分布离散事件仿真的集中同步算法

摘要

分布离散事件仿真和一般的离散事件仿真不同，它没有公共的时钟和事件表，通过分布在多台主机上的仿真模型进行消息传递来实现现实世界中的事件的并发执行。同步算法是分布离散事件仿真中确保模型之间发生事件的因果关系正确性的机制。文中提出了一个集中同步算法，它采用建立模型之间的关联关系和仿真控制节点的方法，来对仿真模型之间的交互事件和仿真模型内部事件进行调度。其实现机制简单，避免了死锁的产生，并能实现事件的并发执行。文中最后对在WindowsNT环境下实现该算法进行了探讨。%DDES is different from the traditional DEVS for its features ofno common clock and event list.It realizes the concurrent execution of the real world's events by transforming the messages of the simulation models which distributted in multi-hosts。The synchronization algorithms are the mechanism of guaranteeing the rightness of the causality between the models of DDES.This paper put forwards a centralized synchronization algorithm which schedules the interaction events among the simulation models and inner events in models by establishing the association relationship and simulation control node among them.Its realization mechanism is simple and can realize the concurrent execution while avoiding the deadlock.At Last a discussion is how to realize the algorithm in the windowsNT environment.