Category: VPLS

Inter-AS L2VPN MPLS Solutions

  • Option A:
    • each SP treats the other as CE
    • PW terminates as ASBR.
    • The link between ASBR is a AC instead of a PW
    • Granular QoS control between ASBR
  • Option B:
    • PE and P LSR do not learn remote PW endpoint addresses
    • Only the IP for PW endpoint (LER) that are ASBR leaked between ASs
    • T-LDP between ASBR of the SPs
    • PW terminates at ASBR. ASBR in turn build a new PW to its peer ASBR. This new PW connects the original PWs together
  • Option C:
    • PE loopback prefix and labels are exchange between two AS using MP-eBGP on ASBR
    • Instead of terminate the PW on its own ASBR, the PW is terminated on the remote PE on another AS. In another word, the PW is extended end to end between the two ASes
  • Summary:
    • All Inter-AS models (A,B,C) are supported to carry VPWS or VPLS PWs
    • IOS supports LDP for signaling, BGP for Auto-discovery (VPLS)
    • IOS XR supports both LDP and BGP signaling
    • Option B is not supported for BGP signal PWs
    • Per-PW QoS is not support
    • AC interworking is supported in IOS XR
    • L2VPN over TE or GRE is supported

VPLS Hightlights

What does it do?

  1. emulate ethernet network across WAN, mostly across MPLS, but can also be GRE over IP (this article assumes MPLS core)

What are the applications?

  1. For Enterprise, to provide Ethernet connectivity between sites over MPLS
  2. For Service Provider, to provide L2 multipoint ethernet service to customers using the same MPLS network the provides IPv4VPN service

How does it work?

Control Plane:

  1. Targeted LDP sessions (or BGP but NOT common) between PEs to exchange labels associated with each pseudo wire
  2. Support LDP and BGP (NOT common) signaling between PEs
  3. MAC learning: Dynamic learning based on source MAC and VLAN
  4. MAC aging: CAM entries aging timers are refreshed with incoming frames
  5. MAC withdrawer: MAC withdrawal upon topology changes
  6. Loop avoidance in the MPLS core: STP is disable by default, reply to Split Horizon and Full Mesh between PEs

Forwarding Plane:

  1. Ethernet frames are encapsulate with MPLS label and forward across MPLS core through Pseudo Wires within LSP
  2. Loop Prevention in the MPLS core: Split Horizon at the PE


  1. Support Ethernet Multipoint, VPLS network emulate an IEEE Ethernet switch
  2. Can forward Ethernet 802.3 frame and 802.1q VLAN frames
  3. Can forward selected VLAsN, or forward all frames (with or without 802.1q tag) within an Ethernet port
  4. Support IGMP snooping
  5. Can scale to a large MPLS core by using Hierarchical-VPLS
  6. Support MAC withdrawal using LDP Withdraw Message (or BGP NLRI Withdrawal) for fast L2 convergence to avoid black holing
  7. Support AutoDiscovery of new PE for a VFI using BGP Auto-Discovery (BGP advertise PE/VFI mapping)
  8. support dual connections to a site, but only in active-passive mode


  1. Unknown Unicast/Multicast/Broadcast are flood to all PW
  2. Full-messed among PEs in a VFI
  3. Does not support multi-path for a site (redundant active active connections), because MAC learning is on the data-plane.
  4. Need to manually setup LDP sessions between all PEs that are in the same VFI (Virtual Forwarding Instance). However if there is an existing LDP session, NO new session is needed.

VPWS auto-discovery and signaling

What is Auto-Discovery and Signaling

  • Auto-Discovery enables PE to discover other PE on the other end of the VPWS, or in the case of VPLS, in the same VFI
  • Signaling is the exchange of VC labels between PEs


  1. Auto-discovery options:
    1. Manual (most common)
    2. BGP
  2. Signaling Options:
    1. Static (no-signaling)
    2. LDP (most common)
    3. BGP


  1. Auto-discovery options:
    1. Manual (most common)
    2. BGP (use by large SP that add PE into VPLS frequently, or with large number of PEs)
  2. Signaling Options:
    1. Static (no-signaling)
    2. LDP (most common)
    3. BGP (label is sent in the same BGP updates that is used for Auto-Discovery)


What is H-VPLS?

Hierarchical VPLS (H-VPLS) is a VPLS design where the PE are no longer directly attached to CE. New VPLS access layer is introduced between the VPLS core and customers.

The PE routers that connect the core and access VPLS networks are called N-PE (Network PE).

The PE routers that connect to the customer is called U-PE (user PE)

Why do we need H-VPLS?

  1. To scale the VPLS network, the load of MAC learning and packet replication (multicast traffic) are off-load to C-PE.
  2. To scale the VPLS network by removing the need to have full meshed among PE. the C-PE is connected connected to a couple of N-PE. The N-PE themselves are still fully meshed. But it is a much smaller number compared to the original VPLS.

VPLS characteristics

A Ethernet switch has the following characteristics:

  1. forwarding of ethernet frames
  2. forwarding of unicast frames with an unknown destination MAC
  3. replication of broadcast and multicast frames to more than one port
  4. loop prevention
  5. dynamic MAC learning
  6. MAC aging

Since VPLS emulates an ethernet switch, it has the same characteristics.