| The growing number of wireless device users and the proliferation of new wireless applications are imposing increasingly greater performance demands on today's wireless networks. As a result, much effort has been directed towards more efficient or heavily resourced physical layers, aimed at improving performance by reducing communication overheads and increasing spatial reuse. Fundamentally, however, distance and decreasing signal strength limit the performance gains hardware approaches can achieve; clients distant from an Internet gateway (e.g., an access point) will experience poor performance.;This dissertation explores a complementary MAC layer approach to increase the throughput of clients distant from their access points. This approach employs multi-hop relaying to leverage short, high-rate links when network participants can forward traffic for one another, and multiple rates are supported. I develop analytic models to evaluate the potential performance benefits of this approach across a range of design choices including relay placement, spatial reuse, power control, and multiple channels and radios. I find that with relatively few relays, multi-hop relaying can yield order of magnitude throughput speedups for many clients. Surprisingly, I find that the simplest design choices (namely single radio nodes without spatial reuse or power control) offer benefits nearly as large as much more sophisticated ones. I use simulations in an 802.11 network to study the effects of practical constraints (e.g., discrete rates, ACKs, CSMA/CA) on multi-hop relaying, and argue that these performance benefits can be achieved in practice. |
