Usage-based pricing of offered traffic to a data network can be an effective technique for congestion control. To gain insight into the benejits usage-based pricing offers, we propose and study a simple model in which many users wish to transmit packets to a singleserver queue. Based on the announced price per packet and the available Quality of Service (QoS) (e.g., mean delay), each user independently decides whether or not to transmit. Given statistical assumptions about the incoming trafic streams and the QoS as a function of offered load, the equilibrium relationship between price and QoS is determined by a jixed-point equation. The relationships among price, QoS, revenue, and server capacity are illustrated numerically, assuming a particular type of random user population. These examples indicate that adjusting the price to maximize revenue results in an efficient use of service capacity with an associated small mean delay.
Michael L. Honig, Kenneth Steiglitz