Thesis Defense - Sheth

The reduction in cost and ease of installation of 802.11 based wireless LAN (WLAN) hardware has resulted in a surge of large scale deployments. Furthermore, with the flexibility of open source WLAN device drivers, there is growing interest to extend the basic WLAN technology to outdoor mesh and long distance networks. However, due to the unreliable nature of wireless links and chaotic deployments, users of such networks frequently encounter degraded performance and lack of coverage. Existing approaches that aim to diagnose these problems are inefficient because they are unable to distinguish among the root causes of performance degradation. In this thesis, through detailed measurement analysis of WLAN deployments, I have identified the primary sources of performance degradation in indoor WLAN deployments and outdoor mesh networks. I have designed and implemented fine-grained detection algorithms that are capable of distinguishing between root sources of performance anomalies at the depth of the physical layer using stock WLAN hardware. Unlike the stock 802.11 protocol, which always resorts to rate fallback as a remedy, I have developed targeted loss recovery mechanisms to mitigate the fault and improve the performance of the network. In addition to software based recovery mechanisms, this thesis also explores the use of smart antennas in mitigating the loss. An important property that emerges from my system is that diagnostic observations are combined from multiple sources over multiple time instances for improved detection accuracy and network performance.