Within the seismic geophysics community, MATLAB^® is generally considered a useful tool for developing prototype algorithms, but not for scaling to production. It might be easy to accept this bias, given that the seismic data sets exceed terabytes. However, one can argue that measuring the efficiency of a programming language should include time to production deployment, not just CPU time. In this talk, Charles Jones and James Selvage demonstrate how, by keeping parallel computing and storage in mind, BG Group has used MATLAB to build a production framework for processing multiterabyte seismic data sets.

Based on lessons learned, BG Group is facilitating the formation of an International Inversion Initiative (III) between Imperial College London in the U.K., the University of British Columbia in Canada, and Universidade Federal do Rio Grande do Norte in Brazil. This initiative includes running MATLAB Distributed Computing Server™ on a 405 TFlop HPC system that is based at SENAI Cimatec in Salvador Bahia in Brazil.