【作者】 周明强；

【导师】 朱庆生；

【作者基本信息】 重庆大学 ， 计算机应用技术， 2010， 博士

【摘要】 服务注册中心是面向服务架构的重要组成部分之一。随着互联网范围内面向服务计算模式的普及和网络规模日趋复杂,海量服务注册信息的合理组织和管理对面向服务计算的性能显得至关重要,因此,服务注册中心的合理部署是实现上述目标的基础。目前,复杂网络环境下服务注册中心部署存在的主要问题包括缺乏理论支撑,不能适应网络动态演变,缺乏语义信息处理能力,不能理解服务注册信息和自动演算推理。与此同时,在复杂网络的研究中发现万维网等都具有非常有价值的统计特性,诸如网络呈现的社区结构特性,它为服务注册中心的研究提供了有力理论依据。《中国至2050年信息科技发展线路图》已将网络科学列为发展新信息科学和前沿交叉科学的重要分支,因此,在对等网络中利用语义和社区结构特性来搜寻和发现语义社区(根据节点兴趣偏好的语义信息划分的内部联系紧密而外部联系稀疏的逻辑结构),然后在其中设置服务注册中心的研究具有重要的实际应用价值,也对复杂网络的理论研究起到了积极的作用,在语义社区中进行服务注册中心部署策略研究具有理论价值。基于上述背景,论文通过结合对等网的可扩展性、语义网络的智能性和复杂网络的社区结构特性,开展了语义社区中服务注册中心部署策略研究,提出了一个解决服务注册中心部署问题的模型,并针对模型中所涉及的节点选择、网络演化和语义社区划分等具体问题展开了研究。因此,论文的主要内容包括系统模型设计、超级节点选择、网络拓扑演化、语义信息传播和语义社区划分五个方面。论文研究取得的主要成果体现在以下几个方面:(1)提出了以语义特征为基础的具有超级节点的语义路由服务注册中心(Super Peer and Semantic Routing UDDI, SPSR-UDDI)模型。模型以语义信息为基础,将节点兴趣偏好语义信息以“谣言”形式传播,节点通过语义信息获取到与其具有相同兴趣偏好的其它节点信息,把网络中的节点按照兴趣偏好划分语义社区,在社区中选择综合性能相对较好的节点作为超级节点承担服务注册中心的任务,社区间通过语义信息互联,对于给定的查询主要集中在对应的主题域,查询处理范围小,查询处理效率高。(2)提出了基于集对的超级节点选择算法。算法在超级节点选择时引入资源消耗因子和负载能力,量化了负载能力性能指标,增加了算法的普适性和灵活性;采用集对方法选择超级节点,较好地量化节点各种参数,实现了模糊综合评价,该方法还可以通过权重的方式体现用户的对特定参数的重视程度,是一种有效的评估方法;根据节点排队长度涨落的均方根函数,在节点即将出现拥塞之前启动超级节点选举算法,提高了系统整体服务的响应时间。(3)提出了基于节点点强度的网络演化方法。方法采用在密集区投放服务点的思想,通过节点边权优先连接原则让部分节点动态调整连接策略,实现了网络自组织演化过程,并使最终的网络具备小世界特性,获得较好的服务响应时间。(4)提出了采用谣言兴趣衰减机制的语义信息传播模型。模型采用谣言兴趣衰减机制,以较小的消息量在网络中实现节点兴趣偏好信息的有效传播;每个节点仅依据网络中的兴趣偏好信息进行决策,并根据兴趣偏好的语义匹配相似程度改变节点之间的连接度,减少了对全局性知识的依赖。(5)提出了基于社区语义熵的语义社区划分算法。算法通过节点间的语义信息,利用社区语义熵实现了将兴趣偏好相似的节点聚集在同一主题领域的语义社区划分;社区划分后,由超级节点管理每个语义社区,承担服务注册中心任务,节点根据语义信息自主地建立并维护与其它节点的拓扑连接,完成网络自组织过程,达到节点资源的有序组织,为语义社区结构中的服务注册中心部署提供的新途径。论文的研究工作为挖掘网络语义社区结构和部署服务注册中心提供了理论和方法,在复杂网络特性应用研究方面进行了有意义的探索,对网络科学的发展起到了积极的作用。更多还原

【Abstract】 Service registry is an important component in service-oriented architecture, called SOA. With the arising of service-oriented computing mode in Internet and the growing complexity of network, it becomes very important to the efficiency and performance of service-oriented computing that huge volume of service registering information are reasonably organized, managed and fast indexed in complex network, so the appropriated deployment of service registries is the essential to gain this goal. The current service registries in complex network can not adapt the dynamic evolution of network because of the lack of appropriate theory support in organizing, managing and deploying. Due to the lack of ability of semantic processing, service registration information can not be efficiently understood, automatically calculus reasoned and appropriately organized. Meanwhile, in the study of complex network, many valuable statistical properties results are found in the World Wide Web etc., such as the network that presents a community structural characteristics which can provide a strong theoretical basis to study on service registry. The report "The Road Map of China Information Technology Development Until 2050 " has listed the development of new information science and cutting-edge interdisciplinary. Therefore, the study of searching and discovering semantic communities (A logic structure which is constructed through semantic information according to the preference of each node) through the semantic and structural characteristics in peer-to-peer network can be used to direct the setting of service registries. These studies do not only have important application value in organizing service-oriented network, but also theoretical value which promotes the application of complex network theory.The strategy of service registry deployment in semantic community is studied in this dissertation combining the extensibility of P2P network, intelligence of semantic network and characteristic of community structure in complex network. A model is proposed to help the deployment of service registry. A few key issues involved in this model including node selection, network evolution and semantic community construction are studied. So five main topics are discussed in this thesis, including the design of system model, selection of super node, evolution of network topology, propagation of semantic information and the division of semantic community. Some innovative research results are gained as the following:(1) A Super Peer and Semantic Routing Model is proposed, which is based on the semantic characteristics of information, the preferences of each node in the semantic information is propagated in gossiping mode. Each node can get the information of other nodes which have the same preference through the semantic gossip, so the nodes in the network can be divided into different semantic community based on their preferences. The node with the best overall performance is selected as the super node, which is in charge of registry. Different communities can be interconnected through semantic information. Given request of querying, the processing is limited in appropriate domain, so the processing range is decreased, and more efficiency of processing gained.(2) The set-pair based super node selecting algorithm is proposed. The factor of resource consumption and load capability are introduced in super node selecting. The load capacity is qualified, the universal and flexibility of super node selection can be increased. Selecting super node using the method of set pair can qualify different parameters on each node, so the fuzzy evaluation of each node can be achieved. This method can also reflect the user customized emphasis on specific parameter through weight, which is an efficient evaluation method. The root mean square function is used to measure the dynamic changing of node queuing, the election algorithm could be trigger before the congestion of processing, so the whole service response time could be improved.(3) A node intensity based network evolving method is proposed, which uses the idea deploying more service points in dense zone. The connection strategy is adjusted dynamically through the edge-weight priority rule, so the process of network self evolution could be achieved. The final network has the characteristics of small world and a better global response time is gained.(4) A model of semantic information propagation using gossip mechanism is proposed. Gossip decay mechanism is used to achieve the efficient propagating of node preference with lower cost. Each node makes decision based on preferred information. The connection degree between nodes could be changed according to the semantic similarity of preference between nodes through semantic matching. The dependence on global knowledge is magnificently decreased.(5) A semantic community division algorithm based on the entropy of community semantic is proposed. In this algorithm, using the concept of semantic entropy, nodes with similar preference are gathered into a domain of semantic community division based on the semantic information between nodes. After the division of community, each semantic community is managed by a super node, which is responsible for registry. Each node sets up and maintains the topology of connection with other nodes independently according to the semantic information, process the self organization of network to achieve the orderly organizing of resource on each node. This provides a new way to the deployment of service registries in semantic community.This research provides the theory and method for mining semantic community structure in the complex network and deployment of service registry, meaningfully explores in applied research of characteristics in complex network, and plays a positive role for the development of network science.更多还原