Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [RE] Stream identification at the MAC SAP



At 15:04 13/11/2004 -0700, Shvodian William-r63101 wrote:
>While it won't take care of attempts to send more real-time traffic than a link supports, the work from the Congestion Management group should enable throttling back of the file transfers and non-real time traffic so that those applications don't disturb the real-time video applications.

However, the task of distinguishing between real-time and non-real-time traffic is non-trivial. I guess the easiest way would be to require all real-time traffic to have 802.1Q tags with a high priority, and require all tagged non-real-time frames to have a lower priority.


>By the way, there must be a better way to handle fast forward than to send 2, 4, 8, 16 X the traffic across the net, especially since much of it will get thrown away anyway.  Otherwise, I don't see how any QoS scheme prevents you from going over your allocated bandwidth, without serious over-allocation for fast forward.  If you have a 100 Mbps link and you try to send an 8X HD signal without loss you are out of luck.

If you're fast forwarding at, say, 8x speed then in the case of uncompressed video it'd be easy to just send every 8th frame to get back to the same data rate as normal video. But then you'll need around 270 Mbit/s anyway. In the case of MPEG2 the frames you drop will be mostly B-frames (which are small) and you have to keep the I-frames (which are large); a broadcast MPEG2 stream is around 4 Mbit/s while I-frame-only needs around 20 Mbit/s. To get back to the same data rate as normal video would require re-encoding (at a lower quality, because there will be less in common between adjacent frames, though I don't imagine the users will mind that too much), which adds complexity to the player.

If the QoS mechanism allocates a fixed bandwidth for a stream, the player will have to do whatever is needed to fit within that bandwidth. Otherwise its designer will take the easy way out, because that works most of the time.


John Grant
   ___  ___  ___  ___    ___  ___  ___  ___  ___
  |   ||   ||   ||   |  |   ||   ||   ||   ||   |
  | N || i || n || e |  | T || i || l || e || s |
  |___||___||___||___|  |___||___||___||___||___|

Nine Tiles Networks Ltd, Cambridge, England
+44 1223 862599 and +44 1223 511455
http://www.ninetiles.com