MAHA: Migration-based Adaptive Heuristic Algorithm for Large-scale Network Simulations

The scalable wireless network simulation poses huge computation challenges as the execution time needed to perform the simulation can be prohibitively high. Parallel and distributed simulation (PADS) approaches have been proposed that use huge memory and high processing power of multiple execution units [i.e., logical processes (LPs)] to handle scalable simulations. Each LP is comprised of a set of simulation entities (SEs) that can interact local or remote SEs. However, the remote communication among SEs and synchronization management across LPs are two main issues related to PADS execution of large-scale simulations. A number of migration techniques have been used to mitigate the problem of high-end remote communication. The problem is that most of the existing migration strategies result in higher number of migrations that ultimately lead to higher computation overhead. In this paper, we propose a migration-based adaptive heuristic algorithm (MAHA). Considering the run-time dynamics of the wireless network simulations, MAHA provides dynamic partitioning of the simulation model to achieve better local communication ratio (LCR). In addition, an adaptive academic simulation cloud platform, namely A-SIM-Cumulus cloud, is deployed for scalable simulations. The MAHA is implemented on A-SIM-Cumulus Cloud and simulations are executed multiple times with different configurations and execution environments. The results with optimum LCR show that the proposed algorithm significantly reduces the number of migrations and achieves a good speedup in terms of parallel (i.e., both multi-core and distributed) execution