Paul Schmitt

The increased bandwidth fallacy: performance and usage in rural Zambia

Broadband Internet access has become a critical part of socio-economic prosperity; however, only 6 in 100 inhabitants have access to broadband in developing countries. This limited access is driven predominately by subscriptions in urban areas. In rural developing communities, access is often provided through slow satellite, or other low-bandwidth long-distance wireless links, if available at all. As a result, the quality of the Internet access is often poor and at times unusable. In this paper we study the performance and usage implications of an Internet access upgrade, from a 256kbps satellite link to a 2Mbps terrestrial wireless link in rural Zambia. While usage did not immediately change, performance improved soon after the upgrade. By three months post-upgrade, however, subscribers began to use the faster connection for more bandwidth-hungry applications such as video-streaming and content upload. This change in usage resulted in dramatic deterioration of network performance, whereby the average round trip time doubled, the amount of bytes associated with failed uploads increased by 222% and that of failed downloads by 91%. Thus, while an Internet access upgrade should translate to improved performance and user experience, in rural environments with limited access speed and growing demand, it can bring unexpected consequences.

Community-Level Access Divides: A Refugee Camp Case Study

Despite the appearance of uniform availability of mobile services, in many locales granular network analyses reveal the persistence of physical access divides. It stands to reason these divides, similar to those at larger scales, are also reflections of community-level social and economic divides. In this research, we examine community-level physical access divides in the context of a Syrian refugee camp. The investigation combines detailed network and organizational analyses to characterize the divides and identify factors influencing their creation and potential solutions. Our findings show that even in the limited confines of a refugee camp, coverage patterns and bandwidth availability differ significantly both within and between the networks of three mobile cellular carriers. These patterns, together with the overall configuration of network infrastructure, demonstrate three community level divides: an inter-carrier congestion divide, a spatial distribution divide, and an inter-network divide. We identify a number of linkages between these divides and the social, organizational and humanitarian context of the camp. Building on these analyses, we provide recommendations to ameliorate these divides for both residents and camp management.

Internet Media Upload Caching for Poorly-Connected Regions

Media uploads and downloads, even those on the order of a few hundred kilobytes, commonly fail when attempted over lossy, low-bandwidth, and high latency connections. These conditions, which are common for networks in rural, resource-poor areas, result in the inability for residents of these areas to fully participate in the modern Internet. We study traffic traces collected from two networks and find high locality of interest as well as poor performance for Internet media. We also find that users often abort uploads due to network performance. Given that media content produced by local users is often heavily consumed by local users, we propose VillageCache, a system which allows for appropriate local transformation and redistribution of media uploaded through an Internet cache. We build VillageCache and find it successfully delivers cached media, providing an orders of magnitude improvement in file transfer performance, bandwidth reduction, and the virtual elimination of video stalls in the face of poor network connectivity. Caches with upload capabilities placed at the edge of poorly-connected networks will enable users to produce and consume media-rich content while mitigating the constraints present in under-resourced networks.

Navigating connectivity in reduced infrastructure environments

Modern communications systems rely on assumptions of centralized systems and underlying infrastructure that make them vulnerable in computing within limits scenarios. In this paper we offer case studies of issues facing connectivity in three locations in developing regions as a proxy for networks in such scenarios. We explore obstacles presented by such circumstances and future research directions for systems in resource-poor networking environments. We propose democratizing infrastructure deployment in a bottomup manner and re-thinking network architectures to be more decentralized with less reliance on centralized support systems.

Low On Air: Inherent Wireless Channel Capacity Limitations

Wireless connectivity has fundamentally changed the way we connect and interact with the world. Over the past fifteen years there has been an exponential increase in wireless data usage, a trend that is predicted to continue. The overall capacity for wireless connectivity is limited in that it operates over electromagnetic spectrum, and the usable range of spectrum is both finite and already scarce. We argue that the growth in demand that we currently see is unsustainable in the long-term, as spectrum resources will become fully exhausted. While current lines of research seek to increase spectrum efficiency, increases in the future will achieve diminishing returns. In this work we present current technologies as well as cutting-edge research related to maximizing the efficiency of wireless systems, and offer research questions that will become critical as we near the limits of wireless connectivity.

A Study of MVNO Data Paths and Performance

Characterization of mobile data traffic performance is difficult given the inherent complexity and opacity of mobile networks, yet it is increasingly important as emerging wireless standards approach wireline-like latencies. Mobile virtual network operators (MVNOs) increase mobile network topology complexity due to additional infrastructure and network configurations. We collect and analyze traces on mobile carriers in the United States along with MVNO networks on each of the base carriers in order to discover differences in network performance and behavior. Ultimately, we find that traffic on MVNO networks takes more circuitous, less efficient paths to reach content servers compared to base operators. Factors such as location of the destination server as well as the provider network design are critical in better understanding behaviors and implications on performance for each of the mobile carriers.

Smartcell: small-scale mobile congestion awareness

Despite improvements and expansion of cellular coverage in developing regions, a substantial qualitative divide remains. Maps that display the presence or absence of cellular coverage mask critical differences in infrastructure performance and client load. In order to illuminate challenges faced by users of such mobile networks, we collect and analyze GSM network measurements at the local scale. We discover higher network congestion in developing regions as well as performance differences between available carriers in each location. Based on our findings, we propose an app, called SmartCell, that informs and empowers users in near real time to seek out improved mobile connectivity.

PhoneHome: Robust Extension of Cellular Coverage

Ubiquitous cellular coverage is often taken for granted, yet numerous people live outside, or at the fringes, of commercial cellular coverage. Further, natural disasters and human rights violations cause the displacement of millions of people annually worldwide, with many of these people relocating to shelters and camps in areas at or just beyond the margins of existing cellular infrastructure. In this work we design PhoneHome, a system prototype that extends existing cellular coverage to areas with no or damaged cellular infrastructure, or infrastructure that is otherwise poorly performing. We explore the feasibility of PhoneHome and address current limitations and future directions for independently operated, user-extensible cellular infrastructure.

Helping the Lone Operator in the Vast Frontier

While the networking literature is replete with work on managing and operating networks---from the specifics of protocols to the design of management tools and architectures---there is comparatively little work on planning a network to be rolled out. In part this is because the task of network planning typically falls to carriers (for backbones) and cloud providers (for datacenters), which have the resources and the control to meet their specific needs. Here we consider network planning in situations that are quite different: resource poor and highly constrained. Frontier networks are often planned by lone operators, and while these networks are small in size, they are large in number: they individually serve relatively few users but in aggregate serve millions of users. The key challenge here is the mismatch between the resources of these small operators and the complexity of their network planning. In this paper we detail the difficulties in this context, building on our teams first-hand knowledge many networks of this type. We then present our initial efforts at automating frontier network planning, discuss next steps, and outline several open problems.