Cell Relay Archive

[albert.e.manfredi@boeing.com]

In article <3507EB02.6AF754E1@End.Of.Message>,
  Milan <Email@End.Of.Message> wrote:
>
>     I just wonder if the bandwidth allocation scheme of ATM is efficient
> enough for transmitting VBR-compressed video. From some of the research
> papers I've read, an ATM connection is in general characterized by a
> <PCR, SCR, BT> tuple. However, as VBR-compressed video is itself
> periodic and also highly bursty in terms of bit rate, I really doubt if
> this three-parameter only tuple can handle this periodic and bursty
> nature effectively. I would be glad if someone could give me an insight.

First a clarification: PCR/SCR/BT/CLR/CDV are QoS parameters which
characterize a VBR channel, not ATM in general. ATM can be used in many
different ways, each channel type represented by a different set of QoS
'tuples. For example, a CBR channel requires one only to specify PCR, CLR,
and CDV (I believe, assuming I forgot nothing).

The idea of VBR was precisely something like MPEG video, because it is
bursty, requires high peak rates, but hopefully would not require a CBR
channel that wide, as that would be prohibitively expensive. I mean, wasn't
that the whole idea, back in the days when people were still saying "BISDN"?

I suppose that with a pre-recorded movie one would have a pretty good idea as
to how to set those QoS parameters, whereas with a live feed one would not.
But all that means is that if you're going to try and be cheap about it, you
will occasionally drop some video frames, if you don't get past the ATM
traffic cops.

In my head, at least, compressed video would be a VBR candidate.

But your point about whether or not the simple arithmetic involved in those
QoS parameters is adequate to real-world examples is a fundamental one that
arm waving alone can't resolve. I suppose the best arm waving can conclude is
that the simple arithmetic is probably going to be approximately as effective
as any other method short of magic, in this case (I'm assuming a mix of VBR
MPEG channels, channels unsynchronized among one other). For example, a
"talking heads" channel would be manifested by short Max Burst Size (MBS) and
a big difference between PCR and SCR. Whereas a live sports feed would
require large MBS and perhaps not so much gap between PCR and SCR, which more
closely approximates CBR.

I can see where it would make good economic sense to mix in several of those
Charlie Rose kind of programs with few sports shows, over a restricted
bandwidth.

Bert
manfredi@arl.bna.boeing.com

-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/   Now offering spam-free web-based newsreading