Research Of Materialized View Selection Algorithm In Data Warehouse Based On ROLAP

A data warehouse is a, by definition, subject-oriented, integrated, non-volatile and time-variant repository of data to support decision for manager. It stores massive multi-dimensional historical data and it is for rogatory and analytical user. The query statement which user submits is usually necessary to search massive volume data and involve complex multi-table join operation. As to these complex queries, data warehouse must give fast response. It is well known that materialized view technology is an effective method to enhance query response performance. However, materialized view needs to occupy system space to store, and it spends system cost on maintaining.Therefore, how to choose an appropriate group of materialized view in the space constrain is an important research topic.Generally speaking, it needs to apply static materialized view selection algorithm to obtaining a materialized view set, and then apply dynamic materialized view selection algorithm to adjusting the set for its timeliness.The paper makes improvement to materialized view selection algorithm from two aspects, static and dynamic, respectively. The former proposes static materialized view selection algorithm based on improved genetic algorithm with the base of introducing accept probability. It accepts the individual whose fitness becomes lower with a certain probability, and gives genetic algorithm an opportunity to escape local optimum and achieve global optimum finally. In addition, it calculates immediately individual fitness after genetic operation and processes individual, which makes it search in assigned orientation instead of blind search. The experimental result shows that the improved genetic algorithm not only solves the"premature"problem of classic genetic algorithm, but also avoids the blind search of classic genetic algorithm.With the base of proposing materialized view similarity function, the latter proposes clustering-based dynamic materialized view selection algorithm. It firstly clusters materialized view, and then dynamically adjusts materialized view set. So, it eliminates the"jitter", which dynamic materialized view selection algorithm generally has. The experimental result shows that the algorithm not only improves the overall query response performance, but also reduces the computational cost which will be spent during updating materialized view.