Research Of Bitmap Index In Data Warehouse

Data Warehouse is one of the research topics in Database technique. It can manage the historical data produced by traditional database effectively, and to provide applications for various decision-makings with convincing data supports, such as on-line analysis and processing, data mining and so on. It can satisfy business enterprise customers who have the requirement of handling data deeply. Data Warehouse is environment, rather than a product, and it provides users with data now and in history for decision-makings, which are too difficult to get in traditional Database. Data Warehouse Technology is a general definition of all kinds of technologies and modules, which can provide the function of integrating operating data into unified environment for decision-makings data access efficiently to make users query and obtain the necessary information and support decision-makings more quickly and conveniently. Data Warehouse and other related technologies become focus in research and application during these years.For effective data access, most data ware systems support index structure. Nowadays, there are three kinds of index-structures mainly uses in Database, such as B-tree index, R-tree index and bitmap-index. Compared with B-tree index, bitmap-index needs less storage space, and it changes comparison, connection and integration into logic operations, which can reduce the operating time a lot in order to promote the performance greatly.However, most attribute dimensions must be layered, such as time-dimension attributes, it can be layered into year-level, month-level and day-level. Bitmap index available is usually established in a certain single level, without considering the semantics characteristic that dimension attributes have the layers. And each time we only get records of one fixed layer, which leads to time waste and operation inconvenience. And it neither ignores the operation of fast grouping and accumulating computing for case data, nor neglects to make coding prefix according to dimension-level characteristics, which induces an inefficiency in grouping and accumulating operations.This thesis mainly presents a method to code for dimension members based on the characteristic of dimension attribute having layers, in order to produce dimension layers code for mumbers, in replace of the original key words of dimension table to implement the objective of dimension key words compression. We take the dimension layer code which is much smaller than dimension table outer key words to make the function of retrieving the dimension layer code quickly, which matches the retrieval key words, to obtain the query range of all the dimension layer attributes. We can also convert a large number of multi-table connections in OLAP query into scope query in dimension, so as to reduce and simplify the multi-table connections between case table and dimension table significantly. We can make grouping and accumulating computing for the records of case table directly according to dimension layer attributes coding prefix and grouping attribute code, and save the results of cluster in outer storage, so as to reduce the cost of I/O access greatly, thereby increasing the efficiency of the OLAP queries.