Performance evaluation of discrete event systems via stepwise reduction and approximation of stochastic Petri nets

The authors present a stepwise reduction and approximation method for generalized stochastic Petri nets (GSPNs) in order to reduce their state space. When a subnet is reduced to a simpler structure, not only the qualitative properties but also quantitative characteristics such as token flow rates are preserved. The authors first define various kinds of potentially reducible subnets and then present rules on subnet selection, approximation subnet construction, and reduction evaluation. Two criteria for judging a reduction step are the number of states in the subnets and the final net during the reduction process and the error resulting whenever the exact value is possible. A discrete event system which models a computer system is used as an example to illustrate the procedure