| Data warehouse is a subject-oriented, integrated, relatively stable, reflecting the historical changes in the data set used to support management decision-making. Data warehouse data is very large, and increasing year by year. Therefore, the data warehouse to improve query speed is a function must be implemented.An Index can impact the query speed directly in the data warehouse. There are many types of index, create index blindly can improve query execution speed, but brought a hard work to the index maintain and management cost.How to establish an appropriate index on an appropriate column to improve the query performance in a variety of analysis in the multidimensional data warehouse, is an opportunity and a challenge.This paper makes a research and analysis on the index technology in the column-oritented data warehouse system, works as follows:(1) Study the index technology on the current column-oriented data warehouse system (such as Sybase IQ, C-Store, MonetDB, Infobright), analyze the architecture,feature and SSB data set structure in the column-oriented data warehouse management system D3WMS (Donghua Dameng DWMS).(2) Study and analyze the RB+ tree index in the D3WMS, verify its performance in the space utilization and query speed. Change the structure of its leaf nodes to improve the space utilization of it.(3) According to the low base data, design and realize the bitmap index, and proposes a directly calculation method CBT(Compressed Bitmap Index Execute Technology) on compressed bitmap index which can optimize the the query algorithm on bitmap index. RB+ tree index only divides the data type into fixed-length and variable-length, totally creating the rowid and value index on them. This classification ignores the base of data, which will cause a large space redundancy. Bitmap indexes can effectively improve the redundancy of low base data, and greatly improve space utilization and query efficiency.(4) For the high base, hierarchy data, this paper presents the design of the level encoded bitmap indexes, which can significantly reduce the index storage space and improve the query speed on the columns.This experiment is based on D3WMS and SSB data set, and covers six topics as follows:1) Comparition between RB + tree index and traditional B + tree index; 2) the space utilization contrast of the improved RB + tree index and the RB + tree index; 3) The query performance between CBT and origenal bitmap index based on compressed bitmap index;4) The space comparation between compressed bitmap indexes and origenal bitmap indexes; 5) Achieve the performance of grouping queries with CBT; 6) The space using with the levels encoded bitmap index. Experiments show that the proposed indexing method can effectively reduce the index space usage, and improve query efficiency. |
