Research On The Dynamic Selection Strategy Of Materialized Views In Data Warehouse

Data warehouse stores large amounts of data , it supports the complexity of query forms, how to satisfy the standard of the query efficiency is one of the problems faced by Data Warehouses, materialized view is an effective means to solve this problem. Howerver, the materialized view need extra system space to store ,and it need maintenance at price of system cost, we must balance between the response time and resource constraints, selecting appropriate materialized views. Therefore, the selection of materialized views is an important research topic.This paper works on the selection of materialized views,the selection of materialized view is how to minimize the sum of query processing cost and materialized views maintenance cost, there are two ways to solve this problem : the first way is to construct search space for candidate views ; the second way is to optimize the selection algorithm of materialized views. This paper proposed a new strategy of constructing search space for materialized views selection IMVPP(Improved Materialized—View Processing Plan)based on MVPP. It constructs search space by calculating the common subexpression of the input queries, which calculated the max common subexpression can be shared by the other queries,and can rewrite the input queries . It is of benefit to reduce the seach space and improve the querying efficiency.The selection strategy of materialized views is divided into static selection and dynamic selection. Among them, the static selection strategy of materialized view supposes the queries are distributed in integrated data evenly,or the user can provide the probability distribution of their inquiries , in fact, with the data warehouse system operation, dynamic query will lead that the benefit of the part materialized views decreases, which making the total benefit of materialized views decreases. The problem may be resolved through the dynamic selection strategy of materialized views, which collects the information of queries continuously during operation ,and selects materialized views according to the queries in system actually, when discovers the existing materialized view is no longer suited to the new query, reselects materialized views according to the new set of queries.The existing methods being used to select views do not take into account the characteristic of various kinds ofusers , and therefore have some defects. This paper proposes a new dynamic and user-oriented selection strategy of materialized views DSUOMV (dynamic selection algorithm of user-oriented materialized view) based on IMVPP, which is composed of four algorithms: CVSA (candidate views selection algorithm), DSAW (dynamic space allocating in windows), UMVS (user-oriented materialized view selection), and DMVW (dynamic modulation of materialized views in window). The strategy takes into account the diversity and dynamic of user , the users are divided into different group according to their queries characteristics, and different groups is assigned different view window. The DSWA algorithm is used to adjust the allocation of space between the user window, the UMVS algorithm is used to select appropriate view to materialize, and the DMAMV algorithm is used to determine whether it is necessary to change the materialized view set by measuring the change of user queries. From the experimental data and comparative analysis we can find the above strategy has some advantages.