John T. Murphy
Advances in High-Performance Computing (HPC) offer Computational Social Science the opportunity to create simulations at scales vastly larger than in previous times. Here an example is presented showing solutions to practical problems of modeling at these scales. The exercise explores a novel variation on a well-known model (called the ‘Triangles’ simulation), implemented in the Repast HPC toolkit and run at scales up to billions of agents. The results show that model dynamics change when proportions of different kinds of agents are varied, and this is consistent at very large scales. Practical issues of performance, network analysis, file output and data visualization are illustrated with this worked example.