Price-coupled Scheduling for Differentiated Services: GC-mu vs. GPS

We present an integrated approach to pricing and scheduling for services that are differentiated in terms of throughput, delay and loss specifications. The key building block to the model are quality value curves that specify a user's value of higher quality levels. From the analysis emerges a pricing rule that charges based on rate and quality grade, and a dynamic scheduling rule, called the Gc-mu rule. The analysis also derives the economically optimal probabilistic QoS guarantee parameters. We compare our model to the deterministic approach of QoS guarantees using burstiness constraints and fair scheduling rules. The scheduling that arises from such deterministic approach is the well-known Generalized Processor Sharing (GPS). A comparative analysis inspires the fair Gc-mu-PS rule as the scheduling rule that combines the unique strengths of GPS and Gc-mu. This Gc-mu-PS rule is proposed as a tailored scheduling solution for both the Expedited Forwarding class and the four Assured Forwarding classes in the IETF's Differentiated Services.