Why VCG auctions can hardly be applied to the pricing of inter-domain and ad hoc networks

International audienceThe use of Vickrey-Clarke-Groves (VCG) auction mechanisms is gaining popularity in the networking community, where it seems compulsory to incentivize selfish nodes (in ad hoc networks) or domains (in inter-domain communications) to forward the traffic of their peers. Indeed, VCG auctions are known to both be efficient and produce proper incentives. In this note, we argue that, in fact, VCG auctions can hardly be applied to those problems, for different reasons depending on the model studied: if some resource constraints (bandwidth, spectrum, and/or power) have to be taken into account, then computing allocations and prices implies solving optimization problems that are computationally hard for general network topologies. If there are no such resource constraints, then VCG auctions, even if verifying many important and satisfactory properties, cannot verify a major one that is budget balance: the sum of subsidies given to relay nodes exceeds the sum of charges paid by traffic senders. This means that the auction regulator is required to continuously inject money to make the scheme work, which is unlikely to happen. In a second step, we discuss the combinations of properties that can be verified together, and prioritize them for finding out a proper pricing scheme