19^th International CODATA Conference
Category: Multi-disciplinary Use

Cross-domain Database Support for Large-Scale Scientific Raster Data Sets

Dr. Peter Baumann (baumann@rasdaman.com)
rasdaman GmbH, Germany
http://www.rasdaman.com/

Sensor, image, and statistics data appear in practically any technical/ scientific domain. We can abstract and generalize a large class of these data into raster data, i.e., data cubes with single values sitting at grid points in a regular lattice. Examples are 1-D sensor time series, 2-D satellite maps, 3-D voxel tomograms and geophysical data, 4-D climate models, and n-D statistical data. Raster data pose particular data management problems due to their gridded structure which is not supported well by traditional database technology, and due to their extreme volumes - single objects can well be in the multi-Terabyte range, for example as satellite image maps. On the other hand, fast, flexible retrieval support is becoming more and more an issue, as scientists increasingly want to have Web access to large, distributed data assets using not only low-level standard access methods like spatio-temporal subsetting, but also versatile ad hoc analysis. In our talk we show feasibility of domain-independent database support for Web-enabled access to large scientific raster data assets based on the rasdaman system which is implemented and in commercial use. After an overview of architecture and retrieval mechanisms we present implemented applications from the fields of satellite imagery, climate modelling, astrophysics, and human brain research. A small live demo illustrates the concepts.