An Object Flow Computer for Object-Oriented Database Applications

本课题的重点是为大型数据库应用设计和评估一个对象流计算机。这种并行系统的主要特点是：(a)直接支持面向对象的语义关联模型，它可以比现有的数据模型更明确地建模复杂的数据；(b)采用两阶段查询处理技术，可大幅减少在并行处理器之间传送的数据量；(c)使用基于数据语义属性的垂直碎片模型，这增加了并发程度，并最大限度地减少了不必要的数据传输；并且(d)集成了一些软件和硬件技术来解决非常大的数据库管理问题。研究了在硬件和软件中设计面向对象处理函数时的成本-性能权衡分析。面向对象的数据库管理系统具有许多理想的特性，包括数据抽象、继承、多态、信息隐藏以及数据和元数据的统一处理。使用语义建模的面向对象数据库管理系统的扩展在支持高级应用程序（如CAD/CAM、CASE和办公自动化）方面具有很大的前景。然而，许多基本技术及其在硬件和软件中的实现还没有得到很好的理解。本项目从分析和实验的角度考虑了技术和应用需求的设计权衡。开发的结果将导致更有效和更快的面向对象数据库管理系统。

This project focuses on designing and evaluating an Object Flow Computer for large database applications. The main features of this parallel system are (a) the direct support of an object-oriented semantic association model, which can model complex data more explicitly than existing data models; (b) the use of a two-phase query processing technique, which can drastically reduce the amount of data to be transferred among parallel processors; (c) the use of a vertical fragmentation model based on semantic properties of data, which increases the degree of concurrency and minimizes unnecessary data transfers; and (d) the integration of a number of software and hardware techniques to solve very large database management problems. Research is carried out on analyzing the cost-performance tradeoffs in designing object-oriented processing functions in hardware and software. An object-oriented database management system possesses many desirable features, which include data abstraction, inheritance, polymorphism, information hiding, and uniform treatment of data and meta-data. The extension of object-oriented database management systems with semantic modeling holds great promise in supporting advanced application, such as CAD/CAM, CASE, and office automation. However, many of the essential techniques and their implementation in hardware and software are not well understood. This project addresses the design tradeoffs subject to technological and application requirements from an analytical and experimental points of view. Results developed will lead to more efficient and faster object-oriented database management systems.