Combining a shared-memory high performance computer and a heterogeneous cluster for the simulation of light interaction with human skin

When light interacts with human skin, a complex and involved process begins as the light is absorbed and propagated by cells, fibers and other microscopic materials. This interaction happens countless times each day and its accurate simulation is essential to biomedical and computer graphics applications. Simulating this interaction is computationally intensive, yet highly suitable to parallelization. This paper describes the use of both a shared-memory high performance computer and a heterogeneous cluster to accelerate these simulations. With a description of the parallel software used, we present results to show the performance gains from using such a hybrid approach.