GPU implementation of epidemiological behaviour in large social networks
Sri Sathya Sai Institute of Higher Learning, Prashanthi Nilayam, India
IEEE International Conference on High Performance Computing (HiPC), 2012
@article{menon2012gpu,
title={GPU implementation of epidemiological behaviour in large social networks},
author={Menon, S.K.M. and Baruah, PK and Sosic, M.},
year={2012}
}
In a social network, epidemic spread could be a spread of an infection, opinions, trends, fads, diseases or worm propagation in network. Epidemic spread computation on such huge and ever growing social networks is incredibly challenging. High-performance computing using GPUs has become an important tool to solve computationally intensive problems. This paper presents a GPU based implementation(GPU OPT) of Susceptible-InfectedRecovered (SIR) model. GPU OPT performs 1.8x-3.9x faster than an existing CUDA SIR implementation across various types of networks studied. CUDA SIR is 10x faster than FastSIR(a single core CPU implementation) in the worst and so GPU OPT is effectively about 30x faster when compared to FastSIR on an average case. This implementation was tested on social networks of varied types like Condense Matter Physics collaboration network, friendship network, who-trust-whom relationship network and a Email communication network.
December 10, 2012 by hgpu
