We propose an optimal train dispatching module which computes in real-time the train routing and scheduling that minimize delay propagation. For solving this real-time railway traffic management problem, we exploit a fixed-speed mixed-integer linear programming (MILP) formulation (Pellegrini et al., 2014). To take into account the operational constraints, we impose here a maximum computation time to the MILP solver: the resulting algorithm is a heuristic which returns the optimal solution as soon as the search process is completed (the optimal solution is found and its optimality is proven) or returns the incumbent solution if the maximum computation time is reached (Pellegrini et al., 2013). The object of this work is to compare through simulation the solutions proposed by this formulation with the current dispatching rule. The scenarios tackled include both passenger and freight traffic at peack time. The experiments which we present have been produced in the context of the PREDIT project SIGIFret (Simulation d'une gestion innovante des circulations fret).