The H-VPLS model also enables devices that are unable to learn MAC addresses
to participate in VPLS as U-PEs (known, in this case, as ???PE-r??? devices, as they are
capable of routing but not of switching). A PE-r has one pseudowire to its serving
N-PE for each attachment circuit (since it is unable to bridge between those attachment
circuits). Each pseudowire is signalled with a unique PW ID (in the FEC 128 case), to
enable the N-PE to distinguish between them.
The split horizon rule is now modified so that an N-PE applies split horizon only on
the pseudowires to other N-PEs and not on pseudowires to U-PEs.
When a U-PE sends a broadcast, multicast, or unknown frame, it sends a single
copy to its serving N-PE. Unless the N-PE knows the destination MAC address, it will
flood the frame on all pseudowires to other N-PEs or served U-PEs and to any local
attachment circuits. When a remote N-PE receives the frame, it will flood it (unless it
knows the destination MAC address) on all pesudowires to served U-PEs and on any
local attachment circuits. Thus, the replication load is distributed across the ingress
and egress N-PEs.
One potential issue in the H-VPLS model is that if an N-PE fails, then its served U-PEs
will be isolated from the rest of the network.
Pages:
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012