STEAM-Sim: filling the gap between time accuracy and scalability in simulation of wireless sensor networks

After a decade of research in the field of wireless sensor networks the energy consumption remains the dominating constraint. Complex algorithms with non-negligible runtimes must be processed by resource-limited nodes, and therefore require in-depth knowledge of the temporal behavior of the software and hardware components. However, state-of-the-art simulators provide either accuracy or scalability and therefore somehow limit the development of such networks. We present a novel and unique methodology for energy-aware, time-accurate, and scalable simulation of wireless sensor networks that considers software, hardware, and network components. Algorithms implemented in C are annotated with binary runtime information and are executed natively on the host cpu, i.e., the cpu where the simulation is run. Arbitrary hardware can be modeled at various levels of abstraction and is simulated together with the software. Important effects such as interrupt processing are simulated accurately. As a proof of concept we implemented the proposed methodology and present STEAM-Sim, a novel simulation environment. We evaluated STEAM-Sim by means of a proprietary networking scenario typically used in industrial wireless sensor networks. Preliminary results regarding scalability and accuracy are presented.