Expanding the CASEsim Framework to Facilitate Load Balancing of Social Network Simulations

This research has two components, both involving the CASEsim framework. First, we must handle a few practical concerns, such as increasing the framework’s usability and implementing the ability to distribute computation. To achieve this, we use XML configuration files for usability and the Java RMI library for distributed computing. Second, we will study load balancing techniques for 0-dimensional (non-spatial) social network simulations, creating heuristics that allow for general load balancing of such simulations in an optimal way. We will explore two approaches. The first is a centralized technique, examining macroscopic properties of the graph — such as agent clusters — in order to put related agents on the same slave. The second is an individual, slave-level technique that in which individual slaves dynamically pass agents to other slaves in order to equalize the workload. Finally, we will test our heuristics via a variety of simulations using the CASEsim framework.