Geographically-Distributed Databases: A Big Data Technology for Production Analysis in the Oil & Gas Industry
Résumé
The paper discusses some reported shortcomings of state-of-the-start IT technologies currently employed in the data management of Oil & Gas production operations. Most current IT infrastructures connect historian databases, production databases and application servers. This creates complex issues of data consistency between these systems. In the discussion, a particular focus is put on the geographically-distributed nature of the network which suffers from low-bandwidth limitations and un-reliabilities, e.g. due to satellite communication links.Taking the production engineers’ viewpoint, an example of production allocation using Data Validation and Reconciliation (DVR) serves to stress the malicious impacts of the described architecture. Production allocation represents one of the various monitoring and analysis tasks that are performed, on a daily basis, at the centralized level of data management systems. A quantitative study shows that the problem of mis-synchronization of databases is of great practical importance.We propose solutions to improve the robustness to communication outages. To improve data consistency across sites in a decentralized manner, the paper exposes the key concepts of distributed storage, message-based communication, and clustering. More generally, the paper proposes to shine a light on the potential relevance of several recent advances in the scientific field of “big-data” to the world of Oil & Gas upstream industry. These off-the-shelf technologies must be specifically tailored to geographically-distributed networks. The specificities are detailed, the necessary development work is outlined, and the potential qualitative benefits are estimated. A possible implementation is sketched.