multi-preference environment,the adaptive reference vector-based multiobjective differential evolution algorithm has been presented to support multi-objective service composition.The algorithm effectively offers a set of objective-balanced Paretooptimal solutions and meets user needs from multiple dimensions,by utilizing population differential evolution and reference vectors.To sum up,QoS-aware service composition in various complicated environments has been studied by this thesis.And a series of QoS-aware service composition methods have been proposed around the optimized configuration,scale and objectives of this question to provide technical support for better implementation of service-oriented systems,hereby effectively promoting the rapid development of Chinese software industry.These are valuable in theory and practice.