Cognitive Networks: Towards Self-Aware Networks

4.4. Self-Management Intelligence: To Know and to Act

In this section we examine how knowledge can be acquired and represented in self-managing networks to facilitate the understanding of the environment and the autonomic components. Then we will study the theories that drive the decision-making process of autonomic components and their resulting interactions with the environment and each other.

4.4.1. Knowledge

An autonomic component can only monitor a small subset of environmental variables in its local environment at any given time, yet has to make decisions that often have system-wide consequences (e.g. local QoS routing decisions impact the overall end-to-end QoS of user traffic). One way to combat this lack of knowledge is to correlate the monitored data among other autonomic components. This is nontrivial since the component must be aware of what information the others have, how the information is represented and whether it can be accessed in a timely fashion. Intercomponent negotiations and interactions must then take place to solicit the required data. The amount of network and system resource consumed to acquire such data remotely may also be a problem. QMON ^[] attempts to address this issue by offering differentiated monitoring classes, hence prioritized monitor traffic based on the QoS class of the service an autonomic entity is supporting. Alternatively, some decision-making algorithms are designed to deal with imperfect information or localized data, such as the ...