Research On Quotient Cube Technology

Data warehouses and OLAP tools are based on the multidimensional data model. Data cube is the core of the multidimensional data model. We usually need to pre-compute the data cube and materialize it to improve query performance in OLAP. But the size of data cube will increase explosively when all the cuboids are pre-computed. Reducing disk space cost and improving query performance are two very important goals of the data cube research. To solve these problems, we need to explore effective data cube structures.Quotient Cube is a summary structure for a data cube that preserves its semantics. Quotient Cube partitions a cube lattice into disjoint classes and uses them to express the cube lattice. Each class contains one or more cuboids which aggregate from the same set of base relation tuples. For each class, Quotient Cube only saves its upper bound and lower bound(s), other cells contained in the class could be induced from the bounds. Quotient Cube reduces disk space cost dramatically.Recently, people designed a new structure called closed cube, which saves only the upper bounds of the equivalence classes and, therefore, has higher reduction ratio. But closed cube just aims at antimonotonic iceberg conditions. All the classes which satisfy such conditions are in touch with each other in the cube lattice. The lower bound of a class can be worked out by the upper bounds of other classes. If the iceberg condition is nonantimonotonic, equivalence classes are dispersed in different places of the cube lattice, we still need to save both the upper bounds and the lower bounds of equivalence classes with Quotient Cube in order to keep the boundaries of the equivalence classes.In this thesis, we present a new method for computing iceberg Quotient Cube, especially for nonantimonotonic iceberg conditions. We first design a storage structure IQ-Tree, then introduce the way of pruning by computing tight bounds for aggregate functions. Subsequently, we present our method for computing iceberg Quotient Cube, IQ-Cubing. Finally, we compare IQ-Cubing with the early proposed method by experiments, from which we could see that IQ-Cubing has great advantage in performance.