Optimization for Accelerating Large Scale Agent Based Simulation
Parallel agent based simulation is popular used in artificial society. However, it brings great challenges to the execution efficiency when facing large scale artificial society where the number of agents in the simulation is up to millions. A simulation kernel with conservative synchronization strategy and multi-thread scheduling paradigm for large scale parallel agent based simulation is introduced. Based on the simulation kernel, the paper proposes two optimization strategies: a container based agent management scheme and an event based load balance strategy. The paper then design several experiments to evaluate the optimization performance, it shows that the optimization strategies can obtain up to 5x speedup compared to the basic simulation kernel.
KeywordsAgent based simulation Parallel discrete event simulation Synchronization strategy Load balance
This work was supported by the Natural Science Foundation of China (Grant No. 9102403071303252, 61403402, 41201544 and 71343282) and the HPC project (13010502) funded by NUDT.
- 2.Stroud,P.: Spatial dynamics of pandemic influenza in a massive artificial society. JASSS 10(49) 2007Google Scholar
- 3.Fujimoto, R.M.: Parallel and Distributed Simulation Systems. Wiley, New York (2000)Google Scholar
- 5.Wang, Y., Lees, M., Cai, W., Zhou, S., Low, M.Y.H.: Cluster based partitioning for agent-based crowd simulations. In: Proceedings of the 2009 Winter Simulation Conference, pp. 1047–1058. IEEE Press, Austin (2009)Google Scholar
- 7.Carothers, C.D., Bauer, D., Pearce, S.: Ross a high-performance, low memory, modular time warp system. J. Parallel Distributed Computing 53–60 (2000) Google Scholar
- 9.Vulov, G., He, T., Hybinette, M.: Quantitative assessment of an agent-based simulation on a time warp executive. In: Proceedings of the 2008 Winter Simulation Conference, pp. 1068–1076. IEEE Press, Florida (2008)Google Scholar
- 10.Riley, P.F., Riley, G.F.: SPADES—a distributed agent simulation environment with software-in-the-loop execution. In: Proceedings of the 2003 Winter Simulation Conference. IEEE Press, New Orleans, pp. 817–825 (2003)Google Scholar
- 11.Collier, N.: Repast HPC Manual. Argonne National Laboratory, Argonne (2010)Google Scholar
- 12.Miller, R.J.: Optimistic parallel discrete event simulation on a beowulf cluster of multi-core machines. In: Master dissertation, Cincinnati University, July 2010Google Scholar