The prospect for parallel computing in the oil industry

Abstract

About 10% of all ‘supercomputers’ have been delivered to the oil industry. Its two big computational tasks are seismic data processing (to deduce underground geological structures from surface-based probing with elastic waves), and reservoir modeling (to simulate the flows within a producing field, in order to optimize the amount of hydrocarbons that can be recovered). Both tasks are extremely CPU- and, for seismic modeling, also I/O intensive. Trying to forecast the future is always chancy, and especially so for parallel computing. To stay on a somewhat safe ground, we will focus on the two main tasks just mentioned, and on the computational demands they entail. A change of direction towards large-scale parallelism is well on the way in the former area, but held back a few years in the latter (mainly by coding complexities).