The Shell Oil Company, the U.S. subsidiary of the Royal Dutch/Shell Group, first made a study of parallel computers and their possible application to geophysical computing in 1969.

When it became apparent that single-chip microprocessors would make it economically feasible to build, maintain and deploy a reliable parallel computer, a research group was formed in 1981 to survey the technology and initiate or obtain a suitable computer system. In 1984 we became nCube's first customer with a 512-node nCube-1, which was retired only recently. We now have three nCube-2 systems, one in research (128 nodes) and two in production (256 and 32 nodes). The next generation of systems is under consideration.

The primary goal was to use parallel computers for selected geophysical application programs in the corporate setting. In this respect we have been successful, including requisite network and job support.

We have also built parallel versions of several petroleum reservoir simulators, but these have not been put into use. Some of the factors separating the two application classes will be discussed.

Parallel computers can be made to work in practice, but one must be careful about the job mix, expectations, and competition from simpler systems.

Postscript. The research group was disbanded in Sept. 1993, though the larger nCube system remains in use. IBM SP-1 and SP-2 systems have since been adopted for a variety of seismic applications.