A Study on a Parallel Database Processing System based on Data-Driven Paradigm

本项目的主要目标是建立一个基于动态数据驱动原理的高性能、高可靠性的数据库系统的实现和维护的新范式。为自然地描述数据库应用程序，建立了多层次的数学规范描述方案。为了满足用户对数据库应用的需求，在该环境中采用了多边数据结构描述方案，即树型、网络型和表型结构描述方案.一个增强的动态数据驱动处理模型结合历史敏感的过程被证明是最有前途的原则下，从数学规范派生的程序是在高度并行的方式执行.采用这种处理模型，可以直接将数学规范转换为相应的数据驱动的数据库程序.提出了一种在动态数据驱动的多处理机上进行程序和数据划分的算法。通过提供流类型的数据结构，程序可以探索结构化数据处理中固有的流水线和并发并行性。通过引入基于面向数据方法的通用实现，在工程工作站上建立了一个原型系统。在原型系统中，我们的动态数据驱动的多处理器系统的可执行数据库程序生成的数学规范。

The main objective of this project is to establish a new paradigm of implementing and maintaining database system which has high performance and reliability based upon a dynamic data-driven principle.The results obtained during the project are summarized as following :1. Multilateral diagrammatical specification description scheme was established for natural description of database applications. For the satisfaction of users' requirements on database application, we have adopted multilateral data structures description schemes into this environment, i.e. tree, network and table structures description schemes.2. An augmented dynamic data-driven processing model incorporating history-sensitive process was proved to be the most promising principle under which the generated programs derived from the diagrammatical specifications is executable in highly-parallel fashion.3. Adoption of this processing model enables to directly transform the diagrammatical specifications into the corresponding data-driven database programs.4. A partitioning algorithm of the program and data onto a dynamic data-driven multiprocessor was proposed. By supplying a stream type data-structure, the program can explore the pipelined and concurrent parallelism inherent in the structured-data handling.5. A prototype system was built on an engineering workstation by introducing generic implementation based upon data oriented approach. In the prototype system, executable database programs for our dynamic data-driven multiprocessor system are generated from the diagrammatical specifications.