Abstract: This paper introduces a new method for optimally provisioning and pricing differentiated services, that maximizes profit and maintains a small blocking probability. Resources are provisioned per Quality of Service (QoS) class over the long-term (service level agreement duration), then priced based on user demand over the short-term. Unique to this method is the ability to dynamically promote traffic from one QoS class to a higher QoS class, based on estimated demand statistics. This additional flexibility encourages better short-term utilization of the classes, resulting in higher profits while maintaining a low blocking probability. Experimental results will demonstrate QoS class promotion can obtain higher profits, as compared to other provisioning and allocation methods.
Errin W. Fulp, Douglas S. Reeves