Research And Implementation Of Several Key Technology In Self-Managed Database System

With the fast development of Internet and multimedia technology, the application of database has been widely employed and many techniques have been introduced in the study of advanced database strategies. Due to the increasing complexity and data scale, the cost of managing a complex database system is rising, which calls for the self-managed technology to manage a system actively and automatically. Thus, self-managed technology in database management is becoming a recent hot issue, which is studied to reduce the complexity and cost of the management and to automatically optimize the system performance.In this dissertation, the significance of the research in the self-managed database is described first. Then, the developing of the self-managed database technology is reviewed. After analyzing the defect and deficiency of the recent methods, we introduce a query optimizer's system framework integrated with index selections, materialized views and automatic system statistics diagnosis. The presented method can automatically generate the index configuration suggestions, create the optimal query plan based on materialized views and self-adaptively adjust the system statistics according to the variety of the hardware platform so to obtain improved system performance.Firstly, the index selections based on the workload and the disk constraint is studied. A framework based on feedback control principle for automatic physical database design is proposed. Then, we investigate the following key points of the models in the framework: the candidate indexes generation, the"what-if"configuration analysis and the optimal configuration searching method based on greedy algorithms and genetic algorithms.Secondly, we study the implementation and application of the materialized views, the main point lies on the realizations of the maintenance and the query rewriting. For this part, the maintenance method and implementation of materialized view is presented first and the key research focus is on the incremental maintenance algorithm: the method based on the subsumption graph to support the materialized views with outer-join, the maintenance algorithm of materialized view with aggregation based on progressive updating and the updating algorithm based on the partial recalculating of materialized view. In the following part, the query rewriting using materialized views is investigated, including the semantically matching condition and algorithm with a view, and then we propose the optimal rewriting plan selection method base on binary search tree.We also study the method and implementation of the automatic system statistics diagnosis and present the realizable mechanism. Then the automatic collection method of the system statistics is proposedFinally, with Shenzhou OSCAR database as the platform, we implement the above key technologies and provide experiments for functional and performance test. The relevant results demonstrate the validity and high effectiveness of the proposed techniques, which can reduce the work load of DBA and optimize the system performance automatically.