FiDO: A Community-based Web Browsing Agent and CDN for Challenged Network Environments

Homes located on tribal lands, particularly in rural areas of the United States, continue to lack access to broadband Internet and cellular connectivity [19]. Inspired by previous observations of community content similarity in tribal networks, we propose FiDO, a community-based Web browsing and content delivery system that takes advantage of user mobility, opportunistic connectivity, and collaborative filtering to provide relevant Web content to members of disconnected households via opportunistic contact with cellular base stations during a daily commute. We evaluate FiDO using trace-driven simulations with network usage data collected from a tribal-operated ISP that serves the Coeur d’Alene Indian Reservation in Western Idaho. By collecting data about household Web preferences and applying a collaborative filtering technique based on the Web usage patterns of the surrounding reservation community, we are able to opportunistically browse the Web on behalf of members of disconnected households, providing an average of 69.4 Web pages (all content from a specific URL, e.g., “http://gis.cdatribe-nsn.gov/LandBuyBack/”) crawled from 73% of their top 10 most visited Web domains (e.g., “cdatribe-nsn.gov” or “cnn.com/”) per day. Moreover, this content is able to be fetched and pushed to users even when the opportunistic data rate is limited to an average of only 0.99 Mbps (σ = 0.24 Mbps) and the daily opportunistic connection time is an average of 45.9 minutes (σ = 2.3 minutes). Additionally, we demonstrate a hybrid “search and browse” approach that allocates a percentage of opportunistic resources to the download of user-specified content. By dedicating only 10% of opportunistic windows of connectivity to the download of social media content, 51% of households were able to receive all of their daily expected social media content in addition to an average of 55.3 Web pages browsed on their behalf from an average of 4 different Web domains. Critically, we demonstrate the feasibility of a collaborative and community-based Web browsing model that extends access to Web content across the last mile(s) using existing infrastructure and rural patterns of mobility.