Comparative study of two congestion pricing schemes: Auction and tâtonnement

We evaluate two mechanisms for setting prices in a QoS-enhanced network with congestion-sensitive pricing and resource allocation based on users' willingness to pay. In the first model, the congestion-sensitive component of the price is calculated by tâtonnement, with the price adjusted gradually to drive down the user demand to the supply level of network resources. In the second model, each user submits multiple bids for different bandwidths, each bid expressing its willingness to pay a certain premium for the corresponding bandwidth during congestion. Simulations show that both approaches provide greater network availability, revenue, and perceived user-benefit than a congestion-independent policy. Both approaches have generally similar performance in terms of perceived user-benefit. The tâtonnement-based model obtains higher network revenue than the auction-based model. The auction-based model achieves higher network utilization at a given level of perceived user-benefit, but has higher implementation complexity and longer set-up delay for new connections. The proposed auction-based model serves more users than comparable auction-based schemes, and has less signaling overhead and greater certainty of service availability.