Today's complex systems require easy modeling methods as well as eecient analysis tools. A Petri net based approach called Thorns (Timed Hierarchical Object-Related Nets) has shown to be appropriate for these often contradictory goals. Due to the complex concepts of the language, a formal analysis of Thorn models is almost impossible and simulation is the only analysis means. A problem with the simulation of any kind of models is the considerable amount of time necessary for its execution, if large or complex systems have to be treated in a detailed manner. Parallel computers or workstation clusters currently provide a possible solution to this problem. In our current project, we develop and implement methods for the distributed simulation of Thorns. In this paper, we give a rough overview of the Thorn language and present the main concepts and ideas of the distributed simulator for Thorns. We especially focus on support of dynamic subnet instantiation provided by Thorns. The utilization of the hierarchical structure of a Thorn model to achieve a with respect to speedup promising mapping is described in more detail. Most of our considerations are also applicable to optimistic simulation of other modeling languages that support dynamic generation and deletion of submodels.