It also centralises VPLS functionality in the PE in contrast
to some designs where MAC learning has been separated from MPLS signalling or
designs such as ATM LANE where dedicated servers have been used for replication.
Each PE maintains a separate MAC FIB for each VPLS instance??”known as a Virtual
Switching Instance or VSI??”where the destination for each MAC address may be either
a local attachment circuit or a pseudowire to a remote PE. Since there is only one VSI
per VPLS instance, and since MAC addresses are unique per VSI, there is typically only
one VLAN mapped to each VPLS. This is known as qualified learning and is an implementation
of ???Independent VLAN Learning??? as per 802.1Q. If multiple VLANs are
mapped to one VPLS, then the MAC addresses may not overlap between the VLANs.
This is known as unqualified learning and is an implementation of ???Shared VLAN
Learning??? as per 802.1Q.
One key decision made in the LDP VPLS design was to reuse the Ethernet pseudowire
encapsulations without modification. This was not without controversy. The benefit
of using the existing encapsulation, of course, is that no changes need to be made
to pseudowire forwarding for VPLS.
Pages:
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003