Towards an Accurate Bandwidth Estimation Tool for 802.11n Wireless Networks

Available Bandwidth Estimation (ABE) has been the subject of several studies in recent years. ABE tools based on the PRM approach have been the primary choice for production environments since they are more suitable for applications that are far from users (e.g., Cloud applications). At the same time, the 802.11n wireless network standard has been increasingly used in companies and homes. However, the 802.11n introduced significant changes to improve efficiency such as Frame Aggregation (FA) and the Channel Bonding (CB), which directly influence the accuracy of ABE techniques. Therefore, this work aims to provide a solution that reduces the impact of the new 802.11n arrangements in ABE. For this, we initially performed experiments on a testbed to assess how much these 802.11n MAC layer factors influence the estimates of four ABE tools: ASSOLO, PTR, PathChirp, and YAZ. Results showed that YAZ achieved the best efficiency. Nevertheless, its accuracy is significantly affected in FA-enabled scenarios. Then, we performed an indepth analysis of YAZ aiming to identify the causes of that problem. Finally, we proposed and evaluated the YAZ++, a tool that implements a new technique to improve ABE accuracy in FA scenarios. Results showed an improvement in the estimated values.