This paper studies a control strategy for reducing congestions in freeway systems by applying ramp metering as control measure. More specifically, the objective is to define a Model Predictive Control scheme to be applied on line, characterized by a finite-horizon optimal control problem with a mixed-integer linear form, in order to be efficiently solved with commercial solvers. In this optimal control problem the prediction model is obtained by linearizing the first-order macroscopic traffic model, hence binary variables must be introduced and the resulting model has a piecewise linear structure. In the paper, the adopted control scheme is first of all analysed in order to evaluate its effectiveness in improving the traffic conditions; secondly, the analysis has been devoted to evaluate the computational time necessary to solve the finitehorizon optimal control problem depending on the problem sizes and the traffic scenarios.