Implementing Exhaustive Search for the Coil-in-the-box Problem using MPI

This paper addresses parallelization of the exhaustive search for solving the coil-in-the-box problem. Since exhaustively solving the problem takes roughly 20 days, compairing different techniques to prune the search tree is unrealistic. However, every subtree of the search tree can be treated individually, an ideal setup for parallelization techniques. We present a simple client server architecture using MPI and analyze experimental data with respect to overhead; including work towards finding start parameters that minimize overhead.