Experimental Analysis of Optimistic Synchronization Algorithms for Parallel Simulation of Reaction-Diffusion Systems
Bing Wang,Yiping Yao,Yuliang Zhao,Bonan Hou,Shaoliang Peng
DOI: https://doi.org/10.1109/HiBi.2009.22
2009-01-01
Abstract:With the increasing demands for large-scale and fine-resolution models, simulations of the reaction-diffusion systems are becoming more and more time consuming. Combined with the Stochastic Simulation Method (SSA), the Parallel Discrete-Event Simulation (PDES) is a promising approach to utilize the parallelism in these models. Since synchronization algorithms play the key role in PDES, in this paper, we experimentally investigate the performance and scalability of optimistic synchronization algorithms in simulations of reaction-diffusion systems. First, the Abstract Next Subvolume Method(ANSM), a variant of the Next Subvolume Method (NSM), is presented. It integrates the logical process (LP) based modeling paradigm with several simulation algorithms including both sequential and parallel execution. Second, based on ANSM, three optimistic synchronization algorithms, including a pure optimistic approach, an optimistic approach with risk-free message sending,and a hybrid approach combined the above two are respectively plugged into the simulation. Third, a group of experiments are conducted to study the characteristics of the synchronization algorithms in the parallel simulation of a typical reaction-diffusion systems. The results show that comparing with the pure optimistic approaches, moderate optimistic approaches are more suitable for the stochastic simulation of reaction-diffusion systems, with respect to both the performance and the scalability.