Pricing for heterogeneous services at a discriminatory processor sharing queue

Abstract

In order to deal with applications with different quality of service requirements, service differentiation has to be implemented, especially in case of congestion. Different scheduling policies can be applied at a queue, such as strict priorities, generalized processor sharing, or discriminatory processor sharing. While prices optimizing the network revenue have been determined in the first two above cases, and the optimal revenue compared, nothing had been done yet on discriminatory processor sharing (DPS). Though, at the session level, processor sharing is known to properly model TCP flows behavior. DPS then models multiple TCP flows at a router providing differentiated services. We study here what pricing induces on a DPS router when two types of application compete for service, what is the resulting equilibrium, and explain how optimal prices can be found. © IFIP International Federation for Information Processing 2005.