Simulation of the water flood problems often leads to solving of large sparse systems of partial differential equations. For such systems, the numerical method is very CPU-time consuming. Therefore, the parallel simulation is essential for water flood study with satisfactory accuracy. In this paper, we present some experimental results of parallel numerical solutions done on Linux clusters. Our measurements indicate that Linux cluster can provide satisfactory power for parallel numerical solutions, especially for large problems. We also provide experimental results done on a SGI Origin2000 machine for comparison with Linux clusters.