Discrete-event simulation is a widely used technique for the study of systems which are too complex to be modeled with analytical methods. Usually parallelism becomes an effective tool for reducing the running times involved in the simulation of large scale systems. However the actual realization of an efficient parallel simulator is highly dependent on the particular features of the system being modeled. As a result, a number of alternative strategies for parallel simulation, the so called synchronization protocols, have been developed. None of them is the most efficient one for all kind of systems, or even different instances of a given system. It is then relevant to provide the designer of a parallel simulator with information relative to the factors affecting the performance of known protocols. We present an analysis of such factors in the context of the comparison of an optimistic synchronization protocol with alternative approaches which are suitable for the bulk-synchronous parallel model of computing. It is well-known that analytical analysis in this field is mathematically intractable due to the irregular nature of the workload. However, rather than resorting to benchmarks whose results are largely influenced by programming details, we devised a strategy to get quantitative results from an implementation independent and yet empirical framework.