Online Scheduling for Resource Allocation of Differentiated Services: Optimal Settings and Sensitivity Analysis

In this paper we investigate in detail the properties of a dynamic resource allocation scheme that utilizes online measurements to optimally adjust scheduling weights and to achieve the required QoS, under a given pricing structure. Extending our previous work, we consider here an additional important QoS parameter, namely, the average queueing delay, in order to provision favorable QoS guarantees for all classes of traffic, especially the delay-sensitive one. The objective of this paper is to formally investigate optimal settings that guarantee an improved QoS performance, and develop fundamental insights based on a detailed, case-by-case mathematic model that takes into account all the relevant QoS parameter constraints.