论文部分内容阅读
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。
The simulation model is more and more complicated. Due to the limitation of computing power and storage capacity, it takes longer and longer to simulate. PDES (Parallel Discrete Event Simulation) strategy to accelerate the implementation of simulation programs, it has become a research hot spot. However, the parallel simulation has not been widely used in the industry in the end because of the lack of parallel simulation modeling theory, the unpredictability of parallel simulation performance and the unpredictability of parallel program behavior. Based on the general method of simulator parallelization, this paper presents an SSSS-based sensor network parallel simulation environment SensorSSF. SensorSSF design follows: scalability and simplicity. Scalability ensures that the CPU execution time increases linearly with the size of the problem solving and the complexity of the simulation model; Simplicity allows simulation applications to write efficient simulation programs without having to know too much about parallel programming knowledge. Experimental results show that SensorSSF has good scalability and better time performance than NS2.