Scalable Routing for Mechanical Backhaul Networks

Providing Internet connectivity to rural regions in developing countries is a challenging proposition. To effectively serve the needs of the rural poor, solutions need to be low-cost, in addition to being reliable and scalable. Mechanical backhaul networks are a low-cost solution to providing delay tolerant connectivity to such regions. In these networks, shared kiosk-based computers in villages connect opportunistically to wireless router nodes present on buses and other similar vehicles. These vehicles “mechanically” ferry data between kiosks and Internet gateways. With only as much as a minute of contact duration, the ferry can transfer several tens of MB of data over a wireless link. In this work, we design, implement, and evaluate a protocol for scalable and robust routing in such a setting. We achieve scaling by the appropriate use of hierarchy in routing; and we achieve robustness through the use of a smart flooding approach. During our design, we also take into consideration constraints peculiar to our problem setting, which include use of low power, low speed machines to keep the costs low, availability of low speed (100 Kbps) Internet connections at various locations, and availability of cheap non-volatile storage. Using a combination of simulations and a prototype implementation, we show that our solution can scale to large-scale, country-wide deployments. Further, we show that for the problem setting under consideration, with a network size of 100,000 nodes, a kiosk-based computer can be allowed close to 260MB of upload per month. This upload capacity is more than sufficient for applications like e-mail, offline web browsing, e-governance services, transfer of agricultural knowledge to rural regions, etc. We identify and analyze various aspects of the problem setting that can be bottlenecks for the system performance. We show that low power machines form the primary bottleneck of performance in our solution.