Monitoring Autonomic Networks through Signatures of Emergence

This paper addresses the problems of delivering autonomic management of large-scale networks. It encompasses both the governance of networks of autonomic components and the autonomic governance of networks and indeed the provision of the latter by the former. For this it is necessary to consider the complexity of the systems involved and the mastering of this complexity by distributed self-^* functions. The complexity arises as a natural result of the engineered robustness; as with all autonomic systems the components added to provide self-^*operations also add to the complexity. In addition the feedback control loops within larger scale systems will interact causing emergent outcome to the system as a whole and to individual self-^* functions. This often means that the system is robust to large environmental perturbations yet remains vulnerable to cascading failures initiated by small perturbations. This is investigated through a formally specified observer system where novel outcome can be grounded to a series of actions and likely outcome reasoned upon. This further demands a range of metrics over which reasoning needs to take place: In this paper the algebraic connectivity of the (autonomic) network (of networks) is considered and a implementation presented based on autonomic monitoring selection by self-organisation characterisation. This addresses many current in establishing models of future computation such as the Internet of Services or Cloud Computing.