With the development of information technology and the increasing amount of data,the traditional row-store database is suffering a performance bottleneck in processing and querying massive data.Since the column-store database takes advantage of its unique storage structure,it enhances the system performance significantly.However,due to various reasons,the column-store database system cannot apply to the real scenario universally.Hence,the thought of column-store within row-store,which is called simulative column-store,provides a way to optimize traditional database.The thesis mainly researches the storage and optimization of database system based on simulative column-store.The simulative column-store technologies are investigated including vertical partitioning,all-index,and materialized views.Furthermore,based on the vertical partitioning technology,the thesis optimizes the system performance.Based on the simulative column-store,the thesis introduces light-weight compression algorithms,adaptive multi-projection methods,and table redundancy eliminating optimization strategies.During the research,a variety of classic compression algorithms are implemented in the system.Moreover,a more efficient light-weight compression algorithm called RLE-dictionary encoding and an adaptive multi-projection algorithm called correlative attribute clustering(CAC)are put forward to improve query performance of simulative column-store.Finally,the thesis presents a series of system optimization schemes to optimize the massive transactions data manage system of a financial institution.The thesis processes the tables according to the features of data itself,then applies simulative column-store technology and implements relative optimized modules to the system.The experimental analysis demonstrates that these optimization techniques enhance the performance in the certain OLAP query scenarios. |