Improving the performance of wireless networks using frame aggregation and rate adaptation
MetadataShow full item record
As the data rates supported by the physical layer increase, overheads increasingly dominate the throughput of wireless networks. A promising approach for reducing overheads is to group a number of frames together into one transmission. This can reduce the impact of overheads by sharing headers and the time spent waiting to gain access to the transmission floor. Traditional aggregation schemes require that frames that are aggregated all be destined to the same receiver. These approaches neglect the fact that transmissions are broadcast and a single transmission will potentially be received by many receivers. Thus, by taking advantage of the broadcast nature of wireless transmissions, overheads can be amortized over more data and achieve more performance gain. To show this, we design a series of MAC-based aggregation protocols that take advantage of rate adaptation and the broadcast nature of wireless transmissions. We first show the design of a system that can aggregate both unicast and broadcast frames. Further, the system can classify TCP ACK segments so that they can be aggregated with TCP data flowing in the opposite direction. Second, we develop a rate-adaptive frame aggregation scheme that allows us to find the best aggregation size by tracking the size based on received data frames and the data rate chosen by rate adaptation. Third, we develop a multi-destination frame aggregation scheme to aggregate broadcast frames and unicast frames that are destined for different receivers using delayed ACKs. Using a delayed ACK scheme allows multiple receivers to control transmission time of the ACKs. Finally, we extend multi-destination rate-adaptive frame aggregation to allow piggybacking of various types of metadata with user packets. This promises to lower the impact of metadata-based control protocols on data transport. A novel aspect of our work is that we implement and validate the designs not through simulation, but rather using our wireless node prototype, Hydra, which supports a high performance PHY based on 802.11n. To validate our designs, we conduct extensive experiments both on real and emulator-based channels and measure system performance.