All things related to multicast

[Hugh LaMaster <>]

On Thu, 13 May 2004, John Zwiebel wrote:

> I agree with Tom that multicast isn't going to take off until
> there are high-BW applications.

Agreed.  We had a high-BW imaging application demonstrated 
in 1999 and did HDTV in 2001.  You need multicast to do this,
but, most people are not going to have 19-50 Mbps to the home
for a while.  If we have to wait for high-BW to the home
for multicast to be useful, we're going to keep waiting
for a while.

> Moving on, those high-BW applications are -NOT- going to use
> ASM.

Our high-BW applications *did* use ASM, and although 
they didn't work with PIM-DM, they did work with PIM-SM/MSDP.

ASM/sdr will not scale to billions of home users, but, 
since few are going to have high-BW to the home anyway,
I'm not certain that is a barrier.

My biggest question is, "What is impeding multicast on college/
university/research park campuses?"  Most of the objections to
doing high-BW multicast don't apply there, but, success has still
been limited.  Perhaps even grad students are too busy producing
PowerPoint slides to watch IETF and NANOG broadcasts.

> So, the barriers to multicast are:
>    -- last-mile
>       Where is AMT or how do we break the monopoly that the
>       home providers have.

This is going to take the longest to solve.

>    -- no source discovery protocol that doesn't depend on multicast
>       (ie, sdr doesn't scale to an interdomain environment and
>       continued attempts to make it do so will not work.  Storing
>       this information "in the network" by pushing it out via multicast
>       is OK in a dense environment, but the Internet is not a dense
>       environment.)  This has to be an application level protocol
>       like DNS (ie, putting the SDP file on a web page, although not
>       elegant, does work.  Perhaps a discussion of why this isn't
>       "good enough" is in order?)

I'm at a loss for why this hasn't taken off more.

>    -- MLDv2 and IGMPv3.  Its in KAME so "soon" will be in FreeBSD and
>       therefore OSX (I hope).  Perhaps a little lobbying of the FreeBSD
>       folks to incorporate the KAME IGMPv3 and MLDv2 would be in order?
>       (I don't know the process)
> 
> This forum has been wondering about the sasser attacks on MSDP.  SSM
> would eliminate this kind of an attack.  SSM does open up for a 
> "reverse"
> attack where a "evil-doer" could join every IP address in the world 
> causing
> creation of lots of state.

This is going to be a major issue.  Routers should have  
relatively small defaults for the number of groups 
allowable by default, and folks who need more can configure
it upward.  Something like 64 groups/512 routes by default
on an interface would be a nice limit.  Increase it if you
need to.  If the default is low, it would limit the damage
from a compromised host or incorrect network scan or whatever.

In the long run, this is going to be a big enough issue that
real management tools are going to have to exist.  We're going
to need more than a router log entry to understand if the
need for more groups is legitimate or just the result of a 
worm downstream.


> I'm often accused of being unrealistic because I keep saying "kill asm" 
> and
> "Kill MSDP".  Perhaps I should change my mantra to "allow it to wither".
> I keep seeing efforts (I think) wasted on making ASM better because we 
> have
> to support all those legacy hosts that don't support IGMPv3 so can't do
> SSM.  Yet, I sure don't see the explosion of multicast that was suppose 
> to
> happen because we maintained that backward capability.


We seem to have already saw the explosion, which peaked during 
2000-2001, and traffic and sources have declined severely since then.  



==============================================================================
 Hugh LaMaster, M/S 233-21,   Mail: 

  /~\ The ASCII  
 NASA Ames Research Center    Or:   

 \ / Ribbon Campaign
 Moffett Field, CA 94035-1000                               X  Against HTML 
 Phone: 650/604-1056          Disc: Unofficial *opinion*.  / \ Email!
==============================================================================