Developement and Implementation of Access Protocols for Heterogeneous Distributed Database Systems

本研究的目的是探讨异质联邦分布式数据库的可行性，特别是其在高速广域网上的文档传递中的应用，如互联网上的数字图书馆。在模式相似的分布式数据库中，需要引入一种查询分解方法，将子查询发送到每个数据库服务器，这些服务器的数据库与发出的查询相关。研究了查询的分解方法，引入了虚拟数据库和组件数据库的概念。提出的模式集成方法可以在不考虑组件数据库逻辑分布的情况下，通过自动分解查询并将子查询传输到多个相关服务器的算法来发布数据库查询；在分解算法和查询处理方法的同时，还基于关系理论建立了方法的正确性和可靠性理论。与传统方法分别向每个服务器发布查询相比，该方法通常可以提供更多的查询结果集。研究人员还通过模拟显示了结果集的增加程度，并在保存样本书目信息的原型数据库系统上实现了该算法。该系统在局域网上运行，通过给出大量的查询等结果记录，验证了该方法的有效性，通过对异质分布式数据库环境下查询的理论分析，揭示了基本方法和实现策略，并将在Internet等广域网络上有效。

The aim of this research was to investigate the feasibility of heterogeneous federated distributed databases, especially focusing on its application for the document delivary on high-speed wide area networks, such as digital libraries on the Internet.The first step we have done is a theoretical analysis of the distributed database query processing. In a distributed database with similarities in their schemata, it is required to introduce a query decomposition method for sending sub-queries to each database servers whose database are relevant to the issued query. We studied the decomposition method of queries, introducing the concept of 'virtual database' and 'component database'. The presented method of schema integration allows us to issue a database query without considering logical distribution of component databases, using an algorithm for automatically decomposing the query and transmitting sub-queries to several relevant servers.Along with the decomposition algorithm and query processing method, a theory for the correctness and soundness of the proposed methods is also developed based on the relational theory.The presented method can generally provide us a large set of query results than the set given by the conventional methods which issue queries separately to each server. The investigators also showed that the degree of increase of the result set by simulation.The presented algorithm was implemented on a prototype database system which holds sample bibliographic information. The system, which is operational on a local area network, demonstrated the effectiveness of the presented method by giving larger number of resultant records for example queries.This investigation could reveal fundamental methods and implementation strategy by theoretical analysis of queries over heterogeneous distributed database environment, and it will be effective on wide-area networks such as the Internet.