End-to-end congestion control over wireless networks : analysis and enhancements

The success of the Internet can be contributed to the large number of useful applications running on it. Basically, these applications can be classified into two categories -- bulk data transfer and streaming transfer -- based on the different requirements on the data transmission. As we know, bulk data transfer applications, such as file transfers and emails, usually require high reliability of data transmissions while delays or delay variations caused are less concerned. On the other hand, streaming applications, including real-time audio and video, are very sensitive to delays and delay variations, while certain rate of packet losses is acceptable. In order to provide the end-to-end congestion control for these two different types of applications, two end-to-end congestion control algorithms, namely TCP and TFRC, have been carefully designed over last years. To date, both algorithms work very well in wired networks. However, in wireless networks where random loss is rampant due to environmental noise, both TCP and TFRC suffer significant performance degradation unnecessarily, because they often misinterpret the random loss as the indication of network congestion.DOCTOR OF PHILOSOPHY (SCE