Exploring the predictability of network metrics in the presence of unreliable wireless links

Abstract

In designing, analyzing and monitoring wireless sensor networks (WSN), engineers are interested in global performance metrics such as energy consumption, delivery rate, and delay. Such metrics are often variable and stochastic in nature. In traditional broadband networks the variability is primarily due to traffic variations and bandwidth constraints; in WSN there are additional sources of variability including energy, computation constraints, and most significantly, due to the inherently unreliable nature of wireless links. The goal of our study is to understand how link-layer unreliability impacts global performance metrics.