This paper is dedicated to the modelling and the evaluation of discrete event systems. We focus on manufacturing systems. The first part of the paper concerns cyclic manufacturing systems. The model of such a system is an event graph which encapsulates both the physical and the decision making system models. We propose approaches which aim at maximising the productivity while minimising the Work-In-Process (WIP) in both the deterministic and the stochastic cases. The second part of this paper addresses acyclic manufacturing systems. The Petri net model of such a system is different from the Petri net model of a cyclic system in the sense that only the physical parr of the system and the related constraints are modelled using Petri nets. We show that such a model perfectly fits with a hierarchical management of manufacturing system.