Paper detail

Game Theoretic Iterative Partitioning for Dynamic Load Balancing in Distributed Network Simulation

High fidelity simulation of large-sized complex networks can be realized on a distributed computing platform that leverages the combined resources of multiple processors or machines. In a discrete event driven simulation, the assignment of logical processes (LPs) to machines is a critical step that affects the computational and communication burden on the machines, which in turn affects the simulation execution time of the experiment. We study a network partitioning game wherein each node (LP) acts as a selfish player. We derive two local node-level cost frameworks which are feasible in the sense that the aggregate state information required to be exchanged between the machines is independent of the size of the simulated network model. For both cost frameworks, we prove the existence of stable Nash equilibria in pure strategies. Using iterative partition improvements, we propose game theoretic partitioning algorithms based on the two cost criteria and show that each descends in a global cost. To exploit the distributed nature of the system, the algorithm is distributed, with each node's decision based on its local information and on a few global quantities which can be communicated machine-to-machine. We demonstrate the performance of our partitioning algorithm on an optimistic discrete event driven simulation platform that models an actual parallel simulator.