Parallel Processing Simulator for Separate Sensor of WSN Simulator with GPU

Abstract

Recently wireless sensor networks have been used as the technology that is actively grafted onto industries and daily living. Sensors should have built-in routing functions and basic sensing functions for the self-configuration of topologies. The number of sensors necessary for using them in an actual observation ranges from tens to hundreds or thousands. When theses sensors are wrongly placed in an observation region, they can quickly run out of batteries or be disconnected. These incidents may result in huge losses in terms of sensing data from numerous sensors and their costs. Therefore a number of simulators have been developed as tools for effective design and verification before the actual arrangement of sensors. While a number of simulators have been developed, simulation results can be fairly limited and the execution speed can be markedly slow depending on the function of each simulator. To improve the performance of existing simulators, this paper aimed to develop a parallel processing simulator for separate sensor (P2S3) that enables users to selectively use the GPU mode. It enables parallel and independent operations by matching GPU with many cores in order to resolve the slowdown of the execution speed when numerous sensor nodes are used for simulations. Also, P2S3 include the analyzed of sensor nodes with log data and visualization. The P2S3 supports the GPU mode in an environment that allows the operation of compute unified device architecture (CUDA), and performs the parallel simulation processing of multiple sensors using the mode within a short period of time.