As the number of PEs in the VPLS grows, the replication
overhead also grows. Because PEs are generally located at the edges of the network,
they may have limited bandwidth available for replicating traffic.
?– A full mesh of MPLS LSPs, Ethernet pseudowires, and targeted LDP sessions are
required between all the PEs in the VPLS. If RSVP-TE LSPs are used, this may
severely constrain VPLS scaling (see ???MPLS Scaling,??? earlier in the chapter).
The Hierarchical VPLS (or H-VPLS) model addresses these problems by segmenting
the VPLS service into a core of ???hub??? devices that are fully meshed, as in a standard
VPLS, and a number of ???spoke??? devices that are each connected through a single ???hub.???
The hub devices are generally known as Network-facing PEs (N-PEs), and the spoke
devices as User-facing PEs (U-PEs).
In RFC 4762, the N-PEs are referred to as ???PE-rs??? devices, since they must be PE
devices that are capable of routing and switching. The U-PEs are known as ???MTU-s???
devices since they are often located at Multi-Tenant Units and are capable of switching,
but do not need to be capable of routing as they have only a single path into the
network.
Pages:
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011