|Abstract:||A major problem with leveraging event-driven, packet-level simulation environments, such as ns2 , J-Sim , OpNet ), and QualNet ), in conducting wireless network simulation is the vast number of events generated, a majority of which are related to signal transmission. Due to the broadcast nature of a wireless channel, transmission of a signal has to be received and processed by all nodes operating on the same channel (and neighboring channels if co-channel interference is taken into account). This implies that one signal transmission event will trigger numerous signal arrival notification events.
In this paper, we investigate the operations of signal transmission in the various stages: signal propagation, signal interference, and interaction with the PHY/MAC layers, and identify where events can be reduced without impairing the accuracy. We observe that for each instance of signal transmission, a large number of signal arrival events are generated to notify nodes in the interference range of the signal. A majority of them are, however, redundant, as only nodes in the receiving state or the idle state but intend to transmit will be directly affected by the signal and need be informed. We thus propose to leverage the MAC/PHY state information, and devise (from the perspective of network simulation1) a reactive channel model (RCM) in which nodes explicitly register their interests in receiving certain events according to the MAC/PHY states they are in and the corresponding operations that should be performed. The simulation study indicates that RCM renders an order of magnitude of speed-up without compromising the accuracy of simulation results. The memory required in keeping the extra state information, on the other hand, is minimal. This, coupled with the fact that there is no need to re-design the channel model for each specific MAC layer, and the modification made in the MAC/PHY layer is quite modest (e.g., a few API changes), makes RCM a light-weight candidate mechanism for expediting wireless network simulation.