12:45 PM, March 15, 2001
Gates Building, Room 104

TCP-RTM: Using TCP for Real Time Applications

Sam Liang,
Computer Science Department,
Stanford University

About the talk:

Traditional wisdom has that TCP is not appropriate for real-time applications, because TCP favors reliability over timeliness. The result of this is a proliferation of application-specific real-time transport protocols, none of which has achieved the maturity and wide acceptance of TCP. In this talk, we investigate the basic question: can we make any minimal changes to TCP to make it suitable for real-time applications, either interactive or unidirectional playback? Our answer is yes. We propose a very simple feature to add to TCP, called Real-Time Mode (RTM), and we have implemented it in a Linux kernel. Using such a modified Linux kernel, our emulation-based experiments show that TCP-RTM, along with some other application-level techniques, such as application-level heartbeats and application-level framing, offers good performance for transmitting real-time data. It reduces both application perceived loss rate and data jitter for a flow over a lossy network connection. In addition, TCP-RTM provides congestion control for real-time applications so that they share the network resources fairly with other applications, and this elegantly solves the common problem of not being TCP-friendly of most existing real-time applications.

About the speaker:

Sam Liang is currently a Ph.D candidate of the Department of Electrical Engineering. He holds a Master's degree in Computer Science from the University of Arizona, and a Master's degree in Electrical Engineering from Stanford University. Prior to coming to Stanford, he worked in Silicon Graphics Inc as a Member of the Technical Staff and in LSI Logic as a Senior Software Engineer.

For more information:

Power point slides.