The principal aim of cognitive networking is to equip traditional networks with some sort of intelligence, in order to make them evolve and achieve higher levels of performance than those that can currently be achieved . Typical characteristics of cognitive networks are the ability to monitor the environment they are deployed in, to take reasoned actions based on current conditions towards an end-to-end objective, and to learn from past experience. Such enhanced networks will likely be characterized by a nonnegligible complexity, which can be tolerated if accompanied by relevant benefits, such as performance increase or mitigation of management burden. Different cognitive proposals have been proposed thus far in the literature, ranging from general framework definitions , to cognitive node architectures , to specific implementations . This contribution aims to illustrate the concepts we advanced in previous works , in which we described how the cognitive networking paradigm can fill the gap between service-oriented architectures and network, and extend them, by enabling reasoning with external information, i.e. information that is not locally available and is sensed elsewhere in the network. This enables the cognitive process to achieve a global vision of the network, which should facilitate the achievement of better end-to-end performance.