How does one go about optimizing the most extensive public Wi-Fi network in the nation? That was the task we were asked to perform. Comcast Xfinity WiFi network is a vast Wi-Fi overlay network, with close to 21M access points. These Access Points are deployed outdoors, inside small and medium businesses, and in subscriber’s homes. The purpose of this ongoing project is to improve the customer experience of the Xfinity WiFi network and increase the overall traffic usage, and increase the data offload of Xfinity Mobile users over the Xfinity WiFi network.
The fundamental impact when customer faces a bad experience on Wi-Fi is that they may decide to turnoff Wi-Fi on their mobile devices and may use valuable LTE data. While they may do so while outdoors, our data shows that sometimes these customers will continue to use LTE data while in their own homes and would not connect to their home Wireless Gateway provided by Comcast.
Some of the challenges facing the optimization of Xfinity WiFi network, apart from the sheer size, were the fact that no central or distributed controller manages and coordinates the system from a Wi-Fi perspective. The Access Points are not aware of each other and have no communication path between them. The network was never designed for complete coverage or even to allow seamless roaming between Access Points. Also, over 98% of the network comprises of residential Access Points that act as a Home Hotspot (HHS). We had minimal control over what Wi-Fi configuration changes can be applied to these devices.
The first step we took was to break the optimization of a nationwide network into smaller, more manageable zones. We took several of such zones and used them to prove the methods, processes, and tools. We broke the process into multiple iterations, using traditional RF planning, measurements, and predictive tools to test after each iteration. Thus, establishing a baseline and iterative improvements as we progressed.
In parallel, the CommScope team developed a Big Data pipeline to ingest data from multiple sources of measurement; Multiple Network telemetry data, Access Point telemetry, enriched with geographic information data. This Big Data Analytics platform is turning discrete sources of data into powerful insights and recommendations.
This paper will describe in detail the methods, processes, and tools we developed to address this challenge and the results observed so far.