Time to dump dual-stack networks and get on the IPv6 train – with LW4o6
Deutsche Telekom and others go with subscriber-focused lightweight approach
Despite a decade of efforts, the rollout of IPv6 is still stubbornly sat at less than 25 per cent, in terms of internet traffic, with recent reports suggesting adoption may actually be leveling off.
Among the reasons the world is not shifting en masse to the new addressing system, the most persistent is the fact that people use dual-stack networks to run both IPv4 and IPv6 at the same time in parallel.
It's an approach that has enabled organizations to make the best use of their existing network structures albeit with some extra administration. However, that "temporary" solution has itself become a barrier to wider IPv6 adoption because it effectively locks existing IPv4 networks in place: to move to IPv6 you still need to go through the whole, painful transition that leads people to decide to go with dual-stack in the first place.
Well, several organizations – including Deutsche Telekom – believe they have found the solution in the new-ish lw4o6 standard, standing for lightweight 4-over-6. It effectively encapsulates your local IPv4 traffic in IPv6 packets, and beams them over your shiny new IPv6-only network.
This means, for example, a broadband subscriber can use IPv4 in their home, and open connections over the internet using IPv6-only carrier and backbone links via an lw4o6-supporting home gateway.
Passed as a standard three years ago (RFC 7596), lw4o6 is being developed into a real world solution that can run on today's hardware, and it's open source, so there are few barriers to use. In some respects, lw4o6 is itself a dual-stack solution but with a critically different implementation that enables you to run it on an IPv6-only network, greatly assisting in an eventual move to the new protocol.
It is based on Dual-Stack Lite (RFC 6333) – which also allows you to run IPv4 over an IPv6 network – but it removes a key scalability problem with Dual-Stack Lite by shifting network address translation (NAT) from the carrier's network to a router on the customer's premises.
I did NAT see that coming
To make this work, lw4o6 uses A+P (Address + Port) sharing where each piece of customer equipment gets an IPv4 and IPv6 address and a port range. The NAT function then maps the customer's private address to a public address and port within the provided range.
The carrier keeps a table that maps the A+P (IPv4 address and port) with the IPv6 address and only translates back, or decapsulates, the IPv4/6 packet if there is a match. The downside to this approach is the creation of yet another large mapping table – the eternal IPv4 problem – but the upside is that it not only scales but also puts everything on an IPv6 foundation. It also allows for tunneling which means that lots of applications that break with parallel systems should work.
Encapsulated ... A broadband subscriber's IPv4 traffic carried over IPv6 via lw4o6 using A+P sharing (Source: APNIC)
While there are a range of new translation techniques aimed at weaning people off running IPv4 and IPv6 in parallel, one of the advantages to lw4o6 is that is has the backing of a big telco: Deutsche Telekom is using it in its cloud-based broadband system TeraStream.
Despite having been announced in 2012, TeraStream is finally coming online and offering 1Gbps residential broadband, meaning that it could help provide an IPv6 tipping point. Juniper is also backing the standard.
If you're interested in hearing about real-world implementation of the system, a recent presentation at regional internet registry RIPE by developer Diego Pino Garcia provides plenty of good information. And Garcia has today written a blog post for another RIR, APNIC, which alerted us to the implementation of this new standard.
In short, it's still early days, however, systems like this may be what finally shift the internet from IPv4 resignation to finally embracing IPv6. ®