Consider the following bridged LAN:
This LAN contains cycles, which must be removed before data nodes start sending frames. The spanning tree algorithm results in the following LAN:
Two links have been "removed"; these are the link from port 0 of bridge 86 to port 1 of bridge 98, and the link from port 1 of bridge 86 to port 3 of bridge 61. Neither of these links were on any bridge's shortest path to the root, and so are redundant.
Suppose the tables start out empty, H0 sends to H1, and then H1 sends to H0. Here are the forwarding table contents after H0 sends to H1:
B14:
|
Destination |
Next Hop |
|
H0 |
3 |
B61:
|
Destination |
Next Hop |
|
H0 |
2 |
B73:
|
Destination |
Next Hop |
|
H0 |
2 |
B86:
|
Destination |
Next Hop |
|
H0 |
3 |
B98:
|
Destination |
Next Hop |
|
H0 |
3 |
Now, the situation after H1 replies to H0:
B14:
|
Destination |
Next Hop |
|
H0 |
3 |
|
H1 |
2 |
B61:
|
Destination |
Next Hop |
|
H0 |
2 |
B73:
|
Destination |
Next Hop |
|
H0 |
2 |
B86:
|
Destination |
Next Hop |
|
H0 |
3 |
B98:
|
Destination |
Next Hop |
|
H0 |
3 |
Notes:
The destinations are MAC addresses, so saying, e.g., H0 is shorthand for "H0's MAC address."
The next hops are port numbers.
When H0 sends, none of the bridges know where H1 is, so the transmission floods the network. If no one knows where the destination is, the frame goes everywhere, maximizing the chance of successful delivery.
When H1 sends, B14 already knows where H0 is, and so forwards directly to H0. B14 also learns the whereabouts of H1. No other bridges learn H1's whereabouts.
No bridges appear in any other bridge forwarding table. These are transparent bridges.