【作者】 边爽；

【导师】 李冠宇；

【作者基本信息】 大连海事大学 ， 计算机应用技术， 2005， 硕士

【摘要】 随着计算机网络技术的飞速发展,从网络所获得的各种形式的资源呈爆炸性增长。电子商务的不断发展使得用户对汇总信息和决策信息的需求更加迫切。面对如此境况,人们的注意力逐渐从多数据库集成转向了多数据源集成,异构数据集成研究成为一种必然。但是异构数据集成系统与多数据库集成系统具有不同特性,这使得查询性能成为其实用化的瓶颈问题。 本文以国家自然科学基金项目(批准号:60172043)所研发的智能异构数据集成原型系统SORE系统作为基础,分析影响系统性能的因素,根据我们提出的性能目标逐层次的分析和优化。 论文首先优化原系统的功能结构,实现结构的优化设计。其后对客户层、服务器层、数据层分别进行优化方法的研究。在客户层,采用缓存用户请求和分页显示查询结果的方式来提高用户性能。在服务器层设计一个查询优化器,并研究了查询重写和代价估计优化方法以及缓存机制。在数据层,主要对半结构化数据进行研究,采用多种索引提高其查询性能。而且为了缓解用户跟踪数据更新和多用户同时访问带来的系统压力,提出应用主动服务模式。最终用一些模拟数据验证了我们提出的各种方法的有效性和可行性。更多还原

【Abstract】 With the quick development of computer network technology, the data on network grow in exploding way. And with the development of electronic commerce, the users’ requests for summary information and decision information are urgent. In the circumstances, people deflect attention from multiple database integration to heterogeneous data integration. Because of the different characters of heterogeneous data integration, the query performance becomes the bottleneck problem of its application.Based on the intelligent heterogeneous data integration prototype system SORE that is supported by National Natural Science Fund of China (No.60172043), the performance factors are analyzed. With the performance objective preferred, the paper analyzes all kinds of performance problems and makes them optimization in each architecture hierarchy.Firstly, the functional architecture of the system is optimized. Secondly the paper prefers different optimization approaches according to every architecture hierarchy. In user layer, cache users’ requests and show the results with multiple pages. In the integration server layer, design a query optimizer and research query rewrite, cost estimation model and caches mechanism. In data layer, the semi-structured data optimization is a main task. The different functional indexes are designed and implemented. Finally, in order to relieve the pressure from the users’ requesting the new information and a great deal of requests in the same breath, the active service is adopted. End of the paper using some analog data, these approaches are proved effective and feasible.更多还原