【作者】 陈志勇；

【导师】 李庆忠；

【作者基本信息】 山东大学 ， 计算机软件与理论， 2011， 博士

【摘要】 随着技术的进步和网络应用的普及,一方面,有越来越多的企业或组织将自己的业务功能单元以Web服务的形式发布到网络上。而另一方面,普通消费者对网络服务提出了更高的要求。他们已不再满足于互联网上的简单的服务提供方式,如获取天气资讯、查询航班信息、宾馆预订等单一功能的服务。用户期望通过网络服务平台以生活中惯常的表达方式提出自己的模糊请求,而不管身在何处,都能享受到专门为自己量身定制的、能提供复合功能的高品质服务。这就对当前主流的Web服务的提供方式提出了挑战。显然,任何商家无法独立提供用户需要的完善的服务。在市场竞争日趋激烈的商业背景下,要满足用户的这种服务需求,就需要同一商业价值链上的各企业实体组建动态的虚拟企业联盟,整合各自的优势资源,开展协同商务,将联盟各成员发布的Web服务有机地集成,形成大粒度的具有内部流程逻辑的组合服务,才能动态、灵活、快速地响应持续变化的用户需求。这也是本文开展面向模糊请求的服务组合关键问题研究的意义所在。针对用户服务请求所表现出的即时性、定制性和模糊性的特点,本文对涉及服务组合整个生命周期的Web服务发现和匹配、服务的自动组合、动态服务选择和对组合服务流程执行过程的监控及异常处理等问题开展了深入的研究,取得了以下成果：1、提出了一种面向模糊请求的服务发现和匹配方法,与基于传统本体模型的服务匹配方法相比,不但解决了其无法处理模糊请求的不足,也有效地提高了服务发现的查准率和查全率。通过对传统本体模型的模糊扩展,构建了包括模糊领域本体和模糊QoS本体在内的模糊本体模型。在模糊本体的支持下,将服务分两层(功能语义匹配层和非功能属性的语义匹配层)进行匹配,并给出了服务匹配的算法。2、提出了一种从OWL-S过程模型到流演算形式的映射算法,在此基础上,给出了基于流演算的Web服务自动组合算法,有效地解决了在组合服务流程的规划空间较大时算法复杂度高的问题,提高了服务组合的效率。在领域本体的支持下,利用流演算的规划和推理能力,自动地生成抽象的组合服务流程。提出的BCA-FC算法采用了前推推理机制对状态和动作进行推理,与采用回归机制的其它AI规划方法相比,有效降低了估价属性有效性的计算复杂度。实验结果表明该算法在组合服务流程中任务节点数目较多时,仍有较高的执行效率。3、提出了一种基于多维QoS约束的服务选择模型,为服务选择以及流程执行中的动态服务替换提供多维QoS支持；提出了基于MMKP背包问题的启发式服务选择算法,有效降低了算法的搜索空间,能够快速有效地完成服务选择。针对目前研究中存在的复合服务QoS全面评价标准不明确,未充分考虑用户需求尤其是模糊请求的问题,提出一种服务的QoS度量模型,定义了的服务质量效用评价函数,并给出了不同的全局QoS约束下动态服务选择模型,为服务选择以及流程执行中的动态服务替换提供多维QoS支持。提出了一种基于MMKP背包问题的使用规则库的启发式选择DSSR_KP算法,实验表明,启发式选择算法通过使用规则库降低了搜索空间大小,能够快速有效的完成服务选择。从而使得在满足用户的QoS模糊约束的同时,获得综合性能评价最优的组合服务流程实例成为可能。4、提出了一种流程异常处理模型,有效地实现了流程业务层的快速异常诊断；提出了一种面向业务逻辑层的流程异常处理方法,提高了组合服务流程执行中异常处理的自动化水平和自适应能力。在异常处理模型的状态监控模块中,通过基于WS-N的消息发布和订阅机制及时获取与服务执行密切相关的信息资源,并据此做出诊断,为可能的流程异常处理奠定基础。针对业务层的服务异常,采用了基于撤回业务生成图的组合服务流程异常处理方法,在用户可接受的代价范围内实现服务的撤回,保证了组合服务流程可靠、一致的执行。更多还原

【Abstract】 With the improvement of technology and the popularity of network applications, on one hand, more and more companies issue their business function units to the network by the form of Web services. On the other hand, the general consumer presents higher demands to network services. They are no longer satisfied with the simple service providing approaches, such as getting weather information, inquiring the flight information and reserving a room in a hotel. Many users expect Internet service platform allow them to make fuzzy requests using natural language expression. Wherever they are, users can enjoy their specially tailored, high-quality service which can provide composite function. This poses a challenge to the current Web service product delivery mechanisms. Obviously, users can’t get faultless Service from any businesses. In an increasingly competitive commercial market context, to meet users’ service demand, we need every business entity in the same commercial value chain to set up dynamic virtual enterprise alliance, integrate their respective advantage resources, develop collaborative commerce, organically integrate web services published by union members, then form coarser-grained composite service with internal process logic. Just in this way the constantly changing user demand could be dynamically、flexibly and rapidly respond. So we research the key issues in service composition orienting vague requests.Considering the features of instantaneity, customization, fuzzification user service requests appears, this article delves into Web service discovery and matching related to the whole life cycle of service composition, automatic combination, dynamic service selection, and the monitoring and exception handling about process execution. We achieved the following results:1. We proposed service discovery and mapping methods based on fuzzy request. Compared with the traditional matching methods, the service discovery and matching scheme not only allows the user to make vague service requests, but also improve the recall and precision rate. By fuzzy expanding traditional ontology model, fuzzy ontology model was constructed including fuzzy domain ontology and fuzzy QoS ontology. With the support of fuzzy ontology, we divided the service into two layers (functional semantic matching layer and non-functional properties semantic matching layer) to match and gave the service matching algorithm.2. We proposed a mapping algorithm from OWL-S process model to flow calculus stream. On this basis, we gave flow-calculus-based automatic Web service composition methods. When the room planed room for composite service process is larger, the complexity of algorithms is higher. This method solved the problem abovementioned and improved the efficiency of service composition. With the support of domain ontology, we use the planning and reasoning ability of flow calculations to automatically generate abstract composite service process. The BCA-FC algorithms uses the progression reasoning mechanisms to reasoning the actions and states. Compared to the other AI planning methods using regression mechanism, BCA-FA algorithms reduces the computational complexity when valuating properties. Experimental results show that it still has high efficiency when there are a large number of task nodes in the combined service process.3. We proposed a dynamic service selection model based on multi-dimensional QoS constraints, providing multi-dimensional QoS support for service selection and dynamic service replacement in process execution. We proposed a heuristics services selection algorithms based on MMKP knapsack problem. This algorithm reduce the search space size, and complete service selection quickly and effectively. As current study hasn’t explicit a comprehensive evaluation standard about composite service QoS, and not fully considered user requests especially the fuzzy requests, we proposed a service QoS measurement model, defined service quality effectiveness evaluation function, and gave dynamic service selection model in different global QoS constraints, providing multi-dimensional QoS support for service selection and dynamic service replacement in process execution. We proposed a heuristic selection approach using rule bank based on MMKP knapsack problem. Experiments show that by using the heuristic rule base selection algorithm reduces the search space size, and can complete service selection quickly and effectively. This method can satisfy users’ QoS fuzzy constraints, in the main time, make it possible to obtain the composite service process instance with the best comprehensive performance evaluation.4. We proposed an exception handling model of process, effectively implementing the rapid diagnosis of anomalies in the business process layer. We proposed exceptional handling methods of process based on business logic layer, improving the automation level and adaptive ability of exception handling during the execution of composite service process. In the state monitoring module of exception handling model, we obtained timely the information resources closely related to service execution through message published and subscribed mechanism. Accordingly, we can make diagnosis and hence set basis for possible exception handling of process. Based on generation figure of revoked business, we used exception handling methods of composite service process. This method enables us to withdraw service in an acceptable cost range in the context of use, guaranteeing the reliable and consistent execution of composite service process.更多还原