Parallel Reservoir Simulation on Clusters of PCs

by
Takaaki Uetani, MSE

University of Texas at Austin, 2001
Supervisor: Kamy Sepehrnoori

Numerical simulation of large complex petroleum reservoirs is computationally expensive and time-consuming process. Traditionally, such simulations were performed using powerful workstations, mainframes, and in some cases supercomputers. More recently, some developers have addressed developments of reservoir simulators for massively parallel supercomputers. However, due to high cost, usage of these supercomputers have been mainly limited to the users in major oil companies and national laboratories. But, developments of several hardware and software have enabled application of such parallel reservoir simulators affordable using clusters of PCs.

This report evaluates the performance of various Intel Architecture-based PC clusters with different networks using a parallel compositional reservoir simulator with up to 1.5 million gridblocks for a gas injection process. Results indicate that fast processors are essential for high-performance computing and also high-speed networks contribute effectively to reduce the wall-clock time of the simulation on clusters of PCs.

Back to theses index