摘要
仿真模型越来越复杂,受单机计算能力和存储容量的限制,模拟需要花费的时间也越来越长。PDES(Parallel Discrete Event Simulation)策略能够加快仿真程序的执行,因此一度成为研究热点。但是,并行仿真最终并没有在工业界得到广泛应用,其原因在于:并行仿真建模理论缺乏,并行仿真性能具有不可预测性,以及并行程序行为的不可预测性。本文在讨论模拟器并行化的一般方法基础上,给出了一个基于SSF的传感器网络并行仿真环境SensorSSF。SensorSSF设计遵循:可扩展性和简洁性。可扩展性保证CPU执行时间随求解问题的规模和仿真模型的复杂度线性增长;简洁性使得仿真应用人员无需了解太多并行程序设计知识,就可以编写出高效的仿真程序。实验结果表明,SensorSSF具有良好的可扩展性,同NS2相比具有较好的时间特性。
As the system to be simulated becomes more and more complex, the execution time may be unacceptable due to computational and memory constrain of single processor. PDES（Parallel Discrete Event Simulation） can accelerate the execution of simulation program, thus gained great interest. At the same time, PDES did not become popular in industry. The reasons include at least three aspects： lack of parallel simulation modeling theory,the unpredictable simulation performance of PDES and unpredictable behavior. After discussion general approach of simulator parallelization, this paper gives a detailed description of SensorSSF： a parallel simulator for wireless sensor networks. From the very beginning, designing of SensorSSF obeys two rules： scalability and simplicity. The scalability provides liner CPU performance as the problem size grows; Simplicity relaxes the programmer from knowledge of parallel programming while highly efficiency. The experiment shows that SensorSSF has a better performance than NS at some simulation application and can gain moderate parallelism in some scenery.
出处
《计算机科学》
CSCD
北大核心
2007年第7期58-62,共5页
Computer Science
