RIA: Combining Queries and Navigation in Workstation/ServerDatabase Systems

当前的数据库系统技术不能满足许多应用领域的需求。 关系数据库系统（RDBMS）擅长提供高级的、关联的（即，基于查询的）对大的平面记录集的访问。 另一方面，面向对象数据库系统（OODBMS）提供了更强大的数据建模能力，但主要支持基于导航的数据访问。 本研究的目标是开发灵活的工作站/服务器数据库系统，它可以动态地适应各种应用程序的要求，使这些应用程序利用数据库的能力，并从工作站和局域网的价格/性能的持续快速增长中受益。 这样的数据库系统将能够基于编译时和运行时考虑在面向查询和面向导航的策略之间进行自适应选择。 所采取的方法将是扩展的技术，最近开发的面向导航的OODBMS共存的技术，从更传统的面向查询（关系）的方法。 然而，要实现这一目标，需要在性能建模和分析、开发新的数据管理算法和新的查询优化技术以及解决众多数据库系统架构问题方面做大量工作。 该项目的结果将包括数据运输和查询运输之间的权衡分析，缓存维护，查询优化和内存管理的算法，以及基于这些新技术的模拟和实现的详细性能分析。

The needs of many application domains are not met by current database system technology. Relational database systems (RDBMSs) excel at providing high-level, associative (i.e., query-based) access to large sets of flat records. Object-oriented database systems (OODBMSs) on the other hand, provide more powerful data modelling capabilities, but primarily support navigation-based access to the data. The goal of this research is to develop flexible workstation/server database systems which can dynamically adapt to the requirements of a wide range of applications, allowing those applications to exploit database capabilities and benefit from the continuing rapid gains in workstation and local area network price/performance. Such database systems will be able to adaptively choose between query-oriented and navigation-oriented strategies based on both compile-time and runtime considerations. The approach taken will be to extend the techniques that have been recently developed for navigation-oriented OODBMSs to co-exist with techniques from the more traditional query-oriented (relational) approach. To achieve this end, however, will require significant work in performance modeling and analysis, the development of new data management algorithms and new query optimization techniques, and the resolution of numerous database systems architecture issues. The results of this project will include analyses of the tradeoffs between data-shipping and query-shipping, algorithms for cache maintenance, query optimization, and memory management, and detailed performance analyses based on simulation and implementation of these new techniques.