Seismic imaging aims at providing accurate and detailed 3D maps of the earth subsurface from acoustic wave propagation recording at the surface. Today's seismic imaging algorithms are rather crude compared to what this industry would do if given access to Exascale systems. CGGVeritas applications portfolio and geophysical research activities provide samples of seismic imaging codes representative of current and future trends of the oil and gas industry. From a programming perspective, those codes are both very compute and I/O intensive, but are most of the times featuring enough intrinsic parallelism to make them good candidates for large scale parallelisation.
The DEEP Architecture will allow us to study the level of scalability of our key applications by re-engineering the codes to take advantage of the vector capabilities of the Booster Nodes, while relying on the Cluster Nodes to address the lack of locality of these algorithms by exchanging data during the computations.
These simulations are done by CGGVeritas, France.
"The tremendous amount of data now being processed to obtain ever more accurate images of the earth's subsurface is pushing the limits of today's hardware. Fortunately, seismic imaging algorithms contain a large amount of parallelism, which is a perfect fit for the DEEP project's Booster nodes. Our initial tests have shown that our seismic imaging application performs very well on such nodes and can leverage their considerable computing capabilities given their restricted size and power envelope. At the same time, pre and post processing steps and access to storage can be more efficiently handled on high single-thread performance processors. The fact that the DEEP architecture tightly integrates these two kinds of nodes is of great interest to CGG." - Jean-Yves Blanc, CGG