Bringing together efficiency and effectiveness in distributed simulations: The experience with D-Mason