Speeding up Mobile Delay-Tolerant-networks simulations with CUDA-based event generator

Choosing a suitable network interface for Mobile Delay-Tolerant-Network (DTN) is a challenging task. The tested network interface must cope various density variation and analyzed in long time period. Long period network interface simulation in various environment require very long time to finish. In this study, we proposed Parallel-based event generator to shorten Mobile DTN simulation. The generator is constructed based on Compute Unified Device Architecture (CUDA). Half Cyclic Checking method is used to maximize CUDA capability by avoiding memory access collision. The result is CUDA-based generator outperform CPU-based application and speed up generation process in ten to thousand times faster. In low density network, inter meeting event is rarely occurred. Hence, direct feeding of external event into simulator is the best solution to shorten simulation process. Using external event feeding which is generated by CUDA-based generator, the simulation process can be shorten in more than 400 times faster compared to standard mobility-feeding simulation.