Towards Dynamic, SDN-assisted Interface Bonding for Heterogeneous 802.11 Devices

  • Michael J Dilmore

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

While state of the art mobile devices offer multiple options for wireless connectivity, there are still challenges, open questions, and organisational barriers with respect to the application of interface bonding techniques in real-world networks such as 802.11. Several works have already demonstrated the benefit of leveraging Software Defined Networking towards the interface bonding problem. By centralising bond functions such as link monitoring and interface selection, the SDN controller is able to exploit its global network view
and dynamically update the bond configuration in responses to changes in link state. These works are so far limited to small-scale, hardware-based proof of concepts, and a rigorous analysis of the performance of such approaches has not been provided. Towards this research area, the work presented investigates robust dynamic interface selection and related performance metrics
for heterogeneous 802.11 device types using Software Defined Networking. Using a custom 802.11 interface bonding framework implemented in MatLab, the performance of the traditional static approach to bond configuration is evaluated for dual-11ac and heterogeneous 11ac-11ah, point-to-point interface bonds at Layer 2, under a range of applied load scenarios, with
the aim of boosting access speeds over the highly contested wireless last-hop. The static bond results are then used as a benchmark against which to independently evaluate the performance of the main thesis reference work: the SDN architecture and decision-making framework for dynamic MPTCP link selection for dual-802.11 bonds by the authors Nam et al. The results show that although a substantial performance improvement was seen via the
dynamic approach, the algorithm was unable to protect at all times against the counter-productive scenario where the bond throughput is reduced below that of the single fastest 802.11 link. To address the limitations of the contemporary work, a novel, predicted-throughput-based algorithm
for dynamic selection of heterogeneous 802.11 devices is proposed.
Date of Award7 Dec 2021
Original languageEnglish
Awarding Institution
  • University of Bristol
SupervisorAngela Doufexi (Supervisor) & George Oikonomou (Supervisor)

Cite this

'