Research Of Key Technology Of Index In Column-Oriented DWMS

As the most popular data analysis tool, data warehouse not only provides a strong support for corporate decisions, but also has drawn increasing attention. Because of the amount of data in data warehouse is very huge, and the data analysis and decision support are complex functions, so, addressing the high efficient mass data access and supporting ad hoc queries are two important issues. Index of concern is the data access mechanism, the index of well-designed not only to provide good data organization, but also maximizes the performance of queries.In DBMS, traditional row-store showed good performance, and is widely applied in the enterprise applications. However, it can not provide efficient query analysis performance for enterprise decision. In recent years, column-store has been reviewed. In data warehouse and analysis application, to support read optimization inquires work, column-store showed more superior performance than row-store. The existing column-stored systems, such as commercial database system, patent system and open source system etc., all on business intelligence and data warehouse reflects the obvious superiority.Index is a key technology to promote the accessing and querying efficiency of the mass data. Based on the index technology in the application of column-stored DWMS as the core content, this paper emphatically analyzes the realization of B+tree index in column-stored DWMS, and its improvement. And put forward the join index in the column-stored DWMS.The traditional B+ tree index has good performance in the update transaction environment. However, in the analysis column-stored and data warehouse querying environment, the cost of space and time is greatly. Based on the characteristics of the query in column-stored data warehouse environment, this paper put forward a new kind of tree index - RB+tree (Reduced B+tree). This index is improved from the traditional B+tree. And combined with the bottom-up creation of the index tree, making the space utilization, efficiency of creation and querying the index all is improved significantly.By applying the RB+tree in column-stored data warehouse, we can establish the rowid index and value index. Specifically, to solve the multi-table joining of the star model, we put forward the join index. It effectively improves the efficiency of tuple reconstruction and multi-table joining in column-stored data warehouse. The experiment result on the data warehouse benchmark SSB verified the validity of this method.