Simulation studies of viral advertisement diffusion on multi-GPU

Simulation has become an important method that is widely used in studying the propagation behaviors during the process of viral advertisement diffusion. With the increased computing and memory resources required for large-scale network processing, General Purpose Graphics Processing Units (GPGPUs) have been used in high performance computing platforms to accelerate simulation performance. In this paper, we show optimized simulation strategies of viral advertisement diffusion on a Multi-GPU system. Using our proposed simulation strategies, we examine the spread of viral advertisements over a realistic social network with different tolerance thresholds. We also investigate the effect of different initial nodes selection policies in maximizing the performance of advertisement diffusion. According to our simulation studies of viral advertisement diffusion, we can observe that the number of initial selected nodes is important to the diffusion behaviors. However, we also note that the initial selection policy plays a limited role in the final result of viral advertisement diffusion. Finally, we discuss improved viral advertising strategies that use mass marketing first to increase the willingness of accepting a product and apply viral marketing to facilitate the maximization of advertisement diffusion.