Abstract

The critical parameters of multi-dimensional data analysis are its data size and the number of computations. The data size is typically large and the numbers of computations are immense. The complexity increases further when the dimensions and the resulting intersection space are larger. The processing involves data access, required computations and aggregations considering all dimensions with the multi-level hierarchical grouping. The large amount of data and resulting computations require huge processing power and memory, which are usually found in Symmetric Multi-Processing (SMP) machines. Here we present architecture to analyze multidimensional data using a cluster of regular desktop computers which do not share anything between them. All the tasks like building the multidimensional cube, storage of data, data query and the computations are distributed across the standalone nodes of the cluster. Thus, we can achieve infinite scalability and excellent performance.

Authors and Affiliations