We are looking to use Backhauls and CMM Micros to connect one main office and 10 branch offices together. We have routers at each office and each office is on a different subnet. Is there anything in the Backhauls or CMM Micros to keep broadcast storms from happening within the Canopy? I realize that the routers will stop all traffic that is not intended for its office, but I am worried about an over abundance of denied packets at each router and unnecessary network traffic.
Ethernet broadcasts are stopped on both sides of each router. If you separate each office from the Canopy network with a router, the only Ethernet broadcasts the Canopy units should see are the ARP broadcasts between WAN ports on the routers. This traffic is minimal.
Let me rephrase my question. If a packet is destined for branch 1 from the main office, will it be sent to each branch or just branch 1? If it is just sent to branch 1, what in the system handles that?
You’re correct. All the Canopy SMs will “see” all the traffic destined for all the other SMs. The AP essentially “broadcasts” all traffic to all the SMs. Each SM then discards the traffic intended for other SMs.
This is the nature of point-to-multipoint radio links. Just like the old 10Base2 and 10Based5 coax, and 10BaseT using hubs, the medium is shared. For 10Base2 the coax is shared; for AP/SM Canopy systems the RF link is shared.
Ethernet switches (technically, self-learning Ethernet bridges) get around this limitation by dedicating a single port, and its connected cable segment, to a single host (PC) or a reduced number of hosts. They are able to dedicate the port by learning the MAC address of the PC at the other end of the cable, and sending that PC’s unicast (host-to-host) traffic – as opposed to broadcast traffic – only out that single port.
Canopy units are also self-learning Ethernet bridges; look at the AP’s “Bridge Table” under “Expanded Stats”. This allows the Canopy system to only permit the unicast traffic destined for a remote host to pass through the RF link. Unicast traffic between PCs in a single office won’t travel the RF link.
There is, however, only a single RF link in an AP/SM group. This RF link can be equated to a single cable segment in a wired network.
You're correct. All the Canopy SMs will "see" all the traffic destined for all the other SMs. The AP essentially "broadcasts" all traffic to all the SMs. Each SM then discards the traffic intended for other SMs.
I notice that you have the words "see" and "broadcast" in quotes. I am guessing that you have these in quotes because it is not real broadcast traffic and could not be detected by a packet sniffer.
To my knowledge of the Canopy architecture and Layer-2 switching, when a packet arrives at an access point, the AP Bridge List is checked to make sure that it contains the MAC address of the destination MAC address of the packet. At this point, I believe is where the "broadcast" not not really a "broadcast" is sent.
TCP/IP uses an Ethernet broadcast for ARP (IP) to find the Ethernet MAC associated with an IP address for the host adapters to use for subsequent unicast traffic.
I was confirming your understanding of the nature of the Canopy traffic when I said the AP, in a sense, “broadcasts” both broadcast and unicast packets, and the SMs “see” but disregard these packets.
A packet sniffer passively (without transmitting a response) detects all the traffic – both broadcast and unicast – present at its interface. You’ll have a difficult time sniffing the Canopy traffic, however, because you don’t have a promiscuous Canopy interface. (Promiscuous in the sense that the interface passes all traffic to the sniffer software; this is the accepted industry term, though it seems contradictory for something to be both passive and promiscuous.)