实现适应性系统管理的分布式多智能体框架

Implementation of Distributed Multiagent Framework for Adaptive System Management

为了应对动态、多变环境给管理系统带来的复杂性,提出了基于分布式范型的多智能体框架.该框架由多个分布式智能体结点组成,结点间通过消息传递与会话机制来实现分布式的合作与协同.结点内的功能主体使用了基于信念机制的内核结构,使得每个结点在以目标驱动的方式展开各种诊断与恢复的同时,仍能以反应驱动的方式对环境的变化做出反应,从而高效地实现了本地适应性管理.通过一个网络资源管理实例的分析证明,在复杂、动态的环境下所提框架能有效地实现可扩展的适应性系统管理.

In order to deal with the management complexity resulting from the dynamic and heterogeneous environment, a multi-agent framework consisting of many distributed nodes is proposed based on distributed paradigm. By using the mechanism of message passing and conversation among nodes the distributed cooperation and coordination are realized. The kernel architecture based on faith mechanism is utilized by the functional main agent within the node so as to respond to the change of environment in react-driven way while carrying out various diagnoses and recovery in a goal-driven manner in every node, thus the local adaptive management is complemented efficiently. Through analyzing an instance of a network resource management, it is shown that the extendable adaptive system management can be realized by the proposed framework in dynamic and complex environment.