Selection Of Materialized Views Of Multidimensional Data In Data Warehouse

With the development of the information society, enterprises face large quantities of data. In order to react to the market flexibly in time, how to extract the information and establish the market strategy from these data has become the key for business enterprises to be in an impregnable position in the market competition of the information age. Online Analytical Processing (OLAP) is the main way for user to obtain decision support.Decision supporting applications involve complex queries on data warehouses, which contain large amounts of data and lots of complex operations such as select, project, join and so on. Those are time-consuming processes. Since the decision support system requires its query should be responded quickly, a good physical design of the data warehouse is critical. The data warehouse system pretreats the basic data to those queries which are possibly appeared, then stores the results which called Materialized Views. One of the most important decisions in designing a data warehouse is the selection of materialized views to be maintained at the data warehouse. The goal is to select an appropriate set of views so that the sum cost of processing set of queries and maintaining the materialized vies is minimized. It becomes one of the hotspots of study in the domain of data warehouse. Besides, allied studies of materialized views include queries rewritten using materialized views, the maintenance of materialized views, the dynamic modulation of materialized views and so on.In this paper, we focus on the selection and dynamic modulation of materialized views in OLAP system which base on the relational database. A more suitable benefit model is proposed first. Then four algorithms are proposed. They are VSMF (Views Selection base on Multi-Factor), MVSCA (Modulation of Views under Space Constraint Algorithm), RMMV (Real-time Modulation of materialized views) and DMMF (Dynamic Modulation base on Multi-Factor).The algorithm VSMF can gain a good performance of query and maintenance by using MVPP as its search space and considering all factors which may affect the performance of the system. The algorithm MVSCA modulates the materialized views to meet the space constraint according to the benefit of views in unit space. In order to maintain the system responsiveness, the algorithm RMMV is proposed, which can avoid recalculating the results and estimating the size of the queries. Considering the matter from MVPP, we propose the dynamic modulation algorithm DMMF which has all the excellence of batch modulate algorithm and real-time modulate algorithm. All the experiments and analysis indicate that the algorithms proposed in this paper are effective.