There are several methods for using BGP virtual private LAN service (VPLS) to scale VPLS in an existing LDP-VPLS network. These methods do not require changes on LDP-VPLS provider-edge (PE) routers, but rather enable VPLS to scale by using BGP VPLS and using LDP-BGP VPLS interworking. BGP VPLS can also extend the reach of VPLS from a single LDP-VPLS metro domain to the intermetro WAN.

By Mehul Mehta and Amit Shukla, Juniper Networks, Inc.
CommsDesign Nov 06, 2008

VPLS is a Layer 2 multipoint VPN that emulates LAN service across a WAN. VPLS enables service providers to interconnect several customer sites (each being a LAN segment) over a packet-switched network, effectively making all the customer LAN segments behave as one single LAN. A service provider's network appears as an Ethernet bridge to the service provider's customers using VPLS. With VPLS, no routing interaction occurs between the customer and service providers, and the customer can run any type of Layer 3 protocols between sites. The IETF Layer2 VPN Working Group has two VPLS standards: RFC 4761 and RFC 4762. Though they are almost identical approaches with respect to the VPLS forwarding plane, they are very different approaches to the VPLS control plane.

VPLS Control Plane Choices
The VPLS control plane has two primary functions: autodiscovery and signaling.

• Discovery refers to the process of finding all the PE routers that participate in a given VPLS instance. A PE router can be configured with the identities of all other PE routers in a given VPLS instance, or the PE router can use a protocol to automatically discover the other PE routers. This latter method is called autodiscovery.
• After discovery occurs, each pair of PE routers in a VPLS must be able to establish and tear down pseudowires to each other. This process is known as signaling. Signaling is also used to transmit certain characteristics of the pseudowire that a PE router sets up for a given VPLS.

BGP-VPLS Control Plane
The BGP-VPLS control plane defines a means for a PE router to know which remote PE routers are members of a given VPLS (autodiscovery), and for a PE router to know the pseudowire label expected by a given remote PE router for a given VPLS (signaling). The BGP NLRI contains enough information to provide the autodiscovery and signaling functions simultaneously.

As in the BGP scheme for Layer 2 and Layer 3 VPNs, on each PE router a route target is configured for each VPLS. The route target is the same for a particular VPLS across all PE routers and is used to identify the VPLS to which an incoming BGP message pertains. LDP-VPLS Control Plane
The LDP signaling scheme for VPLS is similar to the LDP scheme for point-to-point Layer 2 connections. LDP is used for signaling the pseudowires that are used to interconnect the VPLS instances of a given customer on the PE routers. In the absence of an autodiscovery mechanism, the identities of all the remote PE routers that are part of a VPLS instance must be configured manually on each PE router.

LDP VPLS defines the hierarchical VPLS (H-VPLS) scheme in which, instead of a PE router being fully meshed with LDP sessions, a two-level hierarchy is created involving hub and spoke PE routers. The hub PE routers are fully meshed with LDP sessions, whereas the spoke PE router has a pseudowire only to a single hub PE router or to multiple hub PE routers for redundancy. Spoke pseudowires can be implemented using any Layer 2 tunneling technology.

Scaling VPLS
Table 1 compares the scaling characteristics of LDP VPLS and BGP VPLS. These scaling characteristics can determine the scope of a VPLS deployment in the context of a metro network and beyond.

Table 1. Comparison of LDP-VPLS and BGP-VPLS Scaling Characteristics

Extending VPLS
Service providers are seeking mechanisms to extend the VPLS out of an autonomous system (AS).

Inter-AS
LDP VPLS faces the following challenges in providing VPLS that spans multiple ASs:

• Inter-AS LDP VPLS may require the setup of LDP sessions between PE routers that are in different ASs and potentially different administrative domains, or it may require the use of multisegment pseudowires, which has its own complexities.
• The globally significant 32-bit VCID used by LDP signaling requires operationally intensive manual coordination between ASs. BGP VPLS also requires site-identifier coordination between Ass.

In contrast, in BGP VPLS, exchange of control traffic between ASs is localized to AS border routers (the so-called option B) or route reflectors (option C), thus facilitating tight control over the information exchanged and such factors as peer authentication. In addition, the use of BGP communities and route target filtering further simplifies the task of determining which VPLS crosses the AS boundary (and to where), and which VPLS remains within the AS.