For years, the transition to IPv6 has been viewed as a gradual process, typically using dual-stack approaches. However, in certain environments such as modern data centers, the option of moving directly to an IPv6-only network is gaining traction. Why opt for this approach? What does this decision entail in practice? What technical challenges does it involve?
These were some of the questions discussed during a recent webinar organized by LACNIC and led by Tomás Lynch, a network engineer at Vultr with extensive experience in large-scale deployments.
Why IPv6-Only?
(Free access, no subscription required)
In data centers where the stack and services can be fully controlled, IPv6-only simplifies the network, eliminates reliance on IPv4 addresses, and reduces operational complexity. As Lynch pointed out, maintaining a dual-stack environment requires managing two protocols, duplicating configurations, and dealing with issues such as traffic asymmetry and diagnostic challenges.
Furthermore, today it’s no longer necessary to wait for all traffic to be natively IPv6: tools such as NAT64, DNS64, and 464XLAT allow translating requests in the case of services that don’t yet offer IPv6 connectivity, thus facilitating a functional transition.
Lynch stressed that dual stack is a transition, not a destination. “The ultimate goal should be IPv6-only networks, as this simplifies network design, security, and overall operations,” he added.
In data centers where the stack and services can be fully controlled, IPv6-only simplifies the network, eliminates reliance on IPv4 addresses, and reduces operational complexity. As Lynch pointed out, maintaining a dual-stack environment requires managing two protocols, duplicating configurations, and dealing with issues such as traffic asymmetry and diagnostic challenges.
Furthermore, today it’s no longer necessary to wait for all traffic to be natively IPv6: tools such as NAT64, DNS64, and 464XLAT allow translating requests in the case of services that don’t yet offer IPv6 connectivity, thus facilitating a functional transition.
Lynch stressed that dual stack is a transition, not a destination. “The ultimate goal should be IPv6-only networks, as this simplifies network design, security, and overall operations,” he added.
Tools, Best Practices, and Lessons Learned
Multiple implementation strategies were discussed during the webinar, focusing on tools and operational considerations:
Differences between ISP and data center networks
Leaf-spine architecture
Routing options and why we shouldn’t be afraid of BGP
Unnumbered BGP sessions
Sizing a GPU cluster
RFC5549 and SIIT-DC, NAT64/DNS64
The impact of using IPv4 addresses on each network element was also examined, considering that the use of IPv6-only can simplify configurations, provisioning, monitoring tools, and observability.
More and more organizations are adopting this IPv6-only architecture in segments of their data centers, especially those where they have full control over applications and infrastructure.
Throughout the presentation, Lynch shared real-life examples of how global organizations are adopting IPv6-only in parts of their infrastructure, particularly in internal areas of their data centers.
“The key is to stop thinking about IPv6 in terms of IPv4,” the expert noted, “so we can utilize the full potential of the different types of IPv6 addresses available, such as ULA and link-local, among others.”
IPv6-Only: A Possible and Desirable Future
The final message was clear: IPv6 only is a viable, desirable, and increasingly common technical option in modern network design. For many operators, it represents a more direct, simple, and sustainable path towards the future of the Internet.
We invite you to watch a recording of the webinar.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of LACNIC.
