Distributed simulation is divided into fine-grained and coarse-grained variants. Fine-grained approaches produce a lot of overhead since they need to synchronize the simulation components. Coarse-grained solutions avoid this by distributing whole simulations, which do not need a synchronization. This represents an efficient way to execute a set of simulation runs, especially if there are a lot of replications necessary, for example when stochastic factors influence the simulation runs. As Grid approaches deal with the distribution of tasks in a network, they provide ideas for this concept. To enable a coarse-grained experiment distribution for the simulation framework James II, a parallel execution concept is created and implemented, using ideas from the Grid. The concept is evaluated, using the Stochastic-$\pi$, Beta-binders and PdynDEVS formalism, to investigate the gain compared to sequential and fine-grained distributed simulations, as well as the behavior with memory expensive models.