A Control Plane Architecture to Enhance Network Appliance Agility through Autonomic Functionality

Heterogeneous networking appliances are pervading every aspect of daily life, offering a diverse array of services and capabilities. This has led to increasingly complex systems on a micro and macro scale. The challenging nature of complexity remains a key area of research. Current research focuses upon two key 'planes'; the knowledge plane (applications controlling device behavior through for example, machine learning; a fundamental aspect of autonomic systems) and the control plane (inter-device communication mechanisms). We believe a self- tuning control architecture is required to bring these two planes together, so supporting future autonomic agility. The knowledge plane will discern the required learning and the control plane will control the learning dissemination through a novel, generic dissemination and negotiation control protocol. We have proposed and designed an autonomic self-tuning architecture, which includes a negotiable control protocol, as well as, support for a flexible number and type of algorithm overlays. The control plane enables key learning attributes to be made visible at the control plane. These attributes are used to negotiate and agree an apt learning payload. This paper provides details of a prototype showing how we can extend the existing networking infrastructure by using this architecture. It shows how the control protocol attributes and learning payload can be self- tuned by an appropriate algorithm, such as a cost benefit analysis algorithm, to allow a network device to self-tune and achieve our stated goals.