Measurement-based optimal resource allocation for network services with pricing differentiation

Michael G. Kallitsis, George Michailidis, Michael Devetsikiotis

Abstract: In this paper, we introduce a model for allocating available resources in service-oriented network, with particular focus on delay sensitive services. The model is based on a pricing scheme for the offered services and also takes into consideration the quality of service requirements of each service class through a probabilistic delay-bound constraint. The proposed policy is dynamic in nature and relies on online measurements of the incoming traffic for adjusting the class allocations. We illustrate its performance and its robustness to various tuning parameters through an extensive simulation study that considers various simulation scenarios including experiments based on real network traces.