I was going through the IEEE 802.1Q (2018) standard on my own. The standard speaks about group addresses, and how these group addresses are learned by transit bridges, which is all fine. But how do these bridges register/subscribe/join these multicast MAC addresses? I understand there is IGMP which is a layer 3 protocol for devices to join a multicast address. But I would like to know if it is possible for joining in layer 2 itself. The standard does not go into depth about this.

So my understanding is:

  1. Management defines the devices (static) registered in a group OR
  2. There is a layer 2 protocol for dynamic registration OR
  3. There is no layer 2 protocol, and one needs to use a layer 3 protocol for this.

802.1q is just 4byte layer2 tag added to the ethernet frame in order to create virtual separation of physical LAN (VLAN). I am not quite sure there are any multicast groups used with relation to 802.1q.

Switches/Bridges don't join "Multicast MAC addresses" - they may however take part in Layer3 multicast routing & forwarding. IGMP can be used to join multicast group, but it takes alot more configuration on the network to make sure multicast traffic is propagated throughout. Many network protocols also run multicast for protocol discovery purposes (hello messages - e.g. PIM, OSPF, EIGRP, etc). Once traffic is sent to multicast address, it's layer2 destination MAC will be changed to multicast address too, which will be derived from the Layer3 group address Network Access and Layer 2 Multicast

Typically with multicast is always about Layer3. Of course, there are specifics you should be aware of when it comes to multicast traffic forwarded on Layer2, some of which @Zac67 has explained above. In addition, network bridges (switches) can provide some optimisation (aka IGMP snooping) with regards to BUM traffic (Broadcast, Unknown, Multicast) and thus forward Multicast traffic only on interfaces where interested party is registered (typically via IGMP).

  • Although, traffic destined to any address in the link-local ( multicast range is not subject to being blocked by IGMP snooping; switches must forward such traffic to all other interfaces.
    – Ron Maupin
    May 28 at 12:27

L2 multicast addresses can only be learned by bridges/switches through IGMP snooping. Since these addresses are never used as source addresses, they're not learned by the bridges directly.

Without multicast control (with a dumb switch or an unconfigured managed switch), multicast addresses are simply unknown to the bridges (never been used as source) and accordingly they're flooded to all ports but the one they came from, just like a broadcast.

  • So why would a bridge choose to use protocols like MMRP and ISIS-SPB for learning group addresses? Apr 28 at 7:30
  • A bridge uses the protocol as it is defined. Many protocols on top of L2 use multicast because it's the best way to propagate link-level information (some multicast address ranges are limited to link level). You definitely can't use broadcast there, or unicast since there's no true two-way communication.
    – Zac67
    Apr 28 at 7:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.