Reading files from and writing them to disk is a quite slow process, but usually the total time spent with i/o operations is negligibly small as compared to the total execution time of the program. More important is the amount of data, which have to be stored. Due to the distributed data structures, the size of the finite element models can be increased (almost linearly) with the number of processors, if each of them has the same amount of private memory available. However, the generation of the FE models and the storage of the results to disk can become a problem.