Scaling the Distributed Stochastic Simulation to Exaflop Supercomputers

The Monte-Carlo method (stochastic simulation) is the one of the major tools in statistical physics, complex systems science and many other fields and is considered to be the promising computational scheme to run on nearest future exaflop supercomputers with many thousands and even millions of computational cores. We suggest a technique of the distributed stochastic simulation suitable for running on large amount of computational cores of the supercomputer. An example of the highly scalable application utilizing distributed stochastic simulation on up-to-date tera- and petaflop supercomputers is the program library PARMONC. Thorough examination of the proposed technique was done using simulation model that is based on the multiagent simulation system AGNES. The AGNES in particular enables one to evaluate the performance of the supposed exaflop supercomputer loaded with the distributed stochastic simulation.