Modeling And Analyzing The Perlink Throughput Fairness Of The IEEE 802.11 DCF Protocol

IEEE 802.11 DCF protocol provides a fair way to share the medium for nodes in wirless networks. Although it was originally designed for single hop networks, with the rapid development of wireless network technology, the basic access strategy of DCF has been widely used in multihop wireless ad hoc networks, wireless Mesh networks, wireless sensor network, etc. However, when applied to multiple hop wireless networks, it will appear serious unfairness phenomenon among different links. In multihop wireless networks which employ the DCF protocol, some links can achieve much higher throughput while some others starve. The performance analysis of the DCF has been a hot issue in the field of wireless networks. This paper focus on the analysis of the per-link throughput and the fairness among links of the IEEE 802.11 DCF in multihop wirless networks. The main contribution of this work is as follows.(1) At first, the challenge in modelling and analyzing the per-link throughput performance of IEEE 802.11 DCF protocol in multihop wireless networks is analyzed. Based on that, a fixed-length time based Markov chain model is provided. The Markov model can be classified as the backoff process, successful transmission process, collision process, and freezing process. The Binary Exponent Backoff(BEB) scheme is taken into consideration in the backoff process. In the successful transmission process and the collision process, two different kinds of collisions are considered. A continues time Markov chain model is used to obtain the the freezing probability and the number of slots in a freezing process. The probabilities of the states in the fixed-length time based Markov model are derived in an interative way. By solving the model, the transmission probability, collision probability and throughput performance of each link can be obtained. The simulation results in Qual Net simulator and the theoretical results obtained from the proposed Markov chain model have been compared to validate the accuracy of the proposed model.(2) The cause of the unfairness phenomenon among different links in multihop wireless networks are be classified as the unfairness of transmission probability and the unfairness of collision probability in this paper. The effect of the contention parameters on the throughput performance of each link is analyzed by some experiments. On one hand, since the transmissions of the links can overlap in time with each other, the links that locate in the centre of the topology may sense the channel busy all the time and nearly have no chance to start a transmission while other links can start a transmission at any time. On the other hand, for some links, they may incur serious collision although they have enough chance to start their transmissions. Almost all the packets on the links may be collided due to the hidden terminal problem, leading to the starvation. According to the proposed markov chain model, we calculate the optimized contention window value when all the links share the medium in a fair way which leaving more transmission opportunity and lower collision probability for the starving links. The simultion results in the Qual Net simulator show that the unfairness problem among different links, whatever the cause, can be solved by our optimization strategy of the contention window value.