FW: Multicast tunnel to the AG from my home over a cable modem

[Ivan R. Judson]

Bill,

This is an interesting perspective.  What would be your answer from the
networking side of things?

--Ivan

..........
Ivan R. Judson .~. http://www.mcs.anl.gov/~judson
Futures Laboratory .~.  630 252 0920
Argonne National Laboratory .~. 630 252 6424 Fax
 

> -----Original Message-----
> From: Denis DeLaRoca [mailto:delaroca at ucla.edu] 
> Sent: Thursday, August 15, 2002 5:41 PM
> To: Ivan R. Judson; Larry Rowe
> Subject: RE: Multicast tunnel to the AG from my home over a 
> cable modem
> 
> 
> 
> On Sat, 10 Aug 2002, Ivan R. Judson wrote:
> 
> > I've been meaning to implement a free version of UMTP for awhile, 
> > since this would alleviate some issues we have on the AG with the 
> > current bridge software we have now.  I've got the latest RFC from 
> > Live's website, and was wondering the same thing -- Why 
> didn't think 
> > progress farther?
> 
> Well, Larry's report on UMTP was so *positive* that it is 
> tempting to go spend the $500 needed to license the 
> client/server pieces from Ross Finlayson! I have also been 
> mulling over Ivan's proposition to do a UMTP implementation 
> ourselves...
> 
> But is it worth doing such implementation? UMTP is 
> essentially dead: it didn't advance to RFC status on the 
> IETF, nor is anybody pursuing R&D with it?
> 
> I just been chatting with Sanjay and CMU's ESM sounds 
> promising -- see below for his recap of ESM. Perhaps ESM 
> could give us a better solution for our simple problem to 
> push our mcast trafic the last mile to our remote offices/labs?
> 
> What do you think, Larry?
> 
> -- Denis
> 
> --------------------------------------------------------------
> ------------
> Date: Thu, 15 Aug 2002 10:38:34 -0400
> From: sanjay+ at cs.cmu.edu
> To: Denis DeLaRoca <delaroca at ucla.edu>
> Subject: Re: Live Broadcast of Sigcomm 2002 using End System Multicast
> 
> 
> Hi Denis,
> 
>         ESM is an alternate architecture to IP Multicast that 
> does not require router level support. The idea here is that 
> end systems participating in a group "organize" themselves 
> into an efficient overlay tree automatically. Each link of 
> the overlay corresponds to a regular unicast path on the Internet.
> 
> We have argued for this ESM architecture in papers that 
> appeared in Sigmetrics 2000, Sigcomm 2001 - we have a 
> combined journal paper that is to appear in IEEE JSAC this 
> year. If you have time I recommend checking the journal 
> paper, and the ppt presentation at Sigmetrics. For details: 
http://www.cs.cmu.edu/~narada/


Our arguments are that:
-ubiquitous IP Multicast support does not look likely in the forseeable
future -IP Multicast violates the stateless architectural principles of
the IP layer -Support for higher level functionality like reliablility
and congestion  control is very hard.

There is a considerable momentum building towards this kind of
"overlay/application-layer" multicast in the research/Sigcomm community
(See the links to related research on the page).

The jury is still out on ESM, but we hope by broadcasting real events,
we will gain valuable experience that will shape the future of this
architecture.