Research On Query Optimization And Multi-dimensional Data Model For OLAP

OLAP (Online Analytical Processing) is an important service provided by data warehouse and the primary means of decision support for users. OLAP queries usually take in large data with the aggregation of the complex multi-joining tables, request to provide users with timely analysis of data and can analyze data in multi-direction and great depth for system users. With the practical application, the users may increase the analysis view and update the data. These make how to improve the OLAP query and efficiency of analysis operation, and better maintain the multi-dimensional data model as a key issue of data warehouse in the application.This thesis focuses on issues more systematically and in-depth study. A practical index based heuristic approach for query optimization is proposed and combine the query technology of MDX on the project of business analyses support system in Shanghai Mobile, which improve the OLAP query and the analysis operation of multi-dimensional data. At the same time under the traditional multi-dimensional model, a new one based on dimension hierarchy updating is proposed, which has better support for updating the dimension hierarchy of the multi-dimensional data model. The main contributions of the thesis are as follows:(1) A practical index based heuristic approach for query optimization is proposed. By considering the threshold of evaluation spending,this approach reduces the scan number of fact table with the index.In advance carry out the joining operation to avoid losing the index of nested subquery when multi-joining of fact tables with indexes involve,thus enhancing the efficiency of CPU implementation. The experimental results show that the querying cost and executing time is significantly less than the traditional heuristic approach. Especially when the data numbers of the fact tables increase, the method can improves the efficiency of OLAP queries more effectively.(2) A multi-dimensional data model based on the updating dimension hierarchy is proposed. The thesis increases the Boolean variable and placeholder on the traditional multi-dimensional data model, which make it more flexibility and scalability. The model can more roundly sustain the updating operation of the structure of dimension hierarchy (dimension level attribute,dimension level, dimension correlation between levels). Analyse the examples to prove the validit model which has a strong practicality.