Markov Renewal Process (MRP) is an extremely powerful analyzing tool for concurrent systems. But it is often difficult to determine the correct Markov model for even moderately complex systems. Petri net is very descriptive and flexible in modeling various kinds of systems of concurrency and asynchronization. This paper presents an aggregate approach of Markov Renewal Process and Extended Stochastic Petri Net (ESPN) model, which allows incorporation of arbitrary time distributions. Using a descriptive "language" like the ESPN for specifying a Markov model of a system, the analyzing power of MRP can be exploited more easily. Moreover, an Abstract Partial Reachability Graph (APRG) is introduced to simplify the Markov solution. In addition, the aggregate approach of MRP and ESPN is applied to analyze a parallel message system.