Research And Application Of Materialized View Technology Supporting Fast Online Analytical Processing

By OLAP queries the useful information behind the data for decision-making analysiscan be find quickly. However, OLAP queries often require large amounts of operations suchas selection, projection and join on raw data, which is a very time-consuming and computingresources-consuming process. In order to get fast online analytical processing operations,materialized view is introduced to speed up the OLAP query. Materialized view is a tablewhich saves the pre-computed query results. With the help of the materialized view, OLAPqueries do not have to do complicated selection, projection and join operations on raw data,just get the results information from the materialized view. In this way, it greatly reduces thequery response time and improve query efficiency.But the introduction of materialized view brings new problems. Firstly, materializedview increases the cost of the data warehouse storage. Secondly, maintenance of materializedviews which involves complex computation on mass data consumes a lot of time andcomputing resources. How to choose appropriate views to be materialized so that the queryefficiency is as high as possible and storage cost and maintenance cost is as little as possible isdescribed as materialized view selection problem. How to choose the suitable method toquickly update materialized views is described as materialized view maintenance problem. Inorder to solve these two problems, this paper puts forward the materialized view selectionmethod based on genetic algorithm and materialized view maintenance method based on dataprovenance. The main research work and achievements are as follows:1. It expounds the main problem after the introduction of materialized views in datawarehouse. It describes domestic and foreign research status about materialized viewsselection and maintenance. It elaborate related technologies about materialized view selectionand maintenance in detail.2. According to the fact that materialized view selection problem has proven to be aNP-hard problem. It puts forward a kind of materialized view selection method based ongenetic algorithm. In addition, it converts materialized view selection problem inmulti-dimensional lattice model to the problem of the genetic algorithm get the optimalsolution. The adaptive adjustment mechanism of crossover probability and mutationprobability is introduced to the solve process of genetic algorithm, which speeds up theconvergence speed and avoids premature convergence.3. It utilizes data provenance techniques to realize incremental maintenance. To aggregation materialized view and non-aggregation materialized view, puts forward theirrespective incremental maintenance model and incremental maintenance methods. Theexperiment shows the algorithm is feasible and has better performance.4. It applies the materialized view selection method based on genetic algorithm andmaterialized view maintenance method based on data provenance to select and maintainmaterialized views of vehicle analysis system. The application shows these two methodsgreatly improve the query efficiency of system and consume the least amount of time inresponse to query requests.