并行仿真动态迁移技术研究  

Research on Dynamic Transfer Technology of Parallel Simulation

作　　者：乔海泉[1] 周云[2] 朱恒伟[1] 陈玉波[1] 

机构地区：[1]第二炮兵装备研究院,北京100085 [2]国防科技大学机电工程与自动化学院,长沙410073

出　　处：《计算机仿真》2012年第8期109-111,115,共4页Computer Simulation

摘　　要：针对提高并行仿真效率问题,关于并行离散事件仿真,仿真克隆是提高效率的有效方法,但是存在克隆次数受仿真节点物理内存、CPU数目等有限资源制约的不足。提出将克隆的仿真迁移到其它空闲仿真节点来解决资源制约问题。首先给出了仿真迁移的方法,然后按照仿真的状态保存、状态迁移、状态恢复、状态修改的过程,阐述了利用仿真持久框架实现仿真克隆与迁移的具体步骤,最后通过测试数据验证了仿真动态迁移给仿真性能带来的提高。Cloning is an effective way of improving simulation efficiency in parallel discrete event simulation. But the shortcomingof cloning is that the times of cloning is limited by the resources of simulation nodes, such as physical memory, the number of CPUs. To solve this problem, we can transfer the cloned simulations to other free nodes. The simulation transfer method was introduced first, then the simulation transfer procedure was described in detail in four steps： state saving, state transfer, state recover and state changing, and the implement which utilizes the simulation persistent framework was also given. In the end, the test result shows that the simulation transfer can improve the simulation performance.

关 键 词：并行仿真 仿真克隆 动态迁移 状态保存 状态恢复 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]