Pricing and Performance Analysis for a System with Differentiated Services and Customer Choice

We consider a model of a service system with finite and shared processing capacity and two service classes. Users arrive at the system and select either a high-priority service level where the service requests are processed at a fixed rate, or a low-priority service level where service rate is subject to degradation when the system is congested. A fixed price-per-connection is charged for each service level, and the mean delay in each class class is announced to the users upon arrival. The users, in turn, select the appropriate class of service based on their perceived “cost,” comprised of price and delay-related cost. We demonstrate that the optimal operational mode of this system is in “heavy-traffic” if the demand is elastic, and determine the asymptotically optimal price per service grade. In particular, the magnitude of price-premium for high-priority service is seen to be “small.” Finally, a somewhat surprising feature of the system is that the fraction of users that select each service-level is determined by a “second-order” analysis that hinges on underlying diffusion limits.