Re: [AVT] RTCP unicast feedback for measurement applications?

[Eve Schooler <[e-mail]>]

Fiandra,

Finally, here are pointers to the paper that extends the 
ideas proposed in the I-D, and that I promised below.
  
The paper analyzes the bandwidth savings of summarisation,
and presents a couple additional techniques (biasing and
hierarchical summarisation) that 
- combine well with unicast feedback and generic summarisation, 
- provide further rtcp scalability, and 
- offer additional application contexts for rtcp.


http://www.research.att.com/~schooler/papers/rtcp-summary-paper.techreport.ps

http://www.research.att.com/~schooler/papers/rtcp-summary-paper.techreport.pdf
  
A shorter version will appear in the 2nd IEEE International 
Symposium on Network Computing and Applications, NCA-03, 
April 16-18, 2003, Cambridge, MA.
 
 http://www.research.att.com/~schooler/papers/rtcp-summary-paper.ps
 http://www.research.att.com/~schooler/papers/rtcp-summary-paper.pdf

E.

> -----Original Message-----
> From: Eve Schooler [mailto:[e-mail]]
> Sent: martedì 4 febbraio 2003 2.02
> To: Fiandra Riccardo
> Cc: Eve Schooler; Julian Chesterfield; Joerg Ott; [e-mail]
> Subject: Re: [AVT] RTCP unicast feedback for measurement applications?
> 
> 
> Fiandra Riccardo <[e-mail]> wrote:
> 
> >I was interested in the "RTCP Extensions for Single-Source Multicast
> >Sessions with Unicast Feedback" draft.
> >Actually, I was wondering if it is also targeted to have a very large and
> >distribuited network measurement framework.
> >Where multicast receivers (thousands) use RTCP to send report to a
> >measurement collector about packet loss, jitter, or other.
> >I see really many advantages on that for e2e measurements, especially in a
> >scenario where thousands of receivers are always connected to the network
> >and cannot send multicast (common scenario for an operator, even without
> >SSM...).
> >
> >My questions are:
> >-Was this a scenario of application when writing the draft?
> 
> Absolutely, though the primary focus of the draft was to solve
> the problem posed by the removal/restriction of the backchannel
> caused by SSM and other uni-directional or asymmetric topologies.
> Solving the feedback problem for SSM gives us the blueprint for
> other unicast-based feedback architectures, including ones for
> network measurement in general.
> 
> >-In this case, could I use the RTCP channel multicasted to the receivers to
> >control them?
> 
> Yes you could.  This is certainly what we envision in the RTCP
> context; use the multicast channel in the source-to-receiver
> direction, and use the unicast backchannels in the
> receiver-to-source direction.
> 
> We also propose to use the multicast channel either:
> - to reflect receiver feedback (received on the unicast backchannels)
>   back out on the multicast channel to reach all receivers, or
> - to redistribute to the receivers feedback summaries
>   (receiver feedback collected at the source and aggregated into
>   a single mathematical distribution) back out on the multicast channel.
> 
> >-Do you see any extension possible also for ptp rtp connections (forward
> >rtcp reports to another device?)
> 
> In section 10 of draft-ietf-avt-rtcpssm-02.txt, we use
> an extension to SDP to indicate the device to which the
> feedback gets reported.  This is done following the proposal
> for SDP source filters documented in
> draft-ietf-mmusic-sdp-srcfilter-00.txt.
> 
> We also just completed a paper that expands upon the ideas
> in the I-D and that should provide further application
> scenarios.  I'll post it here shortly.
> 
> Eve
> 
> -=-=-=-=-
> 
> Eve M. Schooler                 Voice:  1-650-330-7913
> AT&T Labs - Research            FAX:    [phone]
> 75 Willow Road                  E-mail: [e-mail]
> Menlo Park, CA, USA 94025       http://www.research.att.com/~schooler
_______________________________________________
Audio/Video Transport Working Group
[e-mail]
https://www1.ietf.org/mailman/listinfo/avt