Establishing more suitable distributed plans for MultiNode-OLAP systems

Abstract

The materialization of multidimensional structures is a common way to speed up mostly aggregation queries. Its proper selection has triggered a huge variety of proposals. Any algorithm that aims to solve the cube selection problem needs a way to estimate query and maintenance costs, as minimizing costs is the objective function. This estimation is harder if a distribution of multidimensional structures is concerned, as algorithms as to deal with an additional dimension-space translated as a set of spatially distributed storage and processing nodes, interconnected with a communication network. In this paper we present some algorithms especially designed to estimate query and maintenance cost at a Multi Node OLAP system in time units, using a distributed non-linear generalized cost model.