Study On The Strategies Of Increnmental Maintenance Of Materialized Views In Data Warehouse

Almost all data that maybe stored in the forms of materialized views (MVs) in data warehouse (DW) are obtained from multiple basic transaction databases that are separated, possibly heterogeneous and autonomous. The query responding time can be speeded up and the performance of the DW maybe improved obviously by using the presorte of materialized views. However, the data in the basic transaction databases are often changed, in order to make the updating of MVs synchronism with the updating of basic transaction databases, the related data in DW must be maintained in time.The maintenance of MVs is a fundamental technique in DW. With the continual increment of the amount of data source and volume of data, higher requirements are needed for the efficiency of maintenance of MVs in DW, so the completeness updating strategy may not be a feasible. Study of the incremental maintenance of MVs may have the important theoretic and practice significances.After introducing basic concepts of DW and MVs, the three tier maintenance architecture of DW is improved by adding a component wrapper between DW and DW base (DWB), so the valid data form of DW can be converted from current data base duplication, and the increments of basic transaction DB can be obtained by using the triggers.Aimed at some shortages such as the lower efficiency and incomplete disposals in the present MVs' incremental updating algorithm, a way of combining the cleaning old data and keeping track of history is proposed to improve the incremental updating of dimension tables. While cleaning the old data, how to make a real-time decision accurately and efficiently is very important, and the users' requirements must be combined to get the best balance point. The current flag indicator and aggregation related indicator are used to improve the MVs' incremental updating algorithm, so the updating process can be simplified, some unnecessary corrections of fact table are avoided and the incremental updating efficiency may be raised.The current incremental algorithms of aggregation functions are only adapted to somespecial conditions and have no generality. According to their characteristics, aggregation functions are divided into two types of self-maintenance and non-self-maintenance, and the incremental updating strategies of aggregation functions are designed completely for different cases.In order to test the function and efficiency of the improved algorithm, both completeness updating and incremental updating are used to deal with MVs' maintenance in the experiment, the improved algorithm is proved effectively in MVs' maintenance in DW by comparing the running time and memory.