Parallel simulation by analysis of population dynamics in high performance environments.

The project aims to develop a simulator to study the demographic evolution of a population, reproducing the internal interactions between its individuals. Given the nature of these social processes, the methodology used is agent-based modeling. Agent-based models have been applied in several areas of economics, archaeology, psychology or sociology, obtaining interesting results.

The long-term objective is to build a tool that will help decision-making in family planning policies. Due to the inherent complexity of social processes, the project uses parallel programming to leverage the resources of a high-performance architecture to study large-scale populations. This is why in this project we collaborated with Lancaster University, which had already developed a first prototype of a microscopic simulator, and the Barcelona Supercomputing Center, which was interested in exploring the use of the Marenostrum supercomputer in the area of social sciences.

The specific objectives of the project are:

• To deepen in the models of demographic simulation of populations.
• Evaluate the flexibility and performance of the simulator, to understand the mechanisms that improve the tool and take advantage of the parallel architecture.
• Design and development of a graphic tool that allows users in the social environment to use the simulator, in order to be able to transparently use parallel resources.

In addition to C++, the standard MPI (Message Passing Interface) and the parallel and distributed simulation libraries μsik and libsynk have been used in the development of the project.