Electrical Engineering
      and Computer Sciences

Electrical Engineering and Computer Sciences

COLLEGE OF ENGINEERING

UC Berkeley

Maximizing Performance in Long Distance Wireless Networks for Developing Regions

Sergiu Nedevschi

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2009-27
February 10, 2009

http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-27.pdf

Today we are witnessing a large disparity between the levels of network connectivity in industrialized countries, and the ones in the developing world.This digital divide is caused by the uneven distribution of wealth around the world, and has the effect of reinforcing this polarization, by providing increased economic opportunities to people that already afford access to information technology. This divide is partially addressed by the growth of wireless and cellular technologies, but these technologies remain financially inviable in rural regions, with sparse and financially-constrained users. To address rural and remote network coverage, we propose the use of multi-hop wireless networks relying on long-distance point-to-point links. By using inexpensive, off-the-shelf Wi-Fi radios, and connecting them to high-gain directional antennas, we can build inexpensive, high-throughput links exceeding tens or even hundreds of kilometers in length. In theory, these wireless long-distance (WiLD) networks have the potential to deliver low-cost connectivity to remote areas. Unfortunately, the performance achieved using the standard 802.11 MAC in long links is very poor, with high and asymmetric packet loss rates, and with low throughput over wireless paths spanning multiple hops. In this dissertation, we understand the causes for low performance in these scenarios, and build MAC- and PHY-layer mechanisms that address these problems and maximize end-to-end network performance. Using extensive measurements we identify the primary sources of performance degradation. To deal with these problems we design and build WiLDNet, a system that includes a spatial-reuse TDMA MAC and a combination of FEC and ARQ-based link-layer loss recovery mechanisms. We deploy WiLDNet in real-world networks, and show that it eliminates most packet losses and increases link utilization, delivering good end-to-end UDP and TCP throughput. We incorporate the lessons learned from our deployments in the design of JazzyMAC, a MAC that maximizes network-wide throughput and minimizes packet delay by using variable-length transmission slots which change dynamically according to traffic. We demonstrate the appropriateness of our solutions by deploying them in several networks in developing countries, including the Aravind Eye Hospital network in India that uses our technology to provide telemedicine services to many thousands of patients.

Advisor: Eric Brewer


BibTeX citation:

@phdthesis{Nedevschi:EECS-2009-27,
    Author = {Nedevschi, Sergiu},
    Title = {Maximizing Performance in Long Distance Wireless Networks for Developing Regions},
    School = {EECS Department, University of California, Berkeley},
    Year = {2009},
    Month = {Feb},
    URL = {http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-27.html},
    Number = {UCB/EECS-2009-27},
    Abstract = {Today we are witnessing a large disparity between the levels of network connectivity in industrialized countries, and the ones in the developing world.This digital divide is caused by the uneven distribution of wealth around the world, and has the effect of reinforcing this polarization, by providing increased economic opportunities to people that already afford access to information technology.  This divide is partially addressed by the growth of wireless and cellular technologies, but these technologies remain financially inviable in rural regions, with sparse and financially-constrained users. 

To address rural and remote network coverage, we propose the use of multi-hop wireless networks relying on long-distance point-to-point links. By using inexpensive, off-the-shelf Wi-Fi radios, and connecting them to high-gain directional antennas, we can build inexpensive, high-throughput links exceeding tens or even hundreds of kilometers in length.  In theory, these wireless long-distance (WiLD) networks have the potential to deliver low-cost connectivity to remote areas.  Unfortunately, the performance achieved using the standard 802.11 MAC in long links is very poor, with high and asymmetric packet loss rates, and with low throughput over wireless paths spanning multiple hops. 

In this dissertation, we understand the causes for low performance in these scenarios, and build MAC- and PHY-layer mechanisms that address these problems and maximize end-to-end network performance.  Using extensive measurements we identify the primary sources of performance degradation.  To deal with these problems we design and build WiLDNet, a system that includes a spatial-reuse TDMA MAC and a combination of FEC and ARQ-based link-layer loss recovery mechanisms.  We deploy WiLDNet in real-world networks, and show that it eliminates most packet losses and increases link utilization, delivering good end-to-end UDP and TCP throughput. We incorporate the lessons learned from our deployments in the design of JazzyMAC, a MAC that maximizes network-wide throughput and minimizes packet delay by using variable-length transmission slots which
change dynamically according to traffic. 

We demonstrate the appropriateness of our solutions by deploying them in several networks in developing countries, including the Aravind Eye Hospital network in India that uses our technology to provide telemedicine services to
many thousands of patients.}
}

EndNote citation:

%0 Thesis
%A Nedevschi, Sergiu
%T Maximizing Performance in Long Distance Wireless Networks for Developing Regions
%I EECS Department, University of California, Berkeley
%D 2009
%8 February 10
%@ UCB/EECS-2009-27
%U http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-27.html
%F Nedevschi:EECS-2009-27