History-aware rate adaptation in 802.11 wireless networks

Rate adaptation (RA) is a mechanism unspecified by the 802.11 standards, yet critical to the system performance. Although many different design directions have been studied the past years, there are still little insights learned of how short-term channel's past performance can be utilized to limit transmissions at low throughput rates. In this paper, we conduct a systematic experimental study to expose the importance of history aware rate adaptation and explore new techniques to address this space. To this end, we design and implement HA-RRAA, a new robust RA algorithm which uses short-term loss ratio to opportunistically guide its rate selection, a cost-effective, adaptive RTS filter to prevent collision losses from triggering rate decrease and an adaptive probe time window to limit excessive probing at high lossy rates. Our experimental results show gains up to 63% of HA-RRAA over RRAA, RRAA+, SampleRate and ARF, in realistic field trials.