You can configure the JUNOS software to tunnel IPv6 over an MPLS-based IPv4 network. This configuration allows you to interconnect a number of smaller IPv6 networks over an IPv4-based network core, giving you the ability to provide IPv6 service without having to upgrade the routers in your core network. Multiprotocol BGP (MP-BGP) is configured to exchange routes between the IPv6 networks, and data is tunneled between these IPv6 networks by means of IPv4-based MPLS.
In Figure 19, Routers PE1 and PE2 are dual-stack BGP routers, meaning they have both IPv4 and IPv6 stacks. The PE routers link the IPv6 networks through the customer edge (CE) routers to the IPv4 core network. The CE routers and the PE routers connect through a link layer that can carry IPv6 traffic. The PE routers use IPv6 on the CE router-facing interfaces and use IPv4 and MPLS on the core-facing interfaces. Note that one of the connected IPv6 networks could be the global IPv6 Internet.
Figure 19: IPv6 Networks Linked by MPLS IPv4 Tunnels
The two PE routers are linked through a MP-BGP session using IPv4 addresses. They use the session to exchange IPv6 routes with an IPv6 (value 2) address family indicator (AFI) and a Subsequent AFI (SAFI) (value 4). Each PE router sets the next hop for the IPv6 routes advertised on this session to its own IPv4 address. Because MP-BGP requires the BGP next hop to correspond to the same address family as the network layer reachability information (NLRI), this IPv4 address needs to be embedded within an IPv6 format.
The PE routers can learn the IPv6 routes from the CE routers connected to them by means of the routing protocols RIP next generation (RIPng) or MP-BGP, or through static configuration. Note that if BGP is used as the PE-router-to-CE-router protocol, the MP-BGP session between the PE router and CE router could occur over an IPv4 or IPv6 Transmission Control Protocol (TCP) session. Also, the BGP routes exchanged on that session would have SAFI unicast. You must configure an export policy to pass routes between IBGP and EBGP, and between BGP and any other protocol.
The PE routers have MPLS LSPs routed to each others’ IPv4 addresses. IPv4 provides signaling for the LSPs by means of either LDP or RSVP. These LSPs are used to resolve the next-hop addresses of the IPv6 routes learned from MP-BGP. The next hops use IPv4-mapped IPv6 addresses, while the LSPs use IPv4 addresses.
The PE routers always advertise IPv6 routes to each other using a label value of 2, the explicit null label for IPv6 as defined in RFC 3032, MPLS Label Stack Encoding. As a consequence, each of the forwarding next hops for the IPv6 routes learned from remote PE routers normally push two labels. The inner label is 2 (this label could be different if the advertising PE router is not a Juniper Networks routing platform), and the outer label is the LSP label. If the LSP is a single-hop LSP, then only label 2 is pushed.
It is also possible for the PE routers to exchange plain IPv6 routes using SAFI unicast. However, there is one major advantage in exchanging labeled IPv6 routes. The penultimate-hop router for an MPLS LSP can pop the outer label and then send the packet with the inner label as an MPLS packet. Without the inner label, the penultimate-hop router would need to discover whether the packet is an IPv4 or IPv6 packet to set the protocol field in the Layer 2 header correctly.
When the PE1 router in Figure 19 receives an IPv6 packet from the CE1 router, it performs a lookup in the IPv6 forwarding table. If the destination matches a prefix learned from the CE2 router, then no labels need to be pushed and the packet is simply sent to the CE2 router. If the destination matches a prefix that was learned from the PE2 router, then the PE1 router pushes two labels onto the packet and sends it to the provider router. The inner label is 2 and the outer label is the LSP label for the PE2 router.
Each provider router in the service provider’s network handles the packet as it would any MPLS packet, swapping labels as it passes from provider router to provider router. The penultimate-hop provider router for the LSP pops the outer label and sends the packet to the PE2 router. When the PE2 router receives the packet, it recognizes the IPv6 explicit null label on the packet (Label 2). It pops this label and treats it as an IPv6 packet, performing a lookup in the IPv6 forwarding table and forwarding the packet to the CE3 router.
This section discusses the following topics:
Detailed information about the Juniper Networks implementation of IPv6 over MPLS is described in the following Internet drafts:
These Internet drafts are available on the IETF Web site at http://www.ietf.org/.
You must perform the following tasks to allow IPv6 to be carried over an IPv4 MPLS tunnel:
In addition to configuring the family inet6 statement on all the CE router–facing interfaces, you must also configure the statement on all the core-facing interfaces running MPLS. Both configurations are necessary because the router must be able to process any IPv6 packets it receives on these interfaces. You should not see any regular IPv6 traffic arrive on these interfaces, but you will receive MPLS packets tagged with label 2. Even though label 2 MPLS packets are sent in IPv4, these packets are treated as native IPv6 packets.
Include the family inet6 statement:
You can include this statement at the following hierarchy levels:
For information about how to configure MPLS and RSVP, see the following sections:
You enable IPv6 tunneling by including the ipv6-tunneling statement in the configuration for the PE routers. This statement allows IPv6 routes to be resolved over an MPLS network by converting all routes stored in the inet.3 routing table to IPv4-compatible IPv6 addresses and then copying them into the inet6.3 routing table. This routing table can be used to resolve next hops for both inet6 and inet6-vpn routes.
To configure IPv6 tunneling, include the ipv6-tunneling statement on the PE routers:
You can include this statement at the following hierarchy levels:
You also need to configure IPv6 tunneling when you configure IPv6 VPNs. For more information, see the JUNOS VPNs Configuration Guide.
When you configure multiprotocol BGP to carry IPv6 traffic, the IPv4 MPLS label is removed at the destination PE router. The remaining IPv6 packet without a label can then be forwarded to the IPv6 network. Include the explicit-null statement:
You can include this statement at the following hierarchy levels: