Abstract
The Internet has already become a platform for multimedia rich communication, collaboration and entertaining applications. A majority of these applications are based on a technology known as audio-video streaming. Streaming conceals the effect of inherent file transmission delay by starting playback before the file transmission is complete. However, the user experience of streaming is not always satisfactory. This is because of the fact that good quality streaming requires adequate and steady end-to-end bandwidth which is often unavailable in many parts of the Internet. Adaptation is one of the means by which streaming applications try to overcome the inherent limitations of the Internet. Not only the Internet lacks mechanisms for minimum bandwidth, delay and loss guarantees, but also the available transport protocols on the Internet provides little support to adaptive streaming applications. In this dissertation, the effect of the available transport protocols on the performance of adaptive streaming media has been illustrated and mechanisms has been proposed to address these limitations. A new framework called the Dynamic Buffer Active Tuning (DBAT) has been proposed to support the adaptability of streaming applications by providing fine grained feedback, preferential packet drop, smooth rate control and automatic buffer tuning. This dissertation also introduces another framework named Clear Water Streaming (CWS) to support Variable Bit Rate (VBR) streaming flows. CWS incorporates router mechanisms to overcome the barrier of the conservative nature of TCP's congestion control, a major obstacle for VBR streaming. Finally, this dissertation proposes New Slow Start (NSS) which is a modification to slow start for providing an improved start up phase by reducing packet losses and buffering delay.