Net level aggregation using nonlinear optimization for the solution of hierarchical generalized stochastic Petri nets

An approach to the hierarchical evaluation of large generalized stochastic Petri net performance models of computer and communication systems is presented. In this method, complex submodels are replaced by substitute nets of lower complexity which bear the same delay time distribution of tokens. An optimum substitute net is found, using nonlinear optimization and matching the delay time distributions of the original subnet and the substitute net. The model reduction approach is demonstrated by means of an example.<
>