Centralized Vs Distributed architectures in Cognitive Radio networks

Secondary users may not have all the necessary information about the cognitive network. As a result, license-exempt systems have to made a tough choice about the architecture to be adopted either a centralized or a distributed control of spectral and temporal resources depending on the pros and cons of each type.

In the centralized model, a set of stations need to be more robust and should incorporate advanced capabilities to sustain the remaining nodes. Backup operations are less complex since only central units need to be replicated. Less control is required given the fact that maintenance issues are principally located at the central station side. All these reasons highlight the cost effectiveness of the centralized model. The major concern with the centralized approach is that the failure of the central units could compromise the proper functioning of the system, although, right backup and storage solutions may overcome such issues.

The strong advantage of the distributed architecture is that each node may act selfishly and can endure various network failures independently. The network contains no weak points in this situation. However, each node must handle backup and control operations locally. The adhoc approach requires additional hardware resources to be implemented at the node level which is very costly, in practice, in terms of deployment.

I think that the centralized infrastructure can be more efficient than the adhoc one which is inherently limited in its knowledge of the exclusive users of the resources. Spectrum incumbents are fearful of disruption and interference so a centralized control seems to match better their expectations and helps gaining stakeholders’ confidence.

I point out that the more the centralized nodes the less likelihood of interference. However, in practice, any deployment of cognitive radio must start with a small number of centralized nodes when the chances of interference are substantial but gradually as the network becomes larger more central units are deployed until the performance of the system stagnates, as the case of cellular networks dimensioning.