The breakdown of simulation nodes or shortage of simulation resources cause failure of distributed simulation system and lower down the reliability of simulation system. To improve fault tolerance performance and decrease the overhead of fault tolerance, a fault tolerance method based on virtualization technology was proposed. According to different failure locations, different fault tolerance methods were adopted. The optimization of checkpoint strategy was analyzed and the optimal interval was obtained. Including selection of nodes, the number of copies and the distribution of the copies, three main problems of backup strategy were concluded. The problems were solved through virtualization technology. Fault tolerance strategy based on virtual machine migration was proposed as a complementary of checkpoint strategy and backup strategy to decrease the overhead. The performance of dynamic fault tolerance strategy and normal fault tolerance strategy were evaluated through experiments. The experimental results prove that the proposed fault tolerance trategy is efficient and the overhead is kept at a low level.