Speaking at the Petroleum Exploration Society of Great Britain’s Data Management Conference, Duncan Irving (Manchester University) presented a seminal pilot study using Teradata’s data warehouse engine to attack complex storage and retrieval issues associated with massive upstream oil and gas data—seismic in particular.
Irving’s study began with a problem set by Hydro, to automate 3D channel location. Experts from the UK’s National Computer Center figured the answer was to load subsets of the data to graphics processing units (GPU) for fast pattern matching. But the early work failed because of the lack of a robust storage infrastructure.
Irving was then approached by Teradata with a proposal for a data warehousing approach. Teradata’s technology is designed for very large data volumes and is used by Wal-Mart to store and analyze its real-time sales data (OITJ June 2006). The approach involved an earth-model-based data structure that could become the focal point of enterprise technical computing.
Seismic data is stored by voxel and referenced by hashing at various spatial scales for speed of retrieval à la Google Earth. The data warehouse has the potential to support reservoir and financial modeling, rolling in weather, production data and more.
Teradata uses a massively parallel ‘shared nothing’ database architecture with ‘Access Module Processors,’ physically distinct units of parallelism that have exclusive access to portions of the database. Seismic data is stored as a 10 byte header and a binary large object. Spatial data management locates traces in 3D space and more data (velocity) and methods can be attached to a trace. Rendering and compression middleware can be farmed out to other hardware—perhaps a a Microsoft Xbox! Irving is currently working on geospatial query and is looking for tie-ins with other vendors.
Irving concluded by noting that data warehousing has been around for a couple of decades—but the oil and gas vertical has yet to ‘get to grips with it.’ There are many spin-off benefits from enterprise-class data warehousing.
Version control means that nothing is ever overwritten—good for Sarbanes-Oxley compliance and workflow management. Data warehouses are amenable to fuzzy logic and neural net processing. The Manchester/Teradata solution supports high performance computing for flow modeling and real time monitoring of production measurements. Teradata’s Data Warehouse is used, albeit in a non technical role, by Repsol-YPF as a business intelligence system to analyze sales data.
© Oil IT Journal - all rights reserved.