Distributed Simulations

Tyler F. Cloutier

Molecular and biological systems are often extremely complex. The structure and function of many proteins encoded by the human genome are not yet understood or known. It requires an enormous effort to identify the biological pathways in which a protein or macromolecule may be involved, and even more still to fully understand the method through with it functions. Much of the mechanism of protein interactions must be inferred from experimental evidence. The importance of molecular simulations is in its potential to offer much detailed information regarding how macromolecules interact to perform there specified function. The promise of large simulations is that this can be done quickly and efficiently even when dealing with large systems.

Parallel computation has become increasingly important as the clock rate of processors as approached what seems to be an upper limit. Fortunately, molecular simulation is an very parallel problem.

This semester I explored the performance and challenges of distributed computing by implementing a distributed water molecular dynamics simulation. The goal of this project was to develop and optimize a distributed simulation system that could be easily generalized for molecular simulation with different models.

My presentation and detailed performance results of my project can be found here. Below you can see a video of a simulation of 2048 water molecules in a box with periodic boundary conditions.