Distributed agent management in a parallel simulation and analysis environment

Abstract

This thesis presents the design, implementation, and evaluation of adding features of automatic distributed injection, diffusion, merger, and termination to reactive agents in a parallel simulation and analysis environment. This provides a powerful simple-to-use agent behavior for information searching or diffusion purposes, improving programmability for the scientist simulating an experiment or analyzing big data. Performance evaluation shows an improvement in execution time over using reactive agents without these features. However, a solution not using agents at all currently shows better performance and an analysis is presented as to possible future improvements.