[Abstract] Execution times of large-scale computational science and engineering parallel applications are usually longer than the mean-time-between-failures. For this reason, hardware failures must be tolerated by the ...