【作者】 刘星宇；

【导师】 于双元；

【作者基本信息】 北京交通大学 ， 计算机应用技术， 2008， 硕士

【摘要】 信息集成系统的目的是通过集成各种可用资源建立一个复杂的信息系统,并最大限度的利用这些资源,包括计算资源。常用的做法是建立一个一致的查询界面(语言和模式表述),对多个数据源通过这个统一的界面建立查询。在当前已实现的查询集成系统中存在几方面的缺陷:数据源的组织不够灵活,不能满足一些特定的应用需求;没有一个能够体现混合多数据源查询定义的中间语言,使得优化推理不能体现多数据源混合的特点;在分布式因素分析及相关优化策略方面做的工作很少。本文在传统数据库架构的基础上,提出并实现了一个基于有向查询图的异构数据源查询集成框架,利用虚拟数据源组织方式实现了数据源的灵活组织,并通过SQL,XQuery为异构数据源建立了一致的访问接口,从而实现了对分布式异构数据的一致化查询。本文还针对分布式查询优化提出了增强浓密树的优化算法,此外在分布式因素分析和相应的优化策略方面作了初步的探索,并在最后给出了相关的分布式数据查询的实例。最后的实例表明,本文提出并实现的异构数据访问系统实现了对异构数据源的统一、透明的访问,并具有较好的可扩展性,最后可通过量化的实验方法还证明了该系统具有较短的访问时间。更多还原

【Abstract】 The aim of information integration is to build sophisticated systems by making use of available information resources, which including the computing resources, to fullest extent and by pushing costly operations to these sources as much as possible. What the queries integration systems do is to create a unified query interface, including query languages and schema that queries built on, and users can query multi-sources through the unified query interface.There are some shortages in current implements of query integration systems. Firstly, the organization of data sources can’t meet the requirement of some applications. Some implements need a middle language to express the definition with multi-sources queries characteristics, with which we can do some optimization reasoning expediently. Thirdly, current implements work little research in distributed factors and related optimization algorithm.In this thesis, base on the traditional database framework, we propose and implement a Query Graph-based Heterogeneous Sources Framework. Based on this architecture, we research in some aspect of query integration system to a quite deep extent and implement a prototype of query integration system. We organize data sources with the structure of virtual sources, which is flexible to use in some scenario that current implements can’t be used in. We use SQL and XQuery as the common query interface of distributed heterogeneous sources. Also, we define a middle algorithm, Enhanced Bushy Tree to limit traditional Bushy tree’s search space and enhance the performance of distributed query execution. One more research in the analysis and optimization strategy of distributed factor is also referred in this thesis. We also give some example based on this architecture at last of this thesis.The example given at last show the successful design and implementation of our Heterogeneous Sources Framework, and additional quantitative performance measure result is available at last.更多还原