Auto-Selecting the Rate Adaptation Algorithm in Wireless Networks

Determining an appropriate data rate is an important problem in wireless LAN. A high data rate could suffer from packet losses in a low-quality link. On the other hand, a low data rate is more resilient to the link’s quality degradation, but it fails to achieve a high throughput in a high-quality link. Therefore, an IEEE 802.11 device is usually equipped with a rate adaptation algorithm (RCA) to find the best rate. Traditionally, the RCAs do not take the application requirement into the consideration. In this paper, we consider the problem of finding the most suitable data rate based on both the link condition and application requirement. To this end, we have proposed an RCA auto-selection algorithm which can automatically select the best RCA from the publicly available ARF, ONOE, SampleRate and fixed-rate algorithms. The algorithm is based on a wireless link classification and the application in use. We have implemented the algorithm in the Madwifi driver and also a visualization tool in Linux. Our test-bed measurement results have shown that any existing RCA cannot always be the best choice under various combinations of link condition and applications. As a result, the RCA auto-selection algorithm could help identify the best RCA under various scenarios.