Analysis, limitations and improvements of voice over IP in IEEE 802.11 WLAN

Wireless voice over IP (VoIP) is an important emerging service in telecommunications due to its potential for replacing todays’ cell phone communication wherever wireless local area networks (WLANs) are deployed. However, the number of VoIP calls that can be supported in the widely deployed IEEE 802.11 WLAN is limited and therefore may hinder the success of wireless VoIP as a viable alternative. In this thesis we investigate the limited VoIP capacity and provide new insight and solutions to improve the number of simultaneous VoIP calls that can be maintained with an acceptable level of voice call quality. To improve the VoIP capacity we first discuss a solution based on a quality of service (QoS) mechanism of the IEEE 802.11 protocol. In particular we utilize a transmission opportunity (TXOP) parameter of the medium access control (MAC) as a simple solution to increase the number of VoIP calls. We develop a detailed analytical model to show that a significant increase (≈ 100%) in call capacity can be achieved. We also discuss the implications of this solution and provide ideal parameter settings to maximize the number of concurrent calls without affecting the overall system performance. Using our analytical model, we study the impact of the buffer size on VoIP capacity. We show that there exists an optimal buffer size that maximizes the number of VoIP calls, and we show that further increasing the buffer beyond this ideal value will not result in an increase in call capacity. In particular, we show that VoIP capacity is independent of the buffer size, given the minimum optimal buffer, and that this finding also holds in conjunction with the previous solution to increase the VoIP capacity. Based on this finding, we develop a closed-form expression for the maximum number of VoIP calls as a function of the TXOP parameter. Using the closed-form expression we then propose a novel VoIP capacity approximation equation. This simple yet accurate approximation formula allows us to provide some further insight into VoIP capacity in WLAN and its limits. We then propose a novel dynamic codec with priority (DCwP) scheme that exploits a tradeoff between call quality and traffic priority to increase call capacity while providing a high level of voice call quality. In this scheme, users are encouraged to switch to a lower quality voice codec during periods of high contention, thus enabling them to maintain the call. To compensate for the reduction in call quality, and to give an incentive to the user to adjust the voice codec, a higher priority is given at the AP for voice calls originating from these users, thus providing a high throughput and a higher than average call quality. To show the benefits of this scheme, we develop a detailed analytical model, and show that depending on the voice codec setting that a VoIP capacity gain of between 16% and almost 300% can be achieved with an above average call quality as indicated by results obtained using the ITU E-model. Our analytical model also allows us to provide further insight into a multi-queue, multi-traffic environment, in particular the impact of internal collisions. Finally, we investigate the impact of TCP traffic on VoIP capacity in an IEEE 802.11 WLAN. Using ns-2 simulation we study the impact of TCP traffic on the call capacity when the channel access is controlled using the distributed coordination function (DCF) or enhanced distributed channel access (EDCA) is used. We confirm that the TCP flow direction is an important parameters to consider, irrespective of the channel access function. We then show that the solution based on the TXOP parameter can provide benefits to VoIP and TCP traffic. In particular, we show that using the TXOP parameter yields a high call capacity while maximizing the TCP throughput. However, we also show that even though the TXOP solutions increase the number of calls and the TCP throughput, our proposed DCwP scheme can provide superior results in terms of VoIP capacity and TCP throughput for different scenarios.