Experimental Comparison of Bandwidth Estimation Tools for Wireless Mesh Networks

Measurement of available bandwidth in a network has always been a topic of great interest. This knowledge can be applied to a wide variety of applications and can be instrumental in providing quality of service to end users. Several probe-based tools have been proposed to measure available bandwidth in wired networks. However, the performance of these tools in the realm of wireless networks has not been evaluated extensively. In recent years, there has also been some work on estimating bandwidth in wireless networks via passively monitoring the channel and determining the 'busy' and 'idle' periods. However, such techniques have primarily been evaluated via simulations only. In this work, we perform an extensive experimental comparison study of both passive and active bandwidth estimation tools for 802.11-based wireless mesh networks. We investigate the impact of interference, packet loss, and 802.11 rate-adaptation, on the performance of these tools. Our results indicate that for wireless networks, a passive technique provides much greater accuracy than the probe-based tools.