Research And Implementation Of Index Selection Strategy In DWMS

In order to provide decision makers with comprehensive information, data warehouse contains a large amount of historical data, which not only reflects the department's overall situation, but also contains the details of a number of business data. Meanwhile, the data warehouse should support a complex multi-dimensional analysis and ad hoc queries, and focus on storage space savings. In order to improve query efficiency, there is an effective way for the index technology to achieve a reasonable query response time and a balance among resource constraints. Therefore, as one of the key issues of designing and building data warehouse, the index selection has become a hot point of researching current data warehousing.Based on Data Warehouse Management System (DWMS), we primarily focus on core issues and key technologies of index selection in data warehouse that are specific to research. This paper takes analysis in depth for DWMS storage system structure and implementation mechanism. Based on this, two kinds of indexing techniques, B+ Tree index and Bitmap index, including their design and implementation principle are introduced in details. Combined with the B+ Tree index and Bitmap index that have been available, this paper designed a planned experiment which is used for analyzing of how index type affecting the query efficiency in data warehouse. And we summed up the relationship between the influencing factors, proposing the index design strategy in data warehouse.According to data organization and query features which affect the performance of index, we design four experiments to test the efficiency of index from different suspects comprehensively; we propose index design strategy of data warehouse through research and experimental comparisons.Finally, this paper addresses an index selection issue in physical design by proposing a heuristic approach which selects an optimal index set to be built in a data warehouse. To achieve this goal we simulate an optimizer which generates query execution plans and define a cost model to evaluate them. And then we outline a greedy algorithm which chooses the most promising ones from a set of candidate indexes.