Abstraction and intent through an autonomics framework

Abstract

Autonomic control systems provide self-management capabilities to networks using closed-loop controllers. The Rainbow framework from Carnegie Mellon University is an example of such a capability that uses an ability to reason on and manipulate a formal model of the network architecture to decide what changes to make in response to the situation. Probes and gauges feed the reasoning capability. These probes and gauges can provide some situational awareness to both systems and human controllers, but at a low level of abstraction making it difficult to gain an understanding of the status of a large network of complex systems. We believe that a side effect of utilizing an autonomic framework is enhanced situational awareness at a higher level of abstraction. This paper describes work in progress to develop gauges for Rainbow that incorporate machine learning to allow for early recognition of situation changes. It also describes how the use of strategy selection not only allows the network to adapt, but also to inform situational awareness.