Web服务描述模型及其信息压缩机制研究

发布时间：2018-06-27 10:03

本文选题：普适计算 + 服务发现　；参考：《安徽工业大学》2012年硕士论文

【摘要】：有线网络环境下的Web服务发现技术研究已经比较成熟，如基于语义的服务匹配和分布式的服务发现架构，但面临人们是否可以快速获取需要的信息和服务，这就需要进一步的研究。在普适计算环境中，服务的使用者和提供者表现出移动性、间歇连接性，其拓扑结构随时间变化而改变，呈现出极强的动态性。而且，移动设备的存储容量和通信带宽都受限，这就要求服务发现所产生的消息负载不能过高。在动态的网络环境下，发现合适的服务是实现服务共享、复用的重要前提，服务发现的效果直接关系服务复用的质量，影响到服务组合的相容性和替换性，关系到能否实现快速的使用个性化服务。在服务发现过程中，有效地服务发现依赖于服务的描述方法；而普适环境又涉及到服务社会关系，那么本文需要建立社会关系模型，在服务描述中增加服务社会属性，并且要对服务的描述信息进行压缩优化。首先，本文在OWL-S本体描述语言基础上，扩展为一种轻量级的Web服务描述语言S-OWL-S。先分析了普适环境下存在的社会关系，建立支持社会上下文表达语义Web服务描述模型。考虑OWL-S没有服务社会上下文描述，但它具有可扩展性，所以设计了S-OWL-S本体服务描述语言，，构建的SCProfile本体囊括了社会上下文属性以及属性参数。其次，本文提出了一种改进的Counting Bloom Filter算法，即分域Counting BloomFilter算法，该算法按照服务的领域划分，并用于服务信息的压缩。压缩的目的是简洁地表示出服务集合信息，在向用户传播服务广告时减少带宽消耗和缓存的占用，此外可以减小误判率。文中分析了标准的Bloom Filter的基本原理、算法以及误判率，同时分析计数式Bloom Filter算法，在此基础上提出了分域计数式Bloom Filter算法，并比较了分域Counting Bloom Filter算法和Bloom Filter算法的误判率以及设计相应的哈希函数。最后，本文对服务属性进行量化，并把发布的服务信息存储到位串向量组以及进行服务的查找。由于Bloom Filter算法只能用于表示数据集合，本文把服务信息分为领域和其他服务属性两部分，并按照量化的领域不同，将量化的服务信息存储到不同位串向量。通过常用的查准率和查全率性能指标以及平均查找时间来检测分域Counting BloomFilter算法的有效性。
[Abstract]:Web service discovery technology in wired network environment has been more mature, such as semantic service matching and distributed service discovery architecture, but whether people can quickly obtain the required information and services. This requires further study. In the pervasive computing environment, the service providers and consumers exhibit mobility and intermittent connectivity, and their topology changes with time, showing a strong dynamic nature. Moreover, the storage capacity and communication bandwidth of mobile devices are limited, which requires that the message load generated by service discovery should not be too high. In dynamic network environment, finding suitable services is an important prerequisite for service sharing and reuse. The effect of service discovery is directly related to the quality of service reuse, and affects the compatibility and substitution of service composition. It is related to whether to achieve the rapid use of personalized services. In the process of service discovery, the effective service discovery depends on the service description method, and the universal environment involves the service social relationship, so this paper needs to establish the social relationship model, and add the service social attribute to the service description. And the service description information should be compressed and optimized. Firstly, based on OWL-S ontology description language, this paper extends to a lightweight Web services description language S-OWL-S. In this paper, we first analyze the existing social relations in the universal environment, and establish a Web service description model that supports the semantic representation of social context. Considering that OWL-S has no service social context description, but it is extensible, S-OWL-S ontology service description language is designed, and the constructed SCProfile ontology includes social context attributes and attribute parameters. Secondly, this paper proposes an improved Counting Bloom filter algorithm, which is called Domain Counting Bloom filter algorithm, which is divided according to the domain of services and used to compress service information. The purpose of compression is to concisely represent the service set information, reduce the bandwidth consumption and cache usage when propagating the service advertisement to the user, and reduce the misjudgment rate. In this paper, the basic principle, algorithm and error rate of standard Bloom filter are analyzed. At the same time, the counting Bloom filter algorithm is analyzed. The error rate of domain Counting Bloom filter algorithm and Bloom filter algorithm are compared and the corresponding hash functions are designed. Finally, the service attributes are quantified, and the published service information is stored in the string vector group and the service search is carried out. Because Bloom filter algorithm can only be used to represent data sets, this paper divides the service information into two parts: domain and other service attributes, and stores the quantized service information to different bit string vectors according to the different quantized domain. The validity of the domain Counting Bloom filter algorithm is detected by the commonly used recall ratio, recall performance index and average lookup time.
【学位授予单位】：安徽工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP393.09

【参考文献】