【作者】 穆莉花；

【导师】 谢冬青； 刘金朝；

【作者基本信息】 湖南大学 ， 软件工程， 2010， 硕士

【摘要】 随着网络信息技术持续快速发展,市场需求驱动的“网络化经济”得以快速的发展。在松耦合、开放异构环境下,Web服务技术使得企业间的异构应用系统得以互操作和互通信,实现了企业间B2B电子商务的协同工作,实现了产品设计与制造的协同以及供应链的协同管理,从而使得整个网络服务供应链的竞争力得以提高,缩短产品研发周期,实现服务的自动发现、自动组合与自动装配。然而市场需求驱动下,服务质量属性偏好和服务选择优化模型都存在多元化趋势,现有的工业标准和研究都未能实现多元化建模。且现有的研究很少针对服务合作网络性质进行定性与定量的研究,无法预测和偏好服务注册网络和服务协作网络的本质性发展趋势。本文针对以上问题,主要的研究工作有：(1)通过归纳、对比分析,研究了常见Web服务选择方法和现有市场需求驱动的服务选择方法及它们存在的问题,并且归纳基于Web服务的推荐系统研究现状、无标度网络研究现状和无标度Web服务研究现状,对它们各自的优缺点和发展趋势进行分析。这些研究为本文的研究目标提供了确定性方向。(2)在松耦合、开放异构环境下,提出的支持推荐系统的Web服务资源管理架构-RCWSRF。RCWSRF在远程客户端实现了实现异地在线动态快速计算客户的QoS属性偏好；在本地服务器端实现了业务流程构造与自动装配,并从大规模的服务候选集合中选择符合服务请求者需求的最优服务组合；RCWSRF通过服务发现代理产生服务爬虫实现了跨分散式UDDI注册中心的服务搜索,在本地实现集中式服务管理,这样实现了混合式服务管理。针对服务请求者QoS属性偏好的不确定性,采用改进的层次分析法,在异地客户信息反馈平台根据服务请求者的属性偏好的相对重要性建立判断矩阵,实时的计算属性偏好,并自动修正不一致的判断矩阵,实现客户属性偏好权重的自动计算。通过实验表明,在线QoS属性偏好计算算法是有效且计算结果是符合用户偏好需求的。(3)本文服务选择推荐系统模块定义了服务匹配方法,考虑虚拟企业之间的合作与竞争关系使得服务组合元模型趋于多元化,针对不同的应用场景,建立两类服务选择元模型-通用全局服务选择模型和基于虚拟企业动态联盟的全局服务选择模型。这是问题复杂的线性或非线性0-1多目标规划问题,本文采用鲁棒性好的遗传算法实现全局服务选择。通过大量的实验与数据分析,两类服务选择模型都能快速的计算出满足用户要求的服务序列。通过RCWSRF的服务选择推荐系统的快速服务匹配与服务选择,使得的基于SOA架构的RCWSRF能真正实现自动装配。(4)本文揭示了基于UDDI注册中心的增长性服务社区网络社团结构性质,并从服务注册拓扑结构和业务服务关联网络两方面来研究与总结服务合作网络的无标度性质。因此本文关于无标度服务合作网络的研究为面向服务的体系结构提供有力的理论与实践依据,也对复杂网络的理论与实证研究起到了一定的推动作用。同时服务合作网无标度特性的揭示,为本文的服务发现与服务选择方法研究提供了更充足的理论依据。现在Web服务组合与服务发现研究领域,往往将各个相互竞争的服务抽象成相关性很弱的独立广告服务,在服务组合时仅从任务流程的微观结构进行定性的功能匹配和QOS分析,而未能更进一步的从宏观结构方面定量的考虑各个服务提供者间的企业组织关系,也未能在服务组合领域考虑UDDI与注册服务的宏观发展趋势。本文结合当前复杂网络研究的方法,从服务注册拓扑结构和业务服务关联网络两方面来研究与总结服务合作网络的无标度性质,研究了无标度服务注册网和无标度业务服务网,给出了基于BA模型和局域择优连接演化模型的演化算法。研究表明基于UDDI注册中心的增长性服务社区网络具有社团结构性质,同时仿真试验揭示了服务合作网无标度特性,在服务网络中亦存在“强者愈强,胜者通吃”的现象。更多还原

【Abstract】 With the development of the network information technology, the market demand-driven "network economy" can get fast development. Under the condition of loose coupling and open heterogeneous environment, Web Service technology can makes the heterogeneous application system among the enterprises co-operate and intercommunicate, which realize the assistant function of the B2B among the enterprises, the assistant function of product design and manufacture and the assistant management of the supply chains. From the Web Service technology, the competition of whole network services supply chains can be increased, the span of the product producing can be decreased, and the services can be discovered, composited and assembled automatically. However, under the market demand-driven, the pluralism trend exists in QoS attribute preference and service optimization models, also the current industry standard and study can’t realize pluralism modeling. Moreover, the current study, pointing at the network attributes of the service cooperation to study qualitatively and quantitatively, is very less. In addition, it can’t predict the development trend of the preference services register network and service assistant network. Therefore, some studies aimed at the above problems have been presented in this thesis:(1) Through the conclusion and contrast, the methods of the common Web Service choosing and the market demand-driven service choosing have been studied, as well as their existing problems. The study current states of the Web Service-based recommending system, non-degree network and the non-degree Web Service have been concluded too. In addition, the development trend, the advantages and disadvantages of the above three have been analyzed. All of the studies have provided the exact direction for the study objective in this paper.(2) Under the condition of loose coupling and open heterogeneous, the Web Service resource management architecture-RCWSRF, providing the recommending system, has been proposed. RCWSRF has realized online dynamic fast computation for the users’QoS preference in the direction of the remote users. In the direction of the local servers, RCWSRF has realized the job flow structure and automatic assembling, and RCWSRF can choose the optimized service composition suiting for the service demanders from the large-scale candidate services. RCWSRF have also realized UDDI register center service retrieval across decentralization through service discovering proxy to produce service crawl. In the local, the centralized service management has been realized, so the composite service management has been realized. Aimed at the uncertainty of the demanders’QoS attribute preference, the method of the modified layer analysis has been adopted to build check matrix according to the demanders’QoS attribute preference in the different clients’ information feedback platform, computed QoS attribute preference real-time, corrected inconsistence of the check matrix automatically and realized the clients’ QoS attribute preference automatic computation. The experiment results show that the computation methods of the online QoS attribute preference are efficient and the computation results are fit for the users’preference.(3) The service matching algorithms have been defined by the service choosing recommending system. The relationships of cooperation and competition between the virtual enterprises make the service composition models pluralism. Pointed at different applications, two kinds of service choosing models (the common global service choosing models and virtual enterprises-based automatic ally global service choosing models) have been built, which is the problems of the complicated liner or nonlinear 0-1 multi-objective planning. In this thesis, the gene algorithms with good robustness have been adopted to realize the global service choosing. A lager number of experiments and data present that the above two kinds of models can compute the clients’demanding service series fast. Through the fast service matching and choosing of the service choosing recommending system of RCWSRF, SOA-based architecture-RCWSRF can realize automatic assembling.(4) In this thesis, the structure property of the UDDI-based increasing services community network has been proposed, and the non-degree property of the service cooperation network has been summed up and studied from the service register topology structure and job service related network. Therefore, the study of the non-degree service cooperation network has supplied strong theory and practice proof for the service-oriented architecture, as the same time the theory and practice proof of the complicated network has pushed the development. Moreover, the non-degree property has provided more adequate theory proof for the study of the service discovering and choosing.Currently, in the field of the Web Service composition and discovering, the services combating with one another are always abstracted as dependent advertisement services with less relation. In the process of the service composition, the function matching and QoS analyses can only be conducted from the microstructure of the task flow qualitatively, can’t allow for the enterprises relationships of the service providers from the perspective of macrostructure quantitively and can’t think over the macroscopical development trend of the UDDI and the register services. In this thesis, combined with the complicated network, the non-degree property of the service cooperation network, the non-degree service register network and the non-degree job service network have been studied and summed up from the service register topology structure and the job service related network, while the evolvement algorithms, based on the BA model and the local optimized connection and, have been proposed. The studies indicate that the increasing services community network of the UDDI register center owes the property of the community structure. Also the simulation experiments display that the non-degree property of the service cooperation network has the phenomenon of the "the strong being stronger, the winning sweeping the deck" in the service network.更多还原