If the customer network generates a large number of
MAC addresses (tens or hundreds of thousands), delays in MAC learning and relearning after a topology
change can have potentially service-affecting impact.
In a large scale or inter-domain deployment scenario, the potential risk is twofold:
1. The FDB size on the PE nodes could be exceeded causing all the unknown destination packets to
flood.
2. When a change of topology occurs, such as a node or pseudowire failure, a MAC flush is propagated
throughout the network, hence massive learning must be performed at once on the core PE nodes.
The MAC learning rate on the Alcatel-Lucent 7750 SR is approximately 7000 MACs per second, which
means a large FDB (many tens of thousands of MACs) could take several seconds to be learned. While
the MACs are being learned by the CPM and downloaded to the IOM FDBs, unknown traffic is flooded.
In addition, if a very restrictive rate limit for unknown traffic is applied at the ingress SAPS, some
traffic could be dropped.
FDB for the Alcatel-Lucent 7750 SR can support over
100,000 MAC addresses.
If the FDB capacity is exceeded, flooding of unknown
packets results.
Time to relearn tens of thousands of addresses may result
in excessive flooding or loss of packets.
—It could take several seconds to relearn these MAC
addresses.
Provider Backbone Bridging (PBB) is also known as MAC-in-MAC, since the customer MAC addresses are
effectively encapsulated using the provider (B-MAC) addresses.
