Research And Implementation Of Query Optimizer Based On Self-Tuning Statistics Management

Database technology is a rapidly growing technology in computer science, and is a widely used technology. It is already the core technology and significant base of the computer information system and application system. As system software, the performance of Database Management System (DBMS) will impact the performance of application system directly.The application of self-tuning technology in DBMS is a researching focus in recent years. Foreign DBMS manufacturers have provided their self-tuning DBMS products, but seldom inland. Self-tuning optimizer is an important aspect of the application of self-tuning technology in DBMS. This dissertation mainly researches the design and implement of self-tuning statistics management based optimizer which could tune database object's statistics according to the need of optimizing and modifying of database data of oneself, tune system statistics according to the DBMS running environment of oneself, tune optimizing strategy according to the characteristic of query of oneself, and tune the method of selectivity estimation according to statistics'state and the optimizing deepness level.This dissertation is started with analyzing the significance of researching self-tuning query optimizer. The author then researches the status quo and common technology of statistics, self-tuning statistics management and self-tuning query optimizer home. A holistic structure of self-tuning statistics management based optimizer is designed in this dissertation, and the functions and the design thinking of every component are introduced detailedly.This dissertation researches self-tuning statistics management by dividing it into automatic statistics management and multidimensional statistics management. Firstly, the author researches the design and implement of automatic statistics management. This dissertation analyzes the holistic structure and module construction of automatic statistics management system, and analyzes every module's design thinking and work principle. This dissertation then researches the statistics computing method and storage method, and researches how to estimate selectivity using statistics home at last.This dissertation also researches the application and implement of multidimensional statistics. After researching wavelet technology and wavelet based histogram technology, this dissertation proposed a novel method to construct wavelet based multidimensional histogram by decomposition of multidimensional histogram, which improves the current method. Of course, how to construct and store multidimensional histogram using this method and how to estimate selectivity using this kind of multidimensional histogram are also researched detailedly in this dissertation.As a system, this dissertation researches the detailed structure design of self-tuning optimizer. The author not only researches the application of self-tuning statistics management in self-tuning optimizer, but also researches other aspects of self-tuning optimizer, including self-tuning statistics opting, self-tuning system statistics adjustment, optimizing deepness analyzing and deep optimizing.The effectivity of the research in this dissertation has been testified by strict experiment. The author then summarizes all researches and the innovation in this dissertation. And as for the defects, the author puts forwards the future research at last.