【作者】 黄宇；

【导师】 文俊浩；

【作者基本信息】 重庆大学 ， 计算机应用技术， 2012， 硕士

【摘要】 本文的研究题目来源于国家“十一五”科技支撑计划课题“勘察设计企业信息化关键技术研究与应用”的子课题“基于SOA技术的勘察设计企业应用软件架构系统”。随着全球企业化业务协作需求的增长，服务组合变得越来越重要。而BPEL由于其实现简单，功能强大，已经被越来越多的企业用户所接受。但由于Internet中功能相似的Web服务数量越来越多，服务选择变得越来越困难，且现有的Web服务业务流程复杂多变，有些业务流程的执行步骤难以在定义时就明确下来，因此，基于BPEL的服务组合应具备良好的动态服务选择能力和根据运行的实际环境来决定业务流程中服务执行顺序的能力。在基于BPEL服务组合的动态优化研究中，业务流程在进行服务选择时以服务的QoS属性为基础的；在进行流程中服务执行顺序优化时，考虑的是流程中服务之间的内在关系。因此，在进行服务选择时，本文以QoS全局最优为基础，改进传统分支限界法，设计了QoS全局最优的服务选择算法，使得流程构建的组合服务在满足约束条件下QoS最优；在进行流程中服务执行顺序优化时，以服务之间执行的交互代价为基础，改进已有的以单亲遗传算法为基础的流程执行顺序优化算法IPGA，使得流程的执行代价最小。在本文中主要完成如下工作：①分析了各种服务组合实现方法，重点研究了基于BPEL的服务组合方法，提出了本文的研究方向和意义。②结合本文研究背景，分析了现有的BPEL服务组合优化方法，总结了它们的不足，提出从流程的服务选择和流程执行顺序两个方面进行优化。③提出了基于改进分支限界法的QoS全局最优服务选择算法，利用矩阵的优点对传统分支限界法解空间树的建立复杂和效率低等不足进行了改进。④将服务执行顺序优化问题转化为TSP问题，提出了基于改进IPGA的流程执行顺序优化算法，对IPGA算法的全局收敛性差和执行效率低做出了改进。⑤通过优化算法，提出了基于BPEL的服务组合动态优化机制。在现有的基于代理和BPEL的动态服务组合模型基础上，将两个优化算法整合成优化模块加入进来，实现了该机制的设计。⑥通过模拟和实验分析，分别对机制的可行性和算法的性能进行了验证。更多还原

【Abstract】 The research subject of this thesis comes from the topic "The Architecture Systemof Application Software for the Survey&Design Enterprise Based on SOATechnology", which is one sub-topic of the topic "Research and Application for KeyTechnology of the Survey&Design Enterprise Infomationization", which is one of theKey Projects in the National Science&Technology Pillar Program during the EleventhFive-Year Plan Period.With the growth of the global enterprise collaboration, service composition isbecoming more and more important. For BPEL, because of its simple implementationand powerful function, has been accepted by more and more corporate users. However,with the increasing number of Web service with similar functions in the Internet, serviceselection has become increasingly difficult. Meanwhile, the existing Web serviceprocess is complicated, some business process is difficult to be finalized in the step ofthe definition. Therefore, BPEL-based service composition should have abilities of gooddynamic service selection and deciding the execution order of business processaccording to the actual environment.In the research on BPEL-based service composition optimization, the businessprocess was based on the service’s QoS during the service selection, moreover, thebusiness process considered the intrinsic relationship between the service during theservice execution order optimization. Therefore, during the service selection, the paperimproved the traditional branch and bound method, designed the algorithm of QoSglobal optimization, and realized the dynamic service selection based on QoS globaloptimization under meeting the constraints, meanwhile, in the process of the serviceexecution order optimization, the paper was based on the cost between Web service andimproved the existing IPGA algorithm based on the partheno-genetic algorithm, andminimised the execution cost.This paper mainly accomplished the following tasks:①Analyzed a variety of service composition, focusing on service compositionbased on BPEL, and proposed the paper’s research direction and meaning.②Combined with the research background, analyzed the existing optimizationmethods of BPEL-based service composition, and summarized their weak points, thenproposed two optimization aspects: service selection and service execution order of the process.③Proposed the service selection algorithm with QoS global optimization based onthe improved branch and bound method, reduced the complexity of building thesolution space tree and improved the efficiency of the traditional branch and boundmethod through using the advantages of the matrix.④Translated the service execution order optimization into the TSP problem,proposed the optimization algorithm of service execution order and overcame theIPGA’s weak points of the bad global convergence and low efficiency.⑤A dynamic optimization mechanism for service composition in BPEL based onoptimization algorithm was proposed. On the basis of the existing agent-based BPELservice composition model, two optimization modules based on the optimizationalgorithm were added to the service agent, and finally the mechanism was designed.⑥The simulation and experimental analysis showed the feasibility of themechanism and the performance of the algorithm.更多还原