Efficient voice and video traffic delivery in IEEE 802.11 WLANs: design, implementation and experimental evaluation
MetadataShow full item record
While the IEEE 802.11 protocol has fostered ubiquitous connectivity for wireless users, it was not originally designed to efficiently handle voice or video traffic, which nowadays accounts for most of the Internet traffic. Voice traffic handling is extremely inefficient in existing WLANs due to the large overhead of the protocol and the time spent in contention. Video multicast streaming is both ineffective and inefficient, as its transmissions lack reliability and use a low Modulation and Coding Scheme (MCS). The more robust, but slower, transmission rates result in the well-known performance anomaly problem that degrades network performance. In this thesis we analyze, design and implement solutions to improve the network performance and efficiency when voice and video traffic are present. We first design and experimentally validate a simple yet very effective scheme (VoIPiggy) to improve the efficiency of WLANs with voice traffic. The key idea of VoIPiggy is to piggyback voice frames onto MAC layer acknowledgments, hereby reducing both the frame overhead and the number of times to access the medium, i.e., the time spent in contention as stations aggregate two frames into a single transmission. To quantify the gains of our proposal, we perform an analysis in terms of capacity and delay. Our experimental and analytical results show a dramatic performance improvement, doubling the number of voice conversations that can be allocated in WLANs. Second, we explore a set of mechanisms included in the 802.11aa standard (namely, Direct Multicast Sequence, Groupcast with Retries Unsolicited Retries and Block Acknowledgment), which allow efficient and robust multicast transmission in WLANs. To that aim, we first perform extensive simulations to understand the trade-offs resulting from using the proposed mechanisms. Our results quantify these trade-offs in terms of robustness, resource consumption and complexity, and provide a set of recommended guidelines for their use according to the network scenario. Finally, we develop the first open source implementation of these mechanisms over commodity hardware. We assess their performance under a variety of real-life scenarios, and outline their effectiveness and efficiency when delivering video traffic. Our results provide key insights on which mechanism results more appropriate for a given scenario, this being specified in terms of number of stations, background traffic and video bandwidth.