When to play
When to play
The actual playout time is not a function of the arrival time, only of the end-to-end delay which can be calculated as shown below:
Figure adapted from slide 11 on page 6 of Kevin Jeffay, “Lecture 9: Networking Performance of Multimedia Delivery on the Internet Today”, Lecture notes for COMP 249: Advanced Distributed Systems Multimedia, Dept. of CS, Univ. Of North Carolina at Chapel Hill, November 9, 1999. http://www.cs.odu.edu/~cs778/jeffay/Lecture9.pdf [Jeffay 1999]
Slide Notes
Kevin Jeffay, “Lecture 9: Networking Performance of Multimedia Delivery on the Internet Today”, Lecture notes for COMP 249: Advanced Distributed Systems Multimedia, University of North Carolina at Chapel Hill, Department of Computer Science, November 9, 1999. http://www.cs.odu.edu/~cs778/jeffay/Lecture9.pdf Links to an external site.
Transcript
[slide86] So if we look at the grand scheme of things, what do we see? Well, we see an NTP clock, right? Everyone knows what NTP is, the Network Time Protocol. So we can have a very, very precise clock that's typically controlled by an atomic clock. From that, we can derive the RTP clock. So we can relate global time to our RTP timestamps. We now send our messages, they arrive at the receiver, and now we play them out. So now we can actually write the playout time, the sample generation time, the local clock synchronization correction, because we have to time align, plus the sender packaging delay, plus the network delay, plus the jitter buffer delay. Those pieces end up contributing all to the end-to-end latency. This is really pretty cool, because now we can think about, how do we go and exploit those different terms?