“The email from APNIC arrived on a Thursday: your final IPv4 /22 allocation has been issued. There were no more IPv4 addresses coming.”

In November 2012, Net4 India hit the IPv4 wall. We were onboarding hundreds of hosted VPS instances weekly, but our IPv4 pool was nearly exhausted.

We had to deploy Dual-Stack IPv6 across our core routers and customer aggregation layers immediately.

The Dual-Stack Architecture

Rather than relying on carrier-grade NAT, we chose to deploy native IPv6 BGP Peering.

Every network interface was configured with parallel protocol stacks:

• IPv4: Legacy /24 subnets for existing routing tables.
• IPv6: Modern /48 enterprise allocations and /64 per-customer subnets.

[!IMPORTANT] Running Dual-Stack doubles the size of neighbor tables and RIB/FIB memory utilization on edge routers. Upgrading router RAM was required before pushing full IPv6 tables.

BGP Dual-Stack Router Configuration

We configured multiprotocol BGP (MP-BGP) on our Cisco ASR 9000 edge routers to peer with upstream transit providers over IPv6.

router bgp 17465
 bgp router-id 121.240.1.1
 !
 ! # IPv4 Unicast Address Family
 address-family ipv4 unicast
  network 121.240.0.0 mask 255.255.0.0
 exit-address-family
 !
 ! # IPv6 Unicast Address Family (MP-BGP)
 address-family ipv6 unicast
  network 2402:7800::/32
  neighbor 2402:7800:0:1::1 activate
  neighbor 2402:7800:0:1::1 route-map IPV6_INBOUND in
 exit-address-family

[!NOTE] MP-BGP (Multiprotocol BGP): Allows BGP to carry routing information for multiple network layer protocols (IPv4, IPv6, VPNv4) over a single BGP peering session.

The Verdict

Key Takeaway

Plan Dual-Stack Address Maps Before Deployment.

Deploying IPv6 requires structured address planning (e.g. allocating /64 per VLAN and /48 per customer site). Clean prefix aggregation keeps routing tables small and efficient.