针对用于卫星导航系统定位、测速和授时性能分析与评估的仿真系统运行时间过长的缺陷,以全球格网精度因子分析为例,提出了利用局域网机群、采用并行仿真技术来提高仿真运行速度的解决方案。为了实现负载均衡,先后采用了格网数目平均分配、根据CPU主频分配格网实体、按照基准测试时间分配格网实体三种实体划分方法;为了减少进程间通信,设计了一种由逻辑进程的第一个格网实体接收来自星座实体的带有各卫星坐标的事件,并在格网实体之间依次转发此事件的事件发送机制。实验结果表明,采用根据CPU主频分配格网实体的并行仿真方法,可获得理想的加速比。
One drawback of the simulation evaluation of Satellite Navigation System's performance on positioning, velocity-measuring and time-telling is that it takes the analyzers too much time to run a simulation. A cluster-basedparallel simulation solution was proposed to solve this problem, using global analysis on dilution of precision as an example. For load balance, three entities partitioning methods were introduced one after the other, including mean grid number based partitioning, CPU's frequency based partitioning and baseline testing time based partitioning. For less communication among processes, an event sending mechanism was designed, i.e., the first grid on a logical process received the events (contain the coordinates of all satellites) sent by the constellation, and then retransmitted these events among grids from one to the other. The results of the experiments show that the parallel simulation using CPU's frequency based partitioning can gain good speedup.