Telecommunications is converging on the use of IP based networks. Due to the low cost of VoIP applications, they are being increasingly used instead of conventional telephony services. IEEE802.11 WLANs are already widely used both commercially and domestically. VoIP applications will also expand from usage over wired networks to voice communications over IEEE802.11 WLANs. This is known as VoWLAN. The use of VoWLAN may reach the maximum capacity of a wireless channel if there are many simultaneous VoIP calls operating close to each other. There is much published research based on a single IEEE802.11 infrastructure WLAN concluding that packet loss, transmission efficiency and latency issues are the major challenges limiting the VoWLAN capacity. The VoIP service quality will drop sharply when the demands exceed the WLAN’s capacity. This thesis demonstrates that these challenges also apply to distributed WLANs. To extend these findings from the existing research, the analysis in this thesis indicates that the capacity of a single IEEE802.11 WLAN channel is 12 VoIP calls. When the number of simultaneous VoIP calls is within the capacity, the WLAN can deliver more than 90% of the voice packets to the receiver within 150 ms (the lowest network performance for supporting acceptable VoIP service). However, as soon as the traffic loads are beyond the wireless channel capacity e.g. the number of simultaneous VoIP calls is greater than 13, the VoIP service quality catastrophically collapses. When the capacity is exceeded there are almost no voice packets that can be delivered to the receiver within 150 ms. Our research results indicate that the delay accumulation for voice packets in the transmitter’s outgoing buffer causes this problem. Our research also found that dropping ‘stale’ voice packets that are already late for delivery to the receiver can give more transmission opportunities to those voice packets that may still be delivered in time. This thesis presents a new strategy called Active Cleaning Queue (ACQ) which actively drops ‘stale’ voice packets from the outgoing buffer and prevents the accumulation of delay in congested conditions. When ACQ is applied in a saturated wireless channel the network performance for supporting VoIP traffic was found to gradually decrease proportional to the numbers of simultaneous VoIP calls rather than catastrophically collapse. There is also published research suggesting that the aggregation of packets can improve the efficiency of WLAN transmissions. An algorithm called Small Packet Aggregation for Wireless Networks (SPAWN) is also presented in this thesis to improve transmission efficiency of small voice packets in WLANs without introducing further delay to VoIP traffic. The evaluation result shows that after applying the SPAWN algorithm, the VoIP capacity of a single wireless channel can be extended up to 24 simultaneous calls.

Made available in DSpace on 2016-04-04T18:31:24Z (GMT). No. of bitstreams: 1 Rafael Rodrigo Martinati.pdf: 892328 bytes, checksum: bb06e7a9c52ea61c8c7545d477f6b0f3 (MD5) Previous issue date: 2009-06-19 / The objective of this paper is to analyze the development of Service Level Agreement to Voice over Internet Protocol over WIFI services in compliance with COBIT, ITIL and ISO / IEC 20.000-1:20005 and ISO 38.500 frameworks, specially using perceived quality of speech as a SLA requirement . The models MOS, PESQ and E MODEL published by the ITU for evaluation of quality of speech analyzed was identified the hypothesis that the E MODEL provide compliance with the frameworks of ICT governance. It was emulated a Voice over Internet Protocol over WIFI scenario and measured is the variation of the R FACTOR, in different encoders and compared with the changes in jitter and rate of loss packets. We can conclude that E model has characteristics that make it able to be inserted into a Service Level Agreements for Voice over Internet Protocol over WIFI service. / Este trabalho discorre sobre acordos de n?vel de servi?os, ou Service Level Agreements (SLA) para servi?os Voice over Internet Protocol (VoIP) sobre redes Wireless Local Area Network (WLAN). Foram considerados como fundamentos os frameworks Control Objectives for Information and related Technologies (CobiT), Information Technology Infrastructure Library (ITIL) e as normas ISO/IEC 38.500 e ISO/IEC 20.000-1 e constatouse que a qualidade de fala percebida deve ser um requisito de um SLA para VoIP. Foram analisados os modelos de avalia??o de qualidade de fala Mean Opinion Score (MOS), Perceptual Evaluation of Speech Quality (PESQ) e Modelo E, publicados pela International Telecommunications Union (ITU) onde foi identificada a hip?tese de que o Modelo E esteja em conformidade com os frameworks de governan?a de Tecnologia da Informa??o e Comunica??o (TIC), por fornecer um indicador, o fator R, que pode ser considerado como um indicador de SLA. Para comprova??o desta hip?tese, foram analisados dados, obtidos em uma emula??o de um ambiente real de redes WLAN para medi??o da varia??o do fator R, utilizando-se de dispositivos codificadores (CODEC) e cen?rios diferentes. Constatou-se que o fator R apresentou a mesma tend?ncia de qualidade de fala nos diferentes dispositivos CODEC. Foram ent?o comparadas as varia??es de jitter e taxa de perda de pacotes nos mesmos cen?rios e codificadores para analisar se a varia??o destas m?tricas possui relacionamento com as medidas do fator R. Identificou-se ent?o que o fator R apresenta caracter?sticas que permitem que ele seja inserido em um SLA de um servi?o VoIP, viabilizando assim a requisitos de governan?a e gest?o de servi?os de TIC.