Communication strategies in distributed agent-based simulations: The experience with D-Mason

Abstract

Agent-Based simulation Models (ABMs) are a very powerful experimental tool of analysis, used in many scientific and technological communities of researchers, to assess and predict the dynamic unfolding of a series of events or processes, according to the imposition of certain conditions, given by the analyst. The computing power usually represents a limit for such simulations and the traditional answer to the need for computing power is to invest in computer resources. D-Mason is a framework for parallelizing simulations developed on top of Mason toolkit. The goal of D-Mason is to exploit wasted computing power in a network of computers, eventually heterogeneous, as a research lab or a cluster of workstation. In this paper we present a novel communication strategy using Publish/ Subscribe paradigm through a layer based on the MPI Standard. © 2014 Springer-Verlag Berlin Heidelberg.