Date of Original Version
Abstract or Description
As wired computer networks support increasingly sophisticated applications and wireless local area networks become ubiquitous and fast, it is more natural for users to seek a "wireless Internet" experience that is qualitatively the same as that provided by the wired Internet. However, wireless LANs pose two fundamental challenges to this vision. Harsh and dynamic error environments challenge end-to-end adaptation at the transport and application layers. In addition, dynamic and location-dependent errors challenge the notion of "fair" scheduling of flows sharing a wireless link. This dissertation advances the claim that a combination of protocol-blind link-level error control and error-sensitive link scheduling effectively addresses these two challenges. The first step is a bit-level trace-based analysis of the error environment provided by a particular wireless link technology (AT&T WaveLAN I) in the face of attenuation and interference. Based on the insights revealed by this analysis, the next step is designing an adaptive link-level error control module. Finally, we propose a new notion of fairness appropriate for error-prone wireless links. This new scheduling approach balances application sensitivity to error-induced throughput changes against the need to preserve link efficiency. The proposed mechanisms are evaluated by deploying them in an actual operating system, running on real hardware, and subjecting them to trace replay of the error environments we observed. Particular attention is given to the interaction between link-level error control and TCP's end-to-end error and congestion control mechanisms. The result is a system that can noticeably improve application-level throughput in a wide variety of error environments and can sensibly allocate limited and dynamic network capacity among network flows.