high performance computing on graphics processing units: hgpu.org

For some years now, there has been an increasing interest in modeling and analyzing the spread of epidemics in both human and computer networks. The obvious advantage a computer simulation of the epidemic spread offers is that the answer is delivered in short time and the number of hosts included in simulation can approach their real-world number. This paper presents a CUDA (Compute Unified Device Architecture) technology based implementation of the simulation algorithm for modeling of the epidemic spread on a network. Spreading of the epidemics over the network is modeled using discrete SIR (Susceptible – Infected – Recovered) model. This implementation offers selection of a starting node and monitoring of the epidemic spread in each cycle. Compared to a common CPU implementation, the CUDA version achieves about 10x faster execution time in the worst case. That speed up is of great significance when running tests on large networks. The implementation was tested on real social networks consisting of more than 5 million nodes. Hence, we believe it can be of a practical value in analysis of the epidemic spreading over large networks. To the best of our knowledge, this is only implementation of SIR model on CUDA.