An adaptive fault-tolerance scheme for distributed load balancing systems
Dan Liu, Robson E. De Grande, Azzedine Boukerche · Annual Simulation Symposium · 2015
Load balancing of distributed virtual simulations has been developing into a critical mechanism for enabling these simulations as their complexity grows to model more realistic scenarios. As the scale of these systems increases, they become more susceptible to load imbalances caused by the heterogeneity and non-dedication of resources and by their own simulation load oscillations. Due to its importance, many balancing systems have been designed for distributed simulations. Nevertheless, none of the previous systems consider the existence of failures in their own systems, which can partially hamper or completely interrupt their balancing capabilities. Therefore, a fault-tolerant mechanism is introduced for load balancing systems to keep some minimal services running properly or enable the recovery of components when faults unpredictably occur. The proposed solution employs election and grouping tools to reconfigure the balancing system dynamically. Experiments have been conducted in order to evaluate the benefit of the proposed fault-tolerant balancing system.