Un Modèle décisionnel d'agent pour une supervision adaptative des systèmes distribués

The advent of physically distributed systems and the need to minimize the down-time of services and production processes require more efficient supervision systems. Traditionally centralized, the anytime supervision of such systems is challenged when communications between supervision and supervised systems become slow, disrupted or too costly. Distributing the diagnosis and repair process using a multi-agent architecture, we propose in this article a decision model for agents allowing the supervision system to automatically and dynamically adapt itself to the communications’ state.Preliminary results on a simulator for distributed systems show that our proposal leads to a relevant autonomous supervision in distributed systems where a short response time prevails over a limited re-