0

Wont it make more sense in term of less congestion to send a multicast to just those interfaces that are not currently in its mac table?

Ariel Ferdman
  • 3
  • 1
  • 2
    Just a note that the unicast frame isn't broadcast in the sense of having a broadcast MAC address, it keeps its original destination MAC address and is sent out of every port. – jonathanjo Jul 20 '20 at 11:19
  • Just flooding that frame to linked ports without any MAC association won't work, as the destination may be connected to an adjacent switch with lots of MAC addresses already associated with its port. – Zac67 Jul 20 '20 at 13:32

2 Answers2

3

No, any interface could be connected to a neighboring hub/switch having more than one MAC address on it. Therefore, unknown unicast traffic is flooded to all ports except the origin port.

Jeff Wheeler
  • 5,439
  • 9
  • 18
  • 1
    Both this answer and the one i accepted are good answers which really helped and clarified what i was missing out with my mental model regarding switches roles in networks. each answer gives a different valid explanation and they are complementary. I chose the accepted answer just because of the length and depth of the explanation. Thanks to both of you. – Ariel Ferdman Jul 20 '20 at 13:58
  • @ArielFerdman, the accepted answer has invalid host MAC addresses as they are multicast addresses for Host B and Host C. Addresses with the I/G bit set are not allowed to be host source addresses, and, as destination addresses, they will be sent to every switch interface. – Ron Maupin Jul 21 '20 at 01:05
1

It wouldn't be ideal if the switch sends multicast to interfaces not currently in its switching table. The ideal thing is for the switch to flood the unknown unicast frame out all interfaces except the port on which the frame came from. The primary reason for this is that hosts might have been moved around or replaced with other hosts. Consider this, if for example in a switch's CAM table there are entries for Port F0/1 and F0/2 with Mac-addresses of 00:BB:CC:FF:11:22 and 11:33:CC:DD:22::FF respectively (host A and host B).

Supposing the network administrator removed host A from interface F0/1 to interface F0/3 and replaced interface F0/1 with host C (Mac-address 11:22:33:AA:CC:00), the switch would be ignorant of this change. Supposing the switch received a frame with destination Mac-address of 11:22:33:AA:CC:00, it would be making a grave mistake by sending a multicast out all other franes except the ones already in its table.

But when the switch floods the unknown unicast out all its interfaces (except the port on which it came from), it would learn that host A has moved from port F0/1 and that port F0/1 has been effectively replaced with host C. The switch then updates this entry in its CAM table. So as you can see, multicast frames will not be ideal in this situation.

PS. CAM table, Mac-address table, switching table etc etc all mean the same. I used the terms 'interface' and 'ports' interchangeably.

Ron Maupin
  • 98,218
  • 26
  • 115
  • 191
Kelvinskell
  • 42
  • 6
  • Your Host B and Host C MAC addresses are multicast MAC addresses, Any odd first byte has the I/G bit set, making it a group address. Those addresses are not allowed to be source addresses assigned to hosts. – Ron Maupin Jul 21 '20 at 00:58
  • I chose the Mac-addresses randomly, just for the sake of illustration. Thanks for pointing that out though. – Kelvinskell Jul 23 '20 at 02:22
  • The problem I saw is that you reference multicast addresses, but your example addresses are multicast addresses, and it is very confusing about what you really meant because of that. – Ron Maupin Jul 23 '20 at 02:26
  • "_it would be making a grave mistake by sending a multicast out all other franes except the ones already in its table._" That is exactly what will happen with any frames send to Host B or Host C using those MAC addresses because they _are_ multicast addresses. – Ron Maupin Jul 23 '20 at 02:28
  • See [this question](https://stackoverflow.com/q/58808705/3745413) and the answer for what really happens. – Ron Maupin Jul 23 '20 at 02:32