A Massively-parallelised Spatially-registered Data Structure for Integrated Subsurface Data Management

We demonstrate the application of a massively parallel architecture to subsurface datasets, specifically Terabyte scale multi-attribute seismic volumes and derived interpretation. We decompose subsurface data into spatial, attribute and metadata components. Subsurface attributes are stored by their location whilst retaining the spatial scope (e.g. bin size), their original format (e.g. SEG-Y) and source file in a MPP storage array. We have developed a hashing algorithm which permits computationally efficient recovery of multi-resolution data as a near-instaneous operation regardless of the size and complexity of the dataset. We demonstrate efficiencies both computationally - when working with derived seismic parameters and attributes (e.g. instantaneous phase, variance); and in retrieval - when extracting arbitrarily oriented seismic profiles in a timely manner for use by a visualisation application. We compare this to the performance of enhanced GIS-based systems and suggest that a fully integrated architecture is more appropriate interacting with enterprise-scale data volumes.