The following section shows how a SPBM network is established, in this case, a Layer 2 VSN.
Discover network topology
IS-IS runs on all nodes of the SPBM domain. Since IS-IS is the basis of SPBM, the IS-IS adjacency must be formed first. After the neighboring nodes see hellos from each other they look for the same Level (Level 1) and the same area (for example, Area 2f.8700.0000.00). After the hellos are confirmed both nodes send Link State Protocol Data Units, which contain connectivity information for the SPBM node. These nodes also send copies of all other LSPs they have in their databases. This establishes a network of connectivity providing the necessary information for each node to find the best and proper path to all destinations in the network.
Each node has a system ID, which is used in the topology announcement. This same System ID also serves as the switch Backbone MAC address (B-MAC), which is used as the source and destination MAC address in the SPBM network.
Each IS-IS node automatically builds trees from itself to all other nodes
When the network topology is discovered and stored in the IS-IS link state database (LSDB), each node calculates shortest path trees for each source node. A unicast path now exists from every node to every other node
With this information, each node populates unicast information received from SPBM into the FIB for forwarding purposes. Multicast FIB is not produced until Layer 2 VSN services are configured and learned.
IS-IS advertises new service communities of interest
When a new service is provisioned, its membership is flooded throughout the topology with an IS-IS advertisement.
BMAC and I-SID information is flooded throughout the network to announce new I-SID memberships. In this case, VLAN 20 is mapped to I-SID 100.
Note
I-SIDs are only used for virtual services (Layer 2 and Layer 3 VSNs). If IP Shortcuts only is enabled on the BEBs, I-SIDs are never exchanged in the network as IP Shortcuts allow for IP networks to be transported across IS-IS.
Each node populates its FDB with the BMAC information derived from the IS-IS shortest path tree calculations. Thus there is no traditional flooding and learning mechanism in place for the B-VLAN, but FDBs are programmed by the IS-IS protocol.
When a node receives notice of a new service AND is on the shortest path, it updates the FDB
In this scenario, where there are three source nodes having a membership on I-SID 100, there are three shortest path trees calculated (not counting the Equal Cost Trees (ECTs).
The paths between any two nodes are always the shortest paths. Also, the paths in either direction are congruent, thus a bidirectional communication stream can be monitored easily by mirroring ingress and egress on a link to a network analyzer.
VLAN traffic arriving on switch A and VLAN 20 is forwarded following the blue path, traffic arriving on switch B and VLAN 20 the orange path and on switch C VLAN 20 traffic is following the green path.
If the destination CMAC is unknown at the SPBM ingress node or the traffic is of type broadcast or multicast, then it is flooded to all members of the topology which spans VLAN 20. If the destination CMAC is already known, then the traffic is only forwarded as a unicast to the appropriate destination. In the SPBM domain, the traffic is switched on the BMAC header only. The bridge filtering database (FDB) at the VLAN to I-SID boundary (backbone edge bridge BEB), maintains a mapping between CMACs and corresponding BMACs.
For example, Switch B learns all CMACs which are on VLAN 20 connected to switch A with the BMAC of A in its FDB and the CMACs which are behind C are learned with the BMAC of C.