Comparison of domain partitioning algorithms in the problem of direct flow simulation within rock samples at pore scale

The combination of computed tomography and computational experiment is an important and promising tool in the study of the properties of various materials. State-of-the-art tomographic methods allow to obtain three-dimensional image of materials with high resolution, which leads to a high dimension of discrete settings (10$^6$–10$^9$ numerical cells). Their analysis is not conceivable without application of parallel computing methods. In its turn, the efficiency of parallel simmulations with a large number of processors depends largely on the balanced distribution of the mesh across the processors. In this work, simulation of a single-phase fluid flow within pore space of a sandstone sample with voxel representation is used to compare the partitions obtained by various methods using parallel partitioning tools ParMETIS, Zoltan, and GridSpiderPar. Average time spent on interprocess exchange during one time step of the considered parallel simulation was compared when the grid was distributed over the cores in accordance with various partitions. The obtained results demonstrate advantages of some algorithms and reveal the criteria, crucial for the problem. As a numerical simulator DiMP-Hydro is used.