Rethinking Spatial Processing in Data-Intensive Science

Abstract

In this paper we address the problem of supporting data-driven science in geo-scientific application domains where scientists need to filter, analyze, and correlate large data sets in an interactive manner. We identify ten fundamental requirements for such a new type of processing system. They span from supporting spatial data types over using workflows for data lineage, to fast and flexible computations for exploratory analysis. Our study of related work looks at a range of established systems of different domains and shows significant drawbacks with respect to our requirements. We propose the Visualization, Aggregation \& Transformation system (VAT) as an architecture to overcome current limitations. It bases on standard software and implements performance-relevant parts using state-of-the-art technology like GPU computing. Our first comparison with other systems shows the validity of our approach.