An efficient disk-based tool for solving large Markov models

Very large Markov models often result when modeling realistic computer systems and networks. We describe an efficient tool for solving general, large Markov models on a typical engineering workstation. It uses a disk to hold the state-transitionrate matrix (possibly compressed), a variant of block Gauss-Seidel as the iterative solution method, and an innovative implementation that involves two parallel processes communicating by shared memory. We demonstrate its use on two large, realistic performance models.