Autonomic network management: some pragmatic considerations

Autonomic Network Management (ANM) has the goal of increasing reliability and performance while reducing management cost using various automated techniques. These range from agent-based approaches relying on explicit models and ontologies to emergent techniques relying on gossip protocols, swarming algorithms or other biologically inspired work. In this paper, we review the failures, growing pains and successes of earlier techniques for automated and adaptive network control and management, from the simple control loops in TCP and OSPF to the more complicated emergent behaviors of BGP dynamics and overlay routing. From these examples we extract several lessons relevant to ongoing research in autonomic network management.