Reasoning Multiagent System for Distributed System Management*

Due to the proliferation of data networks, applications and services, distributed system management have to face the complexity, heterogeneity, uncertainty and dynamic changes in the managed domain. Multiagent technology provides an appropriate method to deal with the full distributed system because of its autonomy and cooperation. On the other side, the probability reasoning is needed for the system management in uncertain environment. In this paper, we consider the collaboration and probability inference between agents for distributed system management. We provide an algorithm of counter inference from effects to causes in Bayesian networks based system management. This is the foundation for further intelligent decision of management system.