Abstract
As an inherent characteristic of any distributed system, the execution and performance of distributed virtual simulations totally rely on underlay communication infrastructure and resources. The performance of such simulations is directly restricted by the communication latencies between interdependent simulation components. The high level architecture (HLA) is a framework designed with the objective of organizing these simulations through management services. However, the framework is unaware of the communication delays caused by the network distances between communicating simulation parts. These delays can result from nonplanned initial deployment or dynamic simulation changes, requiring constant load balancing. Due to the importance of balancing distributed simulations, many approaches have been designed. In order to provide a balancing system aware of the dynamic communication changes, a delay-based redistribution scheme has been designed. The scheme successfully arranges the load, but it lacks precision due to communication delay oscillations. Therefore, extensions are proposed to modify the balancing algorithm in order to avoid unnecessary, mistaken load rearrangements. In the experimental results, the delay-based scheme has been able to reduce the simulation execution time when compared with the distributed balancing scheme, and the proposed extension has been capable of increasing the precision of the balancing.