Improving the performance and efficiency of wireless networks using rate adaptation

Recent years have seen a staggering increase in the deployment and utilization of wireless networks. More and more devices are being equipped with Wireless LAN (WLAN) cards to take advantage of the omnipresence of WLAN networks. Therefore, it has become necessary that the protocols used by WLANs are efficient and provide good performance. Rate Adaptation protocols are an important mechanism employed by WLANs to improve network performance. This dissertation develops three complementary techniques, which use rate adaptation to optimize and improve performance by i) performing rate adaptation to optimize energy consumption, ii) developing a more accurate technique to predict the frame delivery ratio that is used by rate adaptation protocols, and iii) jointly optimizing rate adaptation with data retransmission to maximize throughput. More specifically, in i), we use extensive measurements to develop a simple yet accurate energy consumption model for 802.11n wireless cards. We use the model to drive the design of an energy aware rate adaptation scheme. A major benefit of a model-based rate adaptation is that applying a model allows us to eliminate frequent probes required in many existing rate adaptation schemes. In ii), we find that the accuracy of existing delivery ratio calculation techniques is still limited due to bursty errors inherent to the wireless channel. We develop a new method for computing packet delivery rate that captures the burstiness of errors. Furthermore, we propose a new data interleaving technique, which leverages our framework to reduce the burstiness of errors, thereby improving frame delivery ratio. Finally, in iii), we address the susceptibility of wireless networks to transmission failures due to dynamic channel conditions and unpredictable interference. To efficiently recover from failures, we propose a retransmission scheme where the receiver combines information received from multiple failed transmissions associated with the same frame. The protocol has two distinguishing features. First, it simultaneously supports partial retransmission and combines bits with low confidence. Second, it jointly optimizes the data rate of the retransmission and the information to be retransmitted to maximize throughput.