Design And Implementation Of Database Parallel Query Optimization

This paper adopted the parallel data flow method which is based upon data partition, combined with query processing of Gbase8a database, proposed the two-stage parallel query optimization method, realized the parallel query ability of Gbase8a database on the SE parallel computer. This paper gave the design and implementation of the parallel execution module, and the results of experiment showed that the parallel query optimization module designed in this paper, through the database parallel execution, greatly reduced the response time of database query processing and ensured the correctness and stability of the database product.The first chapter of this paper is a brief introduction of the background, the main object and the structure of the paper. The second chapter focuses on a number of database theories and technical basis of database parallel query optimization, includes the inherent parallelism of database, parallel database technology and development trends.The third chapter is the overall design of parallel query optimization module. According to Gbase8a's query process, this paper designs and implements a two-stage parallel data flow optimization method based on data-partitioning: the first stage is the parallel optimization on Gbase8a's executing plan, completes the partition of the data with a serial number of instructions; the second stage is the optimization on the instructions, executes instructions paralleled with multi-threads. Chapter IV is the concrete realization of Parallel Optimizer. Parallel Optimizer works at the stage of instruction-based query optimization of Gbase8a, and the main function of Parallel Optimizer is to achieve the data partitions and increase the parallel degree of query. Chapter V is the concrete realization of Parallel Executor. Parallel Executor consists of three parts: series-instructions execution module, run queue and single-instruction execution module. Actually, Parallel Executor implements a simple parallel data flow method, due to the dependence between instructions, parallel effect of Parallel Executor is limited if used separately. Combined with Parallel Optimizer and Parallel Executor, this paper implements the parallel data flow method which is based upon data partition.Chapter VI describes the relevant experiments and the results are analyzed. Based on the results and analysis, this paper receives the following conclusions: The parallel query optimization module designed in this paper, can effectively reduce the response time of database query processing through the parallel execution. With the increase of CPUs used, query processing performance of Gbase8a continues to be improved. However, the problems includes: data combine costs too mush time, and as well as the lack of parallel optimization on some query operators, the overall parallel performance of the database is limited.The main job of this paper, content of our research and what will be done in the further are introduced in chapter VII.In conclusion, this paper designed and implemented a complete parallel query optimization module for Gbase8a database system, and the results of experiments showed that the query processing performance has been improved. It is a prototype of parallel database and can be promoted.