| Hi-index | 0.00 |
The first generation of wireless networking protocols have treated time varying wireless characteristics like fading and interference as adversities that need to be combated. The design of a second generation wireless protocol stack must aim at a finer optimization of wireless networks that exploit these characteristics natively in every aspect of stack design. Given the increasingly real-time nature of wireless applications, it must have first class support for latency-sensitive applications. This dissertation explores candidate elements of such a second generation wireless stack. Our main argument is for a "physical layer aware" design that exploits the unique properties of wireless communication, namely its broadcast nature, multipath fading, and multi-access interference effects, as first class entities. Our candidate stack attempts to leverage the wireless nature of broadcast at every layer, using it to reduce floor acquisition costs. It leverages multi-user and multi-channel diversity to exploit fading at different timescales to boost performance. Our candidate stack is tailored to real-time communication. The design of scheduling algorithms that optimally inter-play real-time requirements with the physical characteristics of the wireless medium has been a huge challenge. For unicast applications over a single-rate wireless network, our scheduling algorithms provide real-time guarantees while opportunistically harnessing diverse channel conditions arising from multi-user diversity in a theoretically optimal fashion. When the underlying physical layer provides multiple modulation techniques, our algorithms harness multi rate capability to trade-off increased transmission speed against the increased reliability associated with simpler modulation techniques. When the real-time applications themselves are of a multicast nature, our scheduling policies optimally trade-off "real time" urgency against the "broadcast gain" that is obtained by broadcasting a packet that is simultaneously useful to many receivers. Our candidate stack uses traffic adaptive multipath routing to guide packets probabilistically around network bottlenecks, and automatically create interference-aware routes. Recognizing that theoretical arguments require a cross layer re-design of TCP taking wireless interference into account, our work suggests the use of in-network congestion control to automatically adapt flow rates inside the network.