一种改进的移动Agent通信算法

如何实现远程Agent通信的位置透明性，保证消息不会因为目标Agent迁移而丢失，一直是移动Agent通信所面临的难题，在现有的很多移动Agent系统中都没有得到解决，作者在Mogent系统中提出的通信算法初步实现了通信的位置透明和可靠的消息传输，该文在原有处的基础上提出了一种改进的适于多种迁移和通信模式的移动Agent通信算法，进一步减少了Agent的地址注册开销和迁移受到的限制，并给出了一种避免地址欺骗攻击的解决方案。

An Improved Mobile Agent Communication Algorithm

Mobility makes things complicated in mobile agent communication. On one hand, a practical communication mechanism should make the location of an agent transparent to the agent programmer and thus poses a burden on the agent system to keep track of an agent. On the other hand, because of the asynchronous nature of message passing and agent migration, how to guarantee the message delivery to highly mobile agents is still an open question in mobile agent systems. In many currently available mobile agent systems these problems are left unsolved and the hard nuts are left to agent programmers. The communication mechanism that we proposed in Mogent system implemented location transparency and reliable message delivery. But it also has the shortcomings of large location updating overhead, constraints of agents' mobility and vulnerability to the address spoofing attack. In this paper we introduce an improved algorithm based on that of Mogent system. In our improved algorithm, each mobile agent has a Mailbox which buffers the messages sent to it. The Mailbox is detached from its owner agent in the sense that the agent and its mailbox can reside at different hosts. If an agent will not communicate with others at its target host, it will migrate to the host directly and leave its mailbox at the previously located host. In this way the location updating overhead is saved and the constraints of agents' mobility is decreased. An interlocking mechanism is also employed in our algorithm to avoid the address spoofing attack during the location updating process. The improved algorithm can work adaptively according to different migration and communication paradigms in different applications. As shown in the experiments, the algorithm is preferable in the cases that mobile agents migrate frequently but communicate rarely.