The growth of real-time, streaming media application traffic in the Internet presents a number of challenges because the real-time constraints and interactive nature of these applications render the use of TCP ineffective. These streaming media application flows are usually high bandwidth and long duration, which means they should utilize network congestion control to avoid congestion collapse and ensure fairness. The Internet Engineering Task Force (IETF) developed the Datagram Congestion Control Protocol (DCCP) to provide congestion control for these types of real-time applications. A major factor behind this effort was the desire to eliminate the duplication of effort and potential
for error resulting from each application implementing its own congestion control.
In this research, we examine the difference in network performance and video quality for a typical application, a high bandwidth video telephony client, when using DCCP instead of the default transport protocol. We discuss several challenges to porting an application to DCCP and then examine the impact of DCCP on network and application performance in both testbed and Internet environments. We show that DCCP responds to changing network conditions within a few round trip times and provides better fairness to other network traffic than typical real-time, streaming media congestion control methods. Given fair bandwidth allocation, DCCP provides equivalent or better video quality, as measured by Peak Signal to Noise Ratio (PSNR) and Structural Similarity (SSIM).