Finite element analysis via parallel multi-level frontal solver is studied based upon the MPI (Message Passing Interface). The speedup of the parallel solver versus the number of the cluster nodes is examined for each of the main memory and the hard disk options for data storage and for various levels of multi-frontal analysis. The numerical examples show that the present parallel scheme approaches the ideal speedup for a small number of processes, but yields relatively poor improvement in parallel performance for a large number of processes.