A snooping rate adaptation algorithm for IEEE 802.11 WLANs

The IEEE 802.11 physical layer (PHY) supports multiple transmission rates with varying modulation and channel coding scheme. This feature allows the 802.11 wireless LAN (WLAN) to protect the throughput performance in face of dynamically changing channel condition. For instance, 802.11a standard has 8 PHY "modes" or rates. The procedure of selecting the best among the provided rates is called the rate adaptation. In this paper, we propose and implement at the device driver level a MAC snooping approach that overcomes the short-coming of the ACK-based approaches, which is robust against collisions. It uses information readily available on the 802.11 device driver level, in particular the number of disparate and active transmitters and the number of transmissions in the last time window. We do not use PHY measurements such as signal strength or SNR due to the weak correlation issue. Our experiments with the prototype implementation show that it is much more robust to congestion than existing algorithms, in particular the MIT SampleRate algorithm available in MADWiFi. We use the example of 802.11a in this paper, but the idea directly applies to faster 802.11 networks without any additional overhead.