Forecasting-based Cloud-assisted Dynamic Channel Assignment Mechanism for Mesh WiFi Networks

Abstract

In this paper, we propose a cloud-assisted dynamic channel assignment system for WiFi mesh networks considering both the 2.4 GHz and 5 GHz interfaces to increase the overall performance and user experience in the WiFi network. Our solution utilizes periodic interference level measurements by the access points (AP) in all possible channels via conducting clear channel assessments. These measurements are sent to and processed by a cloud component with a forecasting module that predicts the state of each applicable channel in the near future. Finally, a channel change decision is sent to each AP if there is a better channel than its operating channel in the near future.We have conducted numerous field trials for a good selection of the various key parameters of the system with both the overall system’s performance and impact over time-sensitive critical applications such as real-time applications in mind. We have also conducted a field trial of our proposed system over a large real-life population of fifty thousand APs and compared its performance against the widely deployed Least Congested Channel Search (LCCS) mechanism. Our results show that not only our mechanism outperforms LCCS in terms of operating channel interference level but achieves this goal with much less number of channel changes yielding a much less disruptive user experience.