Definitive MPLS Network Designs [Electronic resources] نسخه متنی

IPv6 Internet Access Service Design

Several large customers in Kingland needed a service to interconnect IPv6 networks and for IPv6 Internet access. For example, Kingland's national academic and research network runs regional IPv6 networks and therefore required high-speed IPv6 interconnection across these networks, along with IPv6 Internet access. In response to this demand, TK launched an IPv6 Internet access service.

Because it is absolutely essential to protect the stability of the MPC, TK decided to use the IPv6 Provider Edge (6PE) approach described in the section "Deploying IPv6 Over an MPLS Network" in Chapter 1. This allowed TK to introduce an IPv6 service incrementally on the edge without any software upgrade or configuration change in the core, which remains purely IPv4/MPLS.

Because the number of IPv6 services is fairly small today and requires pure IPv6 connectivity only, TK decided to support the IPv6 services on dedicated 6PE routers. This isolates the mPE routers supporting IPv4 Internet and Layer 3 MPLS VPN services from any faults or problems that could occur with the IPv6 and 6PE technologies. These newer technologies have been subjected to less-extensive testing and validation, and TK has less production experience with them. 6PE routers have been located in every Level 1 POP as well as in four Level 2 POPs where customers also requested IPv6 services. Two 6PE routers are deployed in each Level 1 POP to support dual-homing of IPv6 customer sites. This represents a total of 16 6PE routers.

As illustrated in Figure 4-39, 6PE routers are dual-attached to the two P routers in the POP. They are generally attached with STM-1 PoS interfaces, but a few of them are attached with STM-16 PoS interfaces so as to support very high-speed IPv6 customer connections.

TK uses a separate set of route reflectors dedicated to the IPv6 Internet service. This is for the same reasons TK elected to use separate sets of route reflectors for IPv4 Internet and for Layer 3 MPLS VPN. Namely, doing so avoids contention across the different services in terms of control-plane convergence in network failure situations. Because the number of 6PE routers involved in the IPv6 route reflection is currently quite small, TK deployed only two IPv6 route reflectors. They are located in two different Level 1 POPs (Center POP and South POP). These are also the POPs that support native IPv6 peering to the IPv6 Internet.

The IPv6 route reflectors contain the full IPv6 Internet routing table, which currently contains about 1000 routes.

As the number of IPv6 Internet access services grows and operational experience is developed, TK plans to migrate the IPv6 Internet service onto the shared mPE routers in line with its overall network evolution strategy.

For routing between the 6PE router and the IPv6 CE routers, TK uses static routes as well as BGP.

TK uses a configuration template for the 6PE routers. It is shown in Example 4-13.

Example 4-13. 6PE Router Configuration Template

! interface to Site Y of Customer X
interface POSx/y
no ip address
ipv6 address IPv6-address-of-interface
ipv6 enable
!
router bgp 32764
!peering with first v6 RR
neighbor IPv4-address-of-first-v6-RR remote-as AS-number-of-TK
neighbor IPv4-address-of-first-v6-RR update-source Loopback0
no neighbor IPv4-address-of-first-v6-RR activate
!peering with second v6 RR
neighbor IPv4-address-of-second-v6-RR remote-as AS-number-of-TK
neighbor IPv4-address-of-second-v6-RR update-source Loopback0
no neighbor IPv4-address-of-second-v6-RR activate
!v6 native peering with customer
neighbor IPv6-address-of-CE remote-as AS-number-of-v6-customer
!
address-family ipv6
neighbor IPv4-address-of-first-v6-RR activate
neighbor IPv4-address-of-first-v6-RR send-label
neighbor IPv4-address-of-second-v6-RR activate
neighbor IPv4-address-of-second-v6-RR send-label
neighbor IPv6-address-of-CE activate
redistribute connected
no synchronization
exit-address-family

TK offers the same QoS support for the IPv6 Internet access service as for the IPv4 Internet access service. A single CoS is supported on the access links; it is identical to the Standard CoS of the Layer 3 MPLS VPN service. In the core, the IPv6 Internet traffic is handled together with the IPv4 Internet traffic and the Layer 3 MPLS VPN Standard CoS. To that end, TK takes advantage of the default behavior of the IOS 6PE implementation, which sets to 0 the EXP field of the label stack entries pushed by the ingress 6PE router. This ensures that the IPv6 traffic receives the desired treatment in the MPC (namely, that it is scheduled in the DF queue) without having to modify the QoS core service policies described in the "Quality of Service Design" section. Note that because the 6PE approach uses two labels, use of Penultimate Hop Popping