IDEAL:A Multl-Agent Architecture for the Management of Complex Networks

摘要

Large telecommunication networks or internets bring out new challenges for management.The key issue is how to efficiently handle huge quantity of information which may come from heterogeneous networks.In this paper,we describe IDEAL, a multi-agent framework to manage such networks.This approach divides the network into domain areas,which can then be managed by separate but interlinking agents(i.e.,computer systems).Each agent has total jurisdiction over its own domain area,but if needed,it can partially take in charge the management of its neighbors.We detail the common architecture of IDEAL agents which consists of five components corresponding to the different activities of an agent.The control component is in charge of coordinating the other components.The communication component allows the agent to communicate with the other agents by message passing.The teamwork component provides some cooperative mechanisms to allow the agent to cooperate with the other agents.The expertise component contains knowledge and reasoning specific to itsdomain area and skills.The presentation component extracts relevant information to allow a developer of an IDEAL application to analyse the agent'evolution.Then,we focus on diagnosis agents to present different types of cooperative mechanisms that can take place during their diagnosis process.To that end,we follow the processing of an alarm set which gives us a global view of cooperation.Furthermore,we analyse this process from an agent perspective to have a local view of cooperation.