分布式实时仿真系统中基于PN的同步模型

Petri Nets Based Synchronization Model for Distributed Real-Time Simulation System

1 引言 基于不同的应用背景与假设,人们已对分布式实时仿真系统中任务间的同步问题进行了大量研究,其主要思想是将系统看成一个离散事件的集合,事件之间存在某种相互依赖的关系,执行时遵循一定的时序约定,在不破坏这种约定的前提下,尽可能提高系统的并行度.

t Synchronization is still an open problem in distributed real-time simulation systems.In such kind of systems, application tasks are allocated on distributed processors. Because of such factors as communication delay and task failure, the multiple source data a task needs form several other tasks could not reach the task at required times. That is, they are not synchronized. Generally, time prediction is used to postpone the sending time of some data to achieve the goal. But this method is not adaptive because the real communication delay could not be predicted accurately. In this paper, the authors propose a new synchronization model- EGM (Entity Group Model). EGM abstracts the application tasks in a distributed real-time system as entities and proxy and uses proxy to take the advantages of both the central strategies and the distributed strategies. We also give the method to describe EGM formally using Petri Nets.