Efficient Multi-Channel Simulation of Wireless Communications

Abstract

Simulation is a key tool for studying new system designs, but its scalability is often limited by the complexity of underlying models. We investigate to what degree different channel models – in particular differently-complex signal representations and loss models – impact simulation performance. Measurements reveal that, if all effects relevant to typical vehicular network simulations are to be captured, use of a highly efficient signal representation that can exploit modern CPU features allows to cut its performance impact by an order of magnitude. Yet, measurements also reveal that in typical vehicular network simulations, runtime performance is dominated by that of loss modeling instead. To address this issue, we also present a universal approach that can speed up loss modeling. We show that this approach can improve the overall runtime performance of simulations by more than an order of magnitude with no loss in precision.