I've noticed a revival in papers that use clustering as part of the network design of a vehciular ad-hoc network. To me it seems intuitively obvious, both from a security and a network management perspective, that performing clustering is an expensive and risky task with unclear gains. I'm wondering whether there is any work that addresses this more general question, rather than trying to show the feasibility of a specific clustering scheme for a specific application. More specifically, I'd like to see some discussion on these questions to either confirm or revise my intuitive standpoint:
- Is real-world traffic really stable enough to have clusters?
- How significant is the overhead needed for cluster management compared to a simple broadcast network?
- Is it really possible to design a clustering scheme that is secure against compromised cluster heads/leaders?
Dear Rens Wouter van der Heijden,
Look the links. May be useful.
http://ieeexplore.ieee.org/document/6256251/
http://thesai.org/Downloads/Volume6No9/Paper_38-Secure_Clustering_in_Vehicular_Ad_Hoc_Networks.pdf
Regards, Shafagat
That clustering survey compares different clustering algorithms. I'm somewhat familiar with those already -- I'm more interested in comparisons between clustering and non-clustering approaches. For most applications that I'm aware of, clustering is not necessary, potentially even detrimental (especially at cluster borders, if the spatial re-use is really used; similarly, it would be detrimental for security). Employing clustering without any gains, with its' inherent disadvantages, seems like a bad idea. Are the benefits for individual "edge case" applications (routing, perhaps also some forms of aggregation) really sufficient to warrant a complete re-work of the communication architecture? Or would clustering be deployed on top of existing applications in an overlay network style (and will that keep the benefits of clustering intact)?
- Badges
- Science topic