All things related to multicast

[Alan Crosswell <>]

John Zwiebel wrote:
> We need to have a standard directory that lists multicast sources.
> We need to move to SSM.

I'm not so sure.  I think the WMT/Real approach of trying multicast and falling 
back to unicast works much better.  Other than network geeks who run SDR to see 
"what's on" the rest of the world has figured out that they find sources on web 
pages.  This argues that SSM is easy since SDP isn't really necessary.
> Shep will say we need AMT, I agree, but feel that without the two
> requirements I've listed, AMT isn't going to be useful.
>
> So Richard, you are the application vendor, where's your "Multicast 
> Source Directory Protocol"?
> (groan, another MSDP.  OK, its SSM, let's call them multicast channels, 
> CDP, Channel Directory
> Protocol.  groan, not another CDP.  OK...)
>
> What efforts is vBrick making toward transition from IGMPv2 to IGMPv3?
> The switch folks described GMRP/GARP, can vBrick take advantage of it?
> (Not that the deployed L2 switches can make use of it, but...)
>
> What efforts are being made to get L2 switches to snoop IGMPv3
> and MLDv2 and are these efforts orthogonal or complementary to GMRP/GARP?

I can tell you that IGMPv3 while "supported" by CGMP doesn't work right -- they 
haven't implemented Leave snooping so an IGMPv3 channel surfer will quickly clog 
the subnet while waiting for soft state timers to expire.  In fact, it's not 
clear to me if the MAC forwarding entries ever get cleaned up.

I think it is broken from a strict layering approach for L2 switches to 
implement (and have to track changes in) L3 protocols.  Just as we got IGMPv2 
working, along comes IGMPv3.  And then, MLD!  We run v6 here too and forget 
about any of the switches we just installed yesterday every supporting v6 
multicast, much less the 3-year old switches that do CGMP.

Isn't it the case though that GMRP/GARP require host stack support as well?  I 
need to learn more about these L2 protocols.


> FWIW, each group is pointing fingers at every other group.  We're all
> right.  How do we get past this?  It would seem that the IETF isn't the
> right place for it because the problem isn't just an L3 problem.
>
>
> Richard you said:
>
> > I think it would be great if someone in this group could 
> > create/sponsor a multicast IMPLEMENTATION FAQ targeted at members IT 
> > staff. We don't need a "what is multicast" but a "how to" FAQ.
http://multicast.internet2.edu has a bunch of links including to the Abilene 
multicast cookbook.  There's a really good inexpensive 2.5 day hands-on 
multicast workshop offered by I2 on a regular basis.  Most of the instructors 
are really good (one just fakes it mostly but they let me come along anyway 
because I'm good at coiling cables;-).

> There are several of these already out there.  It doesn't help that I can't
> point you to any of them.  Nor does it help that none of them can 
> provide answers
> for questions like Alan's L2-switch problem.  There is no answer so it 
> isn't
> mentioned.  If I were an IT guy (I was once -- used it to launch 
> rockets) I'd
> read the FAQ, try configuring it, and then I'd find my L2-switch would hose
> my unicast and spend the next three months trying to figure out why, and 
> I'd
> finally give up as "more intelligent heads" ordered me to turn it off.
>
> Alan is smarter than I am, he figured it out without having to be 
> ordered.  ;-)

Not so smart.  It's still turned on.  In fact, I'm always sourcing some content. 
 Just check SDR.  One day I'll get SSM working.

/a