Connection of IPv6 Islands Across IPv4 Clouds with BGP Overview

If you have not upgraded your core to IPv6, you can still provide IPv6 services to customers by connecting remote IPv6 islands across IPv4 clouds by means of MP-BGP and MPLS. An IPv6 island is a network employing IPv6 addressing, such as a customer site. The IPv4 cloud consists of the PE–P–PE core.

Note: You must configure an IPv6 interface in the parent VR for this feature to work.

Consider Figure 112. Each customer site is connected by means of a CE router to a PE router. The PE routers in this implementation are referred to as dual-stack BGP (DS-BGP) routers because they run both the IPv6 and IPv4 protocol stack.

Figure 112: IPv6 Tunneled over MPLS-IPv4

IPv6 Tunneled over MPLS-IPv4

The PE routers learn IPv6 routes using MP-BGP over TCPv4 or TCPv6 from the CE devices. Alternatively, you can configure IPv6 static routes on the PE routers to reach the customer IPv6 networks through the CE IPv6 link. You can use any IPv6-enabled routing protocol to access the CE routers.

Use any MPLS signaling protocol to establish an MPLS base tunnel in the IPv4 core network. Each PE router runs MP-BGP over an IPv4 stack (MP-BGP/TCP/IPv4). MP-BGP advertises the customer IPv6 routes by exchanging IPv6 NLRI reachability information across the IPv4 cloud.

Each PE router announces the IPv4 address of its core-facing interface (the tunnel endpoint) to its PE peers as the BGP next hop. Because MP-BGP requires the next hop to be in the same address family as the NLRI, the IPv4 next-hop address must be embedded in an IPv6 format. The PE router advertises the IPv6 routes as labeled routes and an IPv6 next hop.

In the topology shown in Figure 112, OSPF advertises reachability of the loopback ( and and core-facing ( and interfaces of the PE routers. LDP binds label L1 to on the P router.

Router CE 1 establishes an MP-BGP session over TCPv4 to PE 1 and advertises its ability to reach the IPv6 network 2001:0430::/32. The MP-BGP update message specifies an AFI value of 2 (IPv6) and a SAFI value of 1 (unicast). As the next hop in the MP-REACH-NLRI attribute, CE 1 advertises the IPv6 address of the CE 1 interface that links to PE 1.

Both IPv4 and IPv6 addresses must be configured on the PE-CE link. The IPv6 address defaults to an IPv4-compatible address that can be overridden with policy.

PE 1 and PE 2 establish an MP-BGP session using their remote loopback IPv4 addresses as neighbor addresses. Router PE 1 installs in its IPv6 global routing table the route advertised by CE 1. MP-BGP on PE 1 then binds a second-level label, L2, and advertises the route to PE 2 with an AFI value of 2 (IPv6) and a SAFI value of 4 (labeled routes). The next hop that PE 1 advertises in the MP-REACH-NLRI attribute is the IPv4 address of its loopback interface,, encoded in IPv6 format as ::

When MP-BGP on router PE 2 receives the advertisement, it associates the base tunnel (to, label L1) with the next hop (:: that was advertised by PE 1 to reach the customer IPv6 island, 2001:0430::/32. Router PE 2 then uses MP-BGP (AFI = 2, SAFI = 1) to advertise to CE 2 its ability to reach this network.

CE 2 sends native IPv6 packets destined for the 2001:0430::/32 network to PE 2. On receipt, PE 2 performs a lookup in its global IPv6 routing table. PE 2 prepends two labels to the IPv6 header (L1–L2–IPv6) and then forwards the packet out its core-facing interface (

The P router does a lookup on L1 and label switches the packet toward PE 1. The P router can either replace L1 with another label or pop L1 if PE 1 requested PHP.

When PE 1 receives the packet on its core-facing interface, it pops all the labels and does a lookup in the global IPv6 routing table using the destination address in the IPv6 header. PE 1 then forwards the native IPv6 packet out to CE 1 on the IPv6 link.

Related Documentation