Study On The Structural Characteristics And Partitioning Methods Of Data Cubes

Data cube is the core data model of data warehouse. Its corresponding data cube lattice retains the semantic relationship of roll-up and drill-down, which is conducive to research and navigate. This paper studies on the internal relations of data cube as a graph. Despite currently intensive studies in random networks and complex networks, it doesn’t form systematic study on structure characteristics and analytical model of data cube lattice. Compared with random networks and complex networks, this paper finds that data cube lattice features different structure characteristics in terms of degree distribution, clustering coefficient and Average shortest path. In addition, this paper analyses the analytical model of data cube lattice.Finally, Based on the structure characteristics and analytical model, partitioning across data cube lattice is designed, then this paper compared the differences of partition methods through experiment.The primary contents are as follows:(1)Lattice data are seen as graph data and its structure statistics in terms of degree distribution, clustering coefficient and Average shortest path are firstly studied. Compared with random networks and social networks, and figured out the relations and differences between data cube lattice and the other networks model.(2) Based on the specific structure characteristics of data cube lattice, this paper deeply studied on the internal laws of data cube lattice, then analyses the analytical model of data cube lattice.(3) Based on the structure characteristics and analytical model of data cube lattice, considering the space complexity and the time complexity of data computing.With high efficiency analysis on the data cube lattice as the target, this paper designs partition methods on lattice partition, so that massive lattice structured data of one to ten million nodes are expected to be queried and analyzed efficiently.