Agent推与拉通信模式性能比较临界条件及推模式算法

Critical Term for Performances of Push Model and Pull Model in Agent Communication and Implementation of Push Model

寻求一种稳定、位置透明的移动Agent通信策略,目前仍然是一个热门的研究课题.将推和拉两种消息传递的基本模式进行了详细分析,给出了两种模式性能比较的临界条件.该理论结果对无线网络中快速移动设备之间的消息传递同样具有借鉴意义.最后在基于邮箱的通信算法基础上,引入区域划分的思想,成功实现了推模式通信算法,算法能有效地进行移动Agent透明通信,大大降低了消息由邮箱到达Agent的时延,并降低了网络通信负载.

In various situations, mobile agents at different hosts must cooperate with one another by exchanging information and making decisions collectively. Communication efficiency is one of the most important factors affecting the performance of the highly dynamic and large-scale mobile agents system. To improve the communication efficiency and find a reliable location-transparency communication infrastructure for mobile agents, push model and pull model are two basic models used in communication algorithms. A detailed analysis of the two models is given and a theoretical formula of their performance comparison is derived. The formula indicates that only when the information traffic between mobile agents is extremely high, the performance of the pull model outperforms the push model, otherwise, the performance of the push model outperforms the pull model. In addition, an efficient mailbox-based communication algorithm using the push model by means of network compartmentalization is implemented. Besides the great reduction of the communication delay, the algorithm alleviates the load of the network and can be used to communicate efficiently and transparently between mobile agents.