why dont you try to simulate a smaller portion of the domain since the problem seems to me as symmetric. Also if you are interested in meshing the second geometry using ICEM, try using the global mesh settings initially and check the mesh quality(determinant, aspect ratio, skewness), and if it meets your expectation, then proceed with the simulation. nThe first geometry would be relatively easier since the roughness has a rectangular shape, you will have to split the blocks and the meshing procedure would be similar to the one used in microchannels.nn