A method for determining an optimal configuration and redundancy allocation for a system containing a plurality of units grouped into a plurality of elements and one or more modules. The method conducts random statistical sampling of a combinatorial space reflecting the possible combinations, redundancies and integrations of the units and elements in a given system. For each sample combination, the method calculates an optimization metric that reflects the reliability and/or the cost of that combination. The optimization metric may incorporate relative weighting of constraints used to evaluate whether the combination is optimal. The optimum configuration will be the configuration having the lowest optimization metric out of all the samples.