Voice quality framework for VoIP over WLANs

Voice over IP (VoIP) has quickly been adopted by commercial and private users due to its low cost of service. To attain an even greater consumer acceptance by offering the convenience of mobility, we have to use a wireless carrier to deliver VoIP packets at mobile phones. The low cost of service is the ultimate bargaining chip of VoIP and using a high cost cellular network, like GSM or CDMA, for the last mile coverage compromises its appeal. The IEEE 802.11 WLANs, in this regard, offer a low cost wireless coverage which intrigued many researchers and device manufacturers to investigate the voice performance in WLANs using theoretical analysis and simulation. Simulation based studies offer only restricted insight since the call capacity after changing any network aspect can not be determined easily and, therefore, analytical capacity models are more helpful to network designers and administrators. But a number of over-simplified assumptions used in the existing voice capacity models make these studies impractical. Standard voice quality measures were ignored and unrealistic assumptions were used in modeling the IEEE 802.11 medium access mechanisms, which make the results unreliable. In this thesis, we model the voice capacity of the DCF and PCF based WLANs with a special attention to the voice quality to assist in network design and planning. We use the ITU-T E-model to determine the voice quality and formulate a quality impairment budget for medium and high quality calls. We develop two novel Markov models to estimate the delay and loss in the DCF and PCF based medium access mechanisms. To model real world environment closely, we consider the impacts of imperfect channel and capture effect. Since most WLANs operate under the unsaturated condition for a considerable amount of time, we consider both saturated and unsaturated traffic conditions. In conjunction to network performance, we also incorporate the characteristics of voice codecs and dejitter buffer. We identify that the delay and loss in the queue degrade voice quality significantly and utilize standard queuing analyses to determine its impact. Extensive simulation studies show a close match with our analytical results. In the current literature, no analytical model exists for the voice capacity of multichannel, multihop WLANs with multi-interface nodes. Especially, the formation of critical zones due to accumulated packet arrivals in multihop networks creating a bottleneck was never considered. We model multi-interface WLAN nodes operating in multichannel, multihop WLANs and determine the voice capacity of such networks. We also identify and analyze the benefit of utilizing multihop networks in order to exercise spatial reuse and improve voice capacity. Considering that DCF offers wider coverage and spatial reuse while PCF offers lower delay by the use of time synchronized medium access, we present a novel medium access mechanism that combines the best features of the above two and yields performance benefits including VoIP call capacity increase through the introduction of “virtual access point” concept. The protocols for different network scenarios are explained and performance comparisons to the DCF and PCF based medium access are illustrated. The capacity models proposed in this thesis will be very useful in designing WLANs to maintain acceptable voice quality using off-the-shelf WLAN devices supporting DCF and PCF while the incorporation of our proposed medium access mechanism will improve the voice performance in future WLANs using virtual access point enabled mobile devices.