Rule Based Reasoning for Network Management

This paper focuses on improving network managementby the adoption of artificial intelligence techniques.We propose a distributed multi-agent architecture for networkmanagement, where a logical reasoner acts as a managing entitycapable of directing, coordinating, and triggering monitoring andmanagement actions in the proposed architecture. The logicalinference system has been devised to enable automated isolation,diagnosis, and to repair network anomalies, thus enhancingthe reliability, performance, and security of the network. Themeasurements of network events are captured by programmablesensors deployed on the network devices and are collected by thenetwork management entity where they are merged with generaldomain knowledge, with a view to identifying the root causesof anomalies, and to decide on reparative actions. The relevantresults inferred by the logical reasoner and the significant eventsoccurred on the network are stored both in a global DB andin local distributed DBs, in order to enable successive analysesof network events. In order to illustrate the advantages andpotential benefits deriving from the reasoning capabilities of ourmanagement system, the results of preliminaries experiments areanalyzed.