Improving Performance for CSMA/CA Based Wireless Networks

Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) based wireless networks are becoming increasingly ubiquitous. With the aim of supporting rich multimedia applications such as high-definition television (HDTV, 20Mbps) and DVD (9.8Mbps), one of the technology trends is towards increasingly higher bandwidth. Some recent IEEE 802.11n proposals seek to provide PHY rates of up to 600 Mbps. In addition to increasing bandwidth, there is also strong interest in extending the coverage of CSMA/CA based wireless networks. One solution is to relay traffic via multiple intermediate stations if the sender and the receiver are far apart. The so called “mesh” networks based on this relay-based approach, if properly designed, may feature both “high speed” and “large coverage” at the same time. This thesis focusses on MAC layer performance enhancements in CSMA/CA based networks in this context. Firstly, we observe that higher PHY rates do not necessarily translate into corresponding increases in MAC layer throughput due to the overhead of the CSMA/CA based MAC/PHY layers. To mitigate the overhead, we propose a novel MAC scheme whereby transported information is partially acknowledged and retransmitted. Theoretical analysis and extensive simulations show that the proposed MAC approach can achieve high efficiency (low MAC overhead) for a wide range of channel variations and realistic traffic types. Secondly, we investigate the close interaction between the MAC layer and the buffer above it to improve performance for real world traffic such as TCP. Surprisingly, the issue of buffer sizing in 802.11 wireless networks has received little attention in the literature yet it poses fundamentally new challenges compared to buffer sizing in wired networks. We propose a new adaptive buffer sizing approach for 802.11e WLANs that maintains a high level of link utilisation, while minimising queueing delay. Thirdly, we highlight that gross unfairness can exist between competing flows in multihop mesh networks even if we assume that orthogonal channels are used in neighbouring hops. That is, even without inter-channel interference and hidden terminals, multi-hop mesh networks which aim to offer a both “high speed” and “large coverage” are not achieved. We propose the use of 802.11e’s TXOP mechanism to restore/enfore fairness. The proposed approach is implementable using off-the-shelf devices and fully decentralised (requires no message passing)