Prioritized Adaptive Max-Min Fair Residual Bandwidth Allocation for Software-Defined Data Center Networks

Abstract

As modern wireless access networks are moving towards packet based wireless access one may expect mobile cellular telephony to be eventually replaced by voice-over-IP (VoIP) applications. The choice of the codec in these applications is not straightforward as packet-based power-aware wireless communications bring new factors into the play. We study interdependencies between the bitrate, energy consumption, and the perceived quality provided by the voice codecs. We show that it is sufficient to equip a software with three codecs only. These are G.729.E, G.711.1 and G.723.1 codecs. Among those, G.723.E provides the best trade-off between the involved factors. When the system is overloaded and/or the power consumption is the most important metric (i.e., a mobile is running out of power) G.723.1 provides the best possible capacity and energy savings at the expense of significant quality degradation. Finally, when the system is underloaded while the amount of power spent for running the service is not important G.711.1 provides the best possible heard quality at exceptionally high power consumption.