Hi Margaret, Chris & Juliusz,
Hi homenet list,

while I haven't completely caught up with discussion on the homenet list
--the routing protocol thread is truly (th|d)readful-- I seem unable to
find significant feedback on the multicast part (sec. 11) of the
comparison draft.  (I apologise & appreciate pointers if I missed it.)

The draft currently assumes IS-IS or Babel would/could be the multicast
routing protocol.  That assumption, IMHO, is wrong.  Long story short,
MOSPF was the archetype for trying to bolt multicast routing onto a
unicast routing protocol, and the best description I've heard of it is
"trainwreck."  The most established protocol for multicast routing would
be some variant of PIM, or for the hipsters you can try BIER.

However, this doesn't mean that multicast is irrelevant in the routing
protocol comparison.  The draft just needs to ask a different question:
does the unicast routing protocol provide the neccessary information for
the multicast routing protocol to perform its job?

What I have in the back of my mind here is a separate MRIB.  For the
non-multicast-savvy on this list, the MRIB is a separate routing table
containing routes to unicast prefixes, used to perform RPF lookups to
locate paths towards multicast sources/senders.

At a large fully multicast-enabled ISP, the MRIB would be an identical
copy of the unicast table/URIB.  But the point is that it's calculated
separately, so that different metrics can be used or links can be
excluded.  I strongly believe this is useful in the homenet, namely to
deprioritise wifi links - or other future media that sucks at multicast.

The concept at work here is that metric/cost for a link isn't the same
for unicast or multicast.  802.11ac with a wide channel and enough
antennas provides great unicast performance - better than a 100MBit
Ethernet link.  Yet, I would never use that wifi link for any kind of
multicast distribution.  This means I want a low unicast metric on the
wifi link, but a (very) high multicast metric.

Hence, with IS-IS, I would put into use a separate multicast routing
topology (topology ID #4 is reserved for this.)  Routers would use the
knowledge of their interfaces and link status to come up with a pair of
unicast and multicast metrics, instead of just one metric.

Babel does not - correct me if I'm wrong - have the capability to work
with more than one topology.

(Also note that a MRIB will never contain source-destination routes.
SADR is simply not supported for multicast RPF lookups;  there is only
the sender address and the multicast group address.)

(I'm blissfully ignorant about BIER for now, yet google tells me there
is some IS-IS integration for it.)

Cheers,


-David


P.S.: I've seen 802.11 multicast rate being mentioned as 1-2MBit/s in some
e-mails.  There is some misconception involved there.  1-2 MBit/s is what
you get on current Linux 802.11 AP/IBSS implementations (= home
routers).  Nothing precludes the AP from using higher rates, but this
requires code that Linux doesn't currently have.

Getting to full-rate multicast requires snooping for group memberships,
determining the lowest rate used by any of the receivers, and optionally
calculating whether it might be better to send unicast.  "Enterprise"
wifi implementations have these features to varying extent in the
controller.  (2 big vendors have quite extensive features there, 1 big
vendor has almost nothing at all - guess it depends on whether some big
customer asked for it at some point.)

P.P.S.: That also means that a better AP implementation would have less
of a difference between the unicast metric and the multicast metric for
wifi links.  I'd say they should still be different.

_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to