Research On Quick Cache Algorithms In Multidimensional Exhibition

The technology of data warehouse (DW) is a new branch of database technology. This technology effectively supports information analysis system and overcomes the limit of traditional database system absolutely. Companies are building their own Data Warehouse and with the help of On-Line Analytical Processing (OLAP), Users access and observe data in multi-dimension from data warehouse in order to verify their viewpoint. Multidimensional exhibition is one of the key techniques for developing enterprise data warehouse.This thesis studies the replacement and prefetching method of Multi- dimensional exhibition caching system, and gives some effective and applicable algorithms for accomplishing this task.(1) We analyses the structure of OLAP Cache Servers (OCS) and its benefit model of cached views. Then it puts forward a weighted computational model (WDM), which uses the workflow to compute views cached benefit. WDM makes the views related with workflow get more cached benefit. By doing this, WDM could meet user's requirement extremely.(2) The traditional cache replacement algorithms can't work well in multidimensional exhibition. The dissertation put forward a replacement algorithm (BLRU) that considers both the frequency and benefit of cached views.(3) Nowadays, the research on prefetching on the result set of multi-dimensional data is quite few. The algorithms in this dissertation consult the materialized views selection technology in DW, which is used in the cache system for selecting views according to workflow and then prefetch the views. Firstly, summarize the AO View Graph that describes the relationship of the views and formalizes the concept and involved cost model. Then, we give a genetic algorithm for selecting views that considers gene automatic amendment and codes basing on AO View Graph, and then prefetch the views selected.