Cyber-physical systems (CPS), by their very nature, mix continuous and discrete behavior and are modeled by heterogeneous components. Formal analysis cannot always handle such complex systems and simulation is a necessary step. In particular, distributed simulation is very useful for the validation of CPS for two main reasons: either the CPS itself is distributed (e.g., a fleet of UAVs) or the CPS is too complex and/or has too much models (e.g., an aircraft). We discuss in this paper the impact of distributing the simulation of a system: which are the rules that must be applied to guarantee a correct behavior between the different simulators? If a centralized simulation already exists (using Discrete Event simulation), which hypothesis must be made for the Distributed Discrete Event simulation? The co-simulation framework used and discussed in this work is Ptolemy-HLA. It allows a Ptolemy model to be distributed using the high-level architecture (HLA) standard.