The World Wide Web is largely unusable or prohibitively slow for a majority of users in the developing world due to poor connectivity. In a large university in India with a good 8 Mbps downlink, we have frequently observed users experiencing page download times in excess of 120 seconds, which is exceedingly slow. The conventional model for Web access is fundamentally ill-suited for emerging regions due to four basic challenges. The first issue is the sluggish growth in connectivity. Akamai quarterly reports consistently indicate that the slowest growths in connectivity have been recorded in South Asia, South America, and Africa – three regions which account for the majority of emerging regions in the world. The second related issue is the growing content-connectivity gap. In contrast to connectivity, Web pages have grown in size and complexity over the past decade. The average web page size has grown roughly by a factor of 40-50 in the last decade. The third problem is the phenomenal growth in the number of Transmission Control Protocol (TCP) flows per web request resulting in a TCP breakdown problem. Standard browsers have evolved to optimize for opening multiple TCP connections (4-6) to each web server yielding as many as 30-50 competing TCP connections to download one page. This level of contention drives a majority of TCP flows in low bandwidth networks to operate in sub-packet regimes, where the fair share per flow is less than one packet per round-trip time. As a result, TCP completely breaks down resulting in extremely poor performance and an unusable Web experience. Finally, we also observed that conventional caching and network-level optimizations yield limited benefits in these settings. Addressing the Web access problem has been a large-scale initiative and several Masters and PhD students have worked on different parts of this project.

We are proposing a new Web stack for emerging regions that makes fundamental changes to different layers of the existing Web stack to enhance web access under different types of poor connectivity scenarios – low bandwidth networks, intermittent connectivity and offline access. We focus mainly on three different scenarios: Web under static wired connections, Mobile with 2G, 3G and LTE, and extremely low connectivity under SMS. The new Web stack consists of the following layers:

We are thankful for the funding support from the Center for Technology and Economic Development (CTED) at NYUAD, NSF, Microsoft Research, and Google for this project.