Answering multidimensional queries on cubes using other cubes

Recently there is an important interest in On-Line Analytical Processing (OLAP) technology. In this context, in order to facilitate complex analysis, data are usually modeled multidimensionally where multiple hierarchies are associated with the dimensions. These multidimensional (MD) data structures are called data cubes. In the existing OLAP products, the user interaction is limited to one operation at a time. Further computing OLAP operations is very expensive since sequential scans are required. In this paper we provide a simple data model for MD databases, and a simple algebraic MD query language that permit the modeling of the principal OLAP operations. The MD query language allows the user to directly specify the result. Therefore, optimization techniques can be applied globally to the MD query evaluation. We state declarative conditions for answering queries on cubes using exclusively one or more precomputed queries (derived cubes). Then, we provide instance independent expressions that compute an MD query on a cube from derived cubes. These results can be used to increase availability of data and to improve MD query performance.