联邦数据库事务研究-智能与分布计算实验室

联邦数据库事务研究

姓名	Terefe Muleta
论文答辩日期	2003.05.08
论文提交日期	2005.08.07
论文级别	硕士
中文题名	联邦数据库事务研究
英文题名	Research on Federated Database Transaction
导师1	卢正鼎
导师2	李瑞轩
中文关键词	事务管理;联邦数据库系统;并发控制;分布式对象核心;数据库安全;数据仓库
英文关键词	Transaction management;Federated Database Systems;Concurrency Control;Distributed Object kernel;Database Security;Data Warehousing
中文文摘	支持多种成员数据库和长事务处理的联邦数据库事务（FDBT）是一个非常棘手的问题，提供联邦数据库事务服务意味着需要从局部和全局视图两个方面来操作多种异构数据源，这使得传统数据库的事务模型和并发控制方法在联邦数据库系统中并不适用。联邦数据库提供了对多种数据源进行统一访问的界面和方法，针对传统事务模型在联邦数据库的缺陷，给出了联邦数据库系统的事务模型，并使用数学的方法对其适宜性进行了评价。这个模型为建立联邦数据库设计了灵活的分布式体系结构，可以不使用全局锁协议而显式地识别和处理事务的并行性。该模型适用于联邦数据库事务、非传统数据库事务模型和复杂的分布式信息系统。在有大量的成员数据库参与的情况下，比如有1000个参与的数据库，依赖于两阶段提交（2PC）和封锁协议的事务模型几乎只会使事务失败或死锁。使用分布式对象核心（DOK）技术来研究联邦数据库事务的访问控制方法是一个合适的选择。基于分布式对象核心技术，给出了一种适用于联邦数据库的安全体系结构，使得DOK系统在自治的、分布的、异构的数据库环境中能够加强其联邦安全策略。将数据仓库的方法引入联邦事务处理会有许多优点，数据仓库的联机事务处理（OLTP）和联机分析处理（OLAP）技术可以加强联邦数据库的事务处理能力。基于时间戳排序的事务管理算法和调度算法，可以实现简单的联邦数据库事务处理。最后，给出了这些算法的性能分析和评价。
英文文摘	This thesis addresses the problem of federated database transaction (FDBT) support for provisioning of large numbers of component databases and/or very long-duration transactions services. Provisioning of large scale federated database services means dealing with various heterogeneous data sources from local and global view. The research reports our approach towards the use of federated database technology to provide uniform access to the various data sources. We evaluated the suitability of a scalable model for large number of federated database systems, something which is not present in the current literature. As such, it makes significant contributions to the areas of Federated Database Transaction, unconventional database transaction models, and complex distributed information system. Such as, flexible and distributed architecture for building federated databases has been designed; development of transactions models that explicitly recognize parallelism and do not use any global locking. Transaction models that depend on two-phase commit with locking are almost guaranteed to fail or deadlock when applied to large number of database like 1,000 participating databases. This thesis attempts to explore the Distributed Object Kernel (DOK) which is a federated database system that study the development of a federated access control, as well as a secure logical architecture allowing the DOK system to enforce federated security policies in the context of autonomous, distributed and heterogeneous databases. This thesis also tries to explain the benefits of data warehousing for the federated database transaction and introduces the OLTP and OLAP of data warehousing to enhance the transaction capability of federated database systems. Finally, the implementation of timestamp ordering transaction manager algorithm and scheduler algorithm in the area of federated database transaction are given, and the performance evaluation and analysis for federated transaction management is discussed.