Research On Technology Of Date Cube Generation

With the foundation and running and developing of the enterprise's online transaction procession (OLTP) system, more and more data were created, data source were founded and distributed all of the world. People found that only depending on the OLTP system couldn't get the full privilege in the competition. The thinking on DW begins to come out and the technology and products of DW develop rapidly. Warehouse technology provides a good technological solution for the technology decision.Data cube is the core of the multi dimensional data model in the data warehouse. Technologists have already done a large amount of research work on the multi-dimensional data model and the data cube. The choice of the materialized views and the formulation algorithm of the view are two focus points in the course of the cube producing.There is no better arithmetic on how to reduce space cost, query response time, maintenance cost of data cubes, as well as to achieving a better tradeoff among these three basic features.In accord with the questions above, based on the systematic summarization of the most recent work on date cube, through the author's effort, some innovations and achievements are made by the author, which will be illustrated in detail as follows.1) By dividing the dimensions in data cube into partition dimensions and non-partition dimensions, the thesis defines an equivalent relation that makes the views with the same partition dimensions compose an equivalent class. By organizing equivalent classes into major pipelines and the views in an equivalent class into minor pipelines, the thesis presents a serial of two tiers pipeline algorithm for computing data cube. The algorithm can make use of physical memory efficiently and dramatically to reduce the times of reading raw data and the communication cost between processor.2) The Cube operator computation plays a very important role in OLAP(online analytical processing) applications. The thesis analyzed the drawbacks of the traditional pipeline method in computing Cube operators and presents the principles to materialize a data cube with multiple dimensions, and also provides an algorithm to determine the nodes in the search lattice that should be materialized.3) There are some limitations of system resources with the process of data cube created, the thesis adopts the hybrid data structure base on one is stored in a relation and the other is stored in a multi-dimension array. It adopts the merits of pipeline aggregation method and array aggregation method. It reduces the amount of pipelines and storage space dramatically and accelerates the computation of data cube.