【作者】 杨冬；

【导师】 刘磊；

【作者基本信息】 吉林大学 ， 计算机软件与理论， 2014， 博士

【摘要】 本文主要研究内容为当前Web服务领域中的几个热点问题，包括Web服务选取方法、多等级Web服务部署问题以及Web服务组合运行时容错处理和异常恢复方法及策略的研究。本文中首先提出了一种改进的人工蜂群算法用来求解服务选取问题。结合遗传操作设计雇用蜂觅食及侦查蜂策略，并针对服务选取问题的特点提出了基于效用值及违反约束度的食物源评价方法及基于邻域的随机贪心观察蜂觅食策略，进而建立求解该问题的人工蜂群算法优化模型，在此基础上实现了求解服务选取问题的混合人工蜂群算法-GABC。其次提出了一种基于协同过滤推荐的混合式Web服务选取方法，将Web服务请求者的个人兴趣偏好和相似用户群体的经验相结合，利用推荐技术对具有相同功能和不同Qos属性值的Web服务进行选取。再次提出了一种改进的遗传算法用来求解多等级服务部署问题。在选择过程中引入了个体被支配强度，通过将个体的支配强度和被支配强度结合到一起建立对个体的评价策略，进而对效用函数进行重新定义，并设计了基于概率的交叉策略及结合局部搜索的个体变异策略和根据评价结果进行环境选择及生成个体的交叉概率方法。此外还设计新的局部搜索策略并将其融入到变异策略中，提高了变异操作的有效性。最后给出了一种基于服务冗余和约束冗余的Web服务组合运行时容错处理和异常恢复策略。通过在组合规划阶段为服务流程中每个抽象成员服务建立后备服务组和将约束分解到每个抽象成员服务上的方法，为运行时的容错处理和异常恢复提供了一种以尽可能小的代价恢复服务组合运行并尽量保留已执行部分结果的处理策略。在此基础上，进一步给出了服务组件库的建立和维护策略及其算法。本文中针对Web服务领域中几个关键问题进行了研究和探讨，所做的主要研究工作和研究成果在提高web服务和服务组合的可靠性和效率等方面具有一定的意义。更多还原

【Abstract】 Due to the rapid development of Internet technology and the rise of e-commerce, greatdemands for large numbers of new resources and applications emerge. Traditional softwaredevelopment architectures and methods could not satisfy the needs in complex enterpriseapplications. In order to handle the differences between platforms and protocols in the Webenvironment encountered in enterprise application integration and the collaboration in looselycoupled approaches, and to improve the efficiency of software development and integration,in addition to the sharing of information resources, a new software development architectureis proposed as a service-oriented architecture, namely SOA.SOA is an open, standard protocol that serves as a basic functional unit, a looselycoupled model framework, and also a design method. It can be applied to the existed isolatedsoftware applications into self-contained services. Applications may access these servicesthrough simple interfaces and communication protocols, which reduces the costs ofinteroperability, and supports the rapid development of enterprise applications. It can be usedswift development of enterprise applications, and may respond in time to the actual needs ofthe rapidly changing Internet environment. The research of SOA has received widespreadattention in recent years, from industry and academia.With the gradual popularization and application of SOA, Web service has been playingan increasingly significant role as a way to achieve SOA, also as a new type of distributedcomputing resources.Web services collect application logic, network technology and a variety of servicefunctional modules so that any company or individual could access them in any place andtime. With the extensive application of Web services, a series of major issues emerge on thelist. For example, in the instantiation of an abstract service process, how to select from a largenumber of candidate emerged Web services with the same quality of service for differentfunctions; in the dynamic Internet environment, how to ensure the reliability of the serviceportfolio under the possibility of a service failure or new services available, and so on.On the basis of Systematic analysis of related methods and the current research status athome and abroad, this dissertation is focused mainly on web services selection, servicesdeployment and fault-tolerance and recovery in runtime. The main achievements and content of this dissertation are summarized as following:(1) This thesis reviews the process of software development architectures and methods inthe past years, especially introduces the emerging SOA architecture. The analysis andsummary of research status of hot issues in the field of Web services is also given. All theseintroduce and summaries provide a solid foundation for the research of this thesis.(2)In order to solve the service selection problem, a discrete artificial bee colonyalgorithm called GABC was proposed and the corresponding optimization model wasestablished. According the character of the service selection problem, the genetic operator iscombined to form the foraging strategy of employed bees, and a greedy exploited strategy isdesigned for the onlookers. The attractive probability of a food source is redefined combiningits constraint violations. Through experiments, the algorithm was tested and compared withrelated algorithms. The results showed that GABC algorithm can effectively solve thisproblem.(3)A web services selection method based on collaborative filtering recommendation isproposed. In this dissertation, collaborative filtering recommendation method is indroduced toweb services selection field.According to the selection of similar web service users andcurrent user’s individual preferences, the unknown Qos values can be predicted, and then theweb services with top Qos values are Personalized recommendated to the current user.(4)For SLA-aware service composition problem, a multi-objective model for thisproblem is built and a multi-objective genetic optimization algorithm is also proposed in thispaper. According to the characteristics of this problem, genetic operations are redefined,including the introduction of individual domination strength into environmental selection, andthe redefinition of crossover strategy and individual mutation strategy incorporated with localsearch. At last, we analyzed our algorithm SMOGA, and compared it with recently proposedalgorithm E3-MOGA and NSGA-II for this problem on different scale test cases; theexperiments results show that algorithm SMOGA can solve this problem more effectively.(5)On the basis of already existing runtime fault-tolerance and recovery methods, a QoSconstrained based on similar web service substitue and Dynamic Partial Reconfigurationstrategy is proposed, and then the corresponding algorithm is given. The core idea of thisstragety is taking the redundant constraints into account and decomposing the globalconstraints into every member of service composition, and then the result of decomposing canbe used as constraint conditions of similar web service substitue and Dynamic PartialReconfiguration. Further, a web service components library and corresponding maintenancealgorithm are proposed. The research work and achievement in this dissertation has some theoretic andapplication value to improve the reliability and efficiency of Web services and servicecomposition.更多还原