T-RATE: A Framework for the Trace-Driven Evaluation of 802.11 Rate Adaptation Algorithms

Rate adaptation algorithms (RAAs) in 802.11 networks can have a significant impact on network throughput. In order to attempt to maximize throughput, these algorithms dynamically adapt to inferred changes in the channel being used. Evaluating and comparing rate adaptation algorithms is extremely challenging due to the variability of channel conditions. Recent work on developing and evaluating new RAAs has used traces to increase the realism of simulators. We propose a new framework for the trace-driven evaluation of RAAs (called T-RATE) in which we collect and process traces in environments where interference is caused by WiFi and non-WiFi devices. T-RATE enables a more comprehensive evaluation of RAAs in environments that are representative of those in which 802.11 devices are actually used. A key to our approach is that we capture, in traces, information that is relevant to RAAs regarding how channel conditions affect channel access and channel error rates. Our approach minimizes the use of wireless channel models and achieves highly realistic results under a wider variety of scenarios than previously possible. Our evaluation of T-RATE demonstrates it can be used to collect and process traces to accurately evaluate a variety of RAAs under more diverse and representative channel conditions than previously possible.