Institutional Repository
| 基于MPI和MySQL的并行数据库系统的研究与实现 | |
| 王璟 | |
| Major | 计算机软件与理论 |
| 2004 | |
| Degree Grantor | 中国科学院软件研究所 |
| Degree Level | 博士 |
| Place of Degree Grantor | 中国科学院软件研究所 |
| Keyword | 并行数据库 集群 |
| English Abstract | 在数据库技术不断发展的今天，数据量也在不断膨胀，现有的串行数据库技术已经越来越无法适应数据增长的要求。因此，象众多其他领域一样，并行化成为提高数据库系统性能的必然手段。本文创新性地提出了以MPI和MySQL相结合来创建并行数据库系统中间件的思想，其核心就是要在MPI及MySQL与上层用户层之间建立一个并行数据库系统的中间件。该中间件的作用就是屏蔽MPI和MySQL操作并向上层提供并行数据库操作界面。在该创新性思想指导下，本文还创新性地对并行集群搭建软件OSCAR进行了彻底的尝试，打破了以往的集群安装的繁琐模式，为中心今后的科研工作提供了一条很好的通路。本文的主要研究成果如下：1．在构建集群过程中创新性地采用了开放源码的软件包。OSCAR(Open Souce of Cluster Application Resource）进行集群系统的构建。2．在OSCAR的基础上建立了一整套软件硬件的研究平台。3．提出了基于MP工和MySQL的并行数据库系统的中间件软件MZ的理论框架。4．实现了很多并行数据库查询操作模块。如并行排序、并行选择、并行连接、并行投影。5．实现了很多提高系统可用性的重要模块。如批量载入模块、格式化输出模块、用户登陆模块。6．对整个系统在多种平台下进行全面的测试并进行了详细的测试结果分析，整理出了提高整个系统性能的关键点。 |
| Abstract | With the hardware technology approaching its physical limitation, building a parallel DBMS that can run on a cluster becomes the only way to increase the performance of database at present. Now, cluster has been applied in many areas such as weather forecasting, oil industry, CAE, CFD, Web, etc. And there are many successful cases prove that cluster can increase the performance of database. This thesis proposed a middle-ware structure based on MPI and MySQL, which can help us use a parallel database without knowing how to use MPI or MySQL. And this thesis also try to build a Beowulf cluster with a good tools package, which is called "OSCAR". The major contributions in this thesis are detailed as follows. Firstly, this thesis tries to use a useful package called "OSCAR" to build a Beowulf cluster. Secondly, this thesis builds a platform on which we can perform our research and development. Thirdly, this thesis try to build a parallel DBMS named M2 using MPI and MySQL, with focus on special application, and not for general situation. Fourthly, this thesis implements some important modules including parallel sort, parallel-join, parallel-projection and parallel-select, which prove that the idea building a parallel database with MPI and MySQL is feasible. Fifthly, this thesis implements some important modules, which can improve the system's usability. These modules include database conversing module, output module and connection module. Sixthly, this thesis performs many tests on M2 on different platform, and gives many analysises in detail. As the result, this thesis get the emphases on increasing the performance of the Parallel Database System. |
| Pages | 95 |
| Language | 中文 |
| Content Type | 学位论文 |
| URI | http://ir.iscas.ac.cn/handle/311060/6200 |
| Collection | 中科院软件所_中科院软件所 |
| Recommended Citation GB/T 7714 | 王璟. 基于MPI和MySQL的并行数据库系统的研究与实现[D]. 中国科学院软件研究所. 中国科学院软件研究所,2004. |
| Files in This Item: | ||||||
| File Name/Size | DocType | Version | Access | License | ||
| LW014051.pdf（2697KB） | 限制开放 | -- | Application Full Text | |||
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment