Dynamic pricing with admission control by market demand in QoS managed network

Abstract

In order to manage congestion problems and allocate network resources fairly in the Internet networks, many researchers have studied dynamic pricing over the last decade. However, most of their research results have been limited by their reliance on an oversimplified demand model, and are not adaptable to emerging multiple class environments such as the Diffserv network. For example, user utility is generally represented by a logarithmic form due to its simplicity in optimization model for pricing. Logarithmic utility, however, is related to unit elasticity demand which is not realistic for representing user demand in the actual Internet service market. We have extended this concept into a dynamic pricing scheme by generalizing a demand model with constant elasticity and applying it to the multi-class Diffserv network. A series of simulation experiments was conducted to compare the engineering and economic performance of the new dynamic pricing model to the static pricing model. Simulation results show that the proposed dynamic pricing model is superior to static pricing for engineering efficiency using the variables of packet loss and link utilization. However, economic efficiency is dependent on whether QoS disutility factors, such as loss rate or queuing delay, are considered or not. Without regarding these factors, some static pricing results have shown better result than that of dynamic pricing. However, the more significant disutility factors, the better the economic efficiency of dynamic pricing. The business implication of these results is that the more the customer is aware of the importance of disutility factors the more the network provider will incur pressure to introduce dynamic pricing. We have also proposed a compatible economic admission control scheme to supplement the dynamic pricing model in order to manage spontaneous congestion. In real situations dynamic pricing may be only effective to control the market demand over mid-or long-t...