In July 1999, an IBM RS/6000 SP was installed at the North Carolina Supercomputing Center. Our fledgling MPI code was up and running smoothly by the end of a training class taught by the Advanced Computing Technology Center (ACTC) of IBM Research. With essentially one key MPI call, our code runs at better than 150 Mflops/processor, giving us 18 Gflops on 120 processors for this disk problem.