Increasing throughput in IEEE 802.11 by optimal selection of backoff parameters. Issue 1 (1st January 2015)

• Record Type: Journal Article
• Authors: Parker, Ben M.; Schormans, John Alexander; Gilmour, Steven G.
• Is Part Of: IET networks. Volume 4:Issue 1(2015)
• Page Start: 21
• Page End: 29
• Publication Date: 2015-01-01
• Subjects: wireless LAN -- Markov processes -- Monte Carlo methods -- carrier sense multiple access; IEEE 802.11 standard -- carrier sense multiple access -- collision avoidance -- Bianchi model -- Markov chain -- heterogeneous device -- Monte Carlo simulation -- maximum contention window time -- minimum contention window time -- CWmin -- CWmax -- trafflc source -- exhaustive search; Computer network architectures -- Periodicals; Computer network protocols -- Periodicals; Information networks -- Periodicals; Telecommunication systems -- Periodicals; 004.605
• DOI: 10.1049/iet-net.2013.0021
• Languages: English
• ISSNs: 2047-4954

Abstract

Abstract : The IEEE 802.11 standard uses Carrier Sense Multiple Access with Collision Avoidance to avoid multiple devices simultaneous transmitting on a shared transmission medium. In this study, Bianchi's model for IEEE 802.11 is studied and the authors suggest some important improvements. Firstly, they expand the state space of the Markov chain to model the evolution of a network, instead of a single device. Secondly, they relax the assumption that the network must be saturated. Thirdly, they extend the model to allow for heterogeneous devices with different transmission profiles. They use this new model to perform Monte Carlo simulation to discover the impact of the minimum and maximum contention window times (CWmin and CWmax ) in the standard on measures of throughput in a network. By exhaustive search over a parameter space, they find optimal values for these devices for any given network model, and show that the recommended parameters in the IEEE 802.11 standard are not optimal. They consider both average and minimum throughput, and show that increases in throughput of around 8% are possible for saturated networks, and that even greater improvements are possible for any case in which the traffic sources are not homogeneous, that is, any real scenario.