Performance modelling of fairness in IEEE 802.11 wireless LAN protocols

Abstract

Wireless communication has become a key technology in the modern world, allowing network services to be delivered in almost any environment, without the need for potentially expensive and invasive fixed cable solutions. However, the level of performance experienced by wireless devices varies tremendously on location and time. Understanding the factors which can cause variability of service is therefore of clear practical and theoretical interest. In this thesis we explore the performance of the IEEE 802.11 family of wireless protocols, which have become the de facto standard for Wireless Local Area Networks (WLANs). The specific performance issue which is investigated is the unfairness which can arise due to the spatial position of nodes in the network. In this work we characterise unfairness in terms of the difference in performance (e.g. throughput) experienced by different pairs of communicating nodes within a network. Models are presented using the Markovian process algebra PEPA which depict different scenarios with three of the main protocols, IEEE 802.11b, IEEE 802.11g and IEEE 802.11n. The analysis shows that performance is affected by the presence of other nodes (including in the well-known hidden node case), by the speed of data and the size of the frames being transmitted. The collection of models and analysis in this thesis collectively provides not only an insight into fairness in IEEE 802.11 networks, but it also represents a significant use case in modelling network protocols using PEPA. PEPA and other stochastic process algebra are extremely powerful tools for efficiently specifying models which might be very complex to study using conventional simulation approaches. Furthermore the tool support for PEPA facilitates the rapid solution of models to derive key metrics which enable the modeller to gain an understanding of the network behaviour across a wide range of operating conditions. From the results we can see that short frames promote a greater fairness due to the more frequent spaces between frames allowing other senders to transmit. An interesting consequence of these findings is the observation that varying frame length can play a role in addressing topological unfairness, which leads to the analysis of a novel model of IEEE 802.11g with variable frame lengths. While varying frame lengths might not always be practically possible, as frames need to be long enough for collisions to be detected, IEEE 802.11n supports a number of mechanisms for frame aggregation, where successive frames may be sent in series with little or no delay between them. We therefore present a novel model of IEEE 802.11n with frame aggregation to explore how this approach affects fairness and, potentially, can be used to address unfairness by allowing affected nodes to transmit longer frame bursts.