In the present paper we are describing the development of a portable computational cluster infrastructure based on Apple’s MacMini systems. The objective of this effort is to initiate the exploration of the feasibility for building and evaluating a low cost computational cluster system that could be portable due to the small form factor of the individual units while it could still provide considerable computational scalability. Here we outline our experiences in the form of a how-to methodology for setting up a MAC OS X 10.4 cluster for Xgrid and three popular versions of Message Passing Interface (MPI), i.e., OpenMPI, LAM-MPI, and MPICH2. Subsequently, we describe some performance measurements through the utilization of the Fire Dynamic Simulator (FDS) for two cases of reactive flow topologies. Finally, Xgrid with OpenMPI throughput and latency performance results are presented and discussed.

This content is only available via PDF.
You do not currently have access to this content.