Understanding the molecular mechanism of mechanical response under a cyclic loading is of fundamental importance in designing high-performance elastomers with a long service life. Herein, we investigated the mechanical response and accompanying microstructural evolution of elastomers under a long-term cyclic loading by using coarse grained molecular dynamics simulation. The typical characteristics of the Mullins effect are successfully reproduced to confirm the validation of our simulation. Through a systematical analysis for the evolution of an intrinsic structure for a pure elastomer, we find that the Mullins effect results from chain extension and chain slippage at a low cross-linking density, whereas at a high cross-linking density, it is mainly caused by bond rupture. Particularly, the optimized cross-linking density of a pure system can lead to optimum mechanical strength and fatigue resistance because of much greater energy dissipation by polymer chains with high mobility. The filled elastomer indicates a much better fatigue resistance because of the incorporation of nanoparticles (NPs) compared to the pure system. However, there is a serious aggregation of NPs during the long-term cycle, which can be prohibited by enhancing the interfacial interaction between the NPs and the matrix to prolong the service life. Furthermore, more bond breakages take place in the system with a chemical interfacial interaction, whereas a pure chain slippage on the surface of the NPs is observed because of a physical interfacial interaction. Notably, most of the bond breakages occur on the chain backbone which undergoes the greatest local deformation. This interesting observation is consistent with the recent work by Creton and co-workers. Generally, this work is expected to provide some guidelines for preparing better fatigue-resistant elastomers by elucidating the Mullins effect and fatigue mechanism at the molecular level.