Feb 1st, 2011: Bye bye IPv4

Today, the IANA has allocated the last two remaining IPv4 address blocks to APNIC, the registry for the Asia/Pacific region. This means that the remaining five blocks will be distributed according to the “Global Policy for the Allocation of the Remaining IPv4 Address Space” evenly amongst the registries. And we’re out.

IPv4 will remain forever

This does not mean that we have NO IPv4 addresses, there are plenty. And the addresses you have won’t be taken away from you. You can still get addresses from RIPE, ARIN and the other registries if you really need them. The clue is in the last part of the previous sentence. You have to prove that you need them and that you will start using them immediately. If you get IPv4 addresses for your services from a hosting provider, getting new addresses will get expensive soon.

IPv6 is the Internet growth path

IPv4 can no longer handle the Internet growth. We will have to start using IPv6 for new services and applications, and new Internet connections. Otherwise, everyone will be behind large NAT servers. Not good for applications, not good for users and not good for performance.

What to do?

The panic is not in the network, your servers and services will continue to run. The panic is in the lack of knowledge and experience of IPv6 in your team. You need to put pressure in starting to work with IPv6 on all fronts, implementing it as a natural part of every project. Regardless if it’s about cloud services, unified communication, web sites, business support systems or e-mail. IPv6 needs to be there and you need everyone to accept that it’s a network just as natural as the IPv4 networks they are used to and know how to work with.

Dual stack solutions – good or bad?

When you start learning about IPv6 you can find a lot of old information out there. The protocol has been around more than 15 years and have changed. In those days, the concept of computers with dual stacks was a common thing. We used a PC-LAN network stack (IPx, netbeui, Arcnet, 10net)  alongside the TCP/IP stack. And there was still a lot of IPv4 addresses around. The concept of dual stacks – one with IPv4 and one with IPv6 was launched as a natural way to migrate. After all, network engineers was used to handling dual stacks.

The world has changed. Network engineers have used single stacks for a long time. And the old dual stack concept was not shared by the same applications. In this dual stack world, with two IP stacks, applications are supposed to use both stacks, which causes a whole new set of issues. And we’re running out of IPv4 addresses, so we can’t assume that all new hosts will get proper addresses.

I strongly think we have to start adding hosts with IPv6 only, and start preparing the architecture for living with two networks. In many cases, an IPv6 only network with the support of NAT64 is easier to handle than a dual stack network. I think we’re beyond the point where dual stacks is the solution for all computers that we will attach to the network.

IPv6 – from snail-pace to VIP lane

The good thing is that even with dual stacks, IPv6 will be the better choice. Today, IPv6 is often implemented with tunnelling solutions, which means that connections have lower performance over IPv6. As we move along and implement non-tunnelled IPv6 in the networks, users will notice that IPv4 will be hidden behind multiple layers of NAT and IPv6 will be the faster network to use. Applications should make the preference a user configurable option, not something hidden in the application.

To summarize:

• The global pool of free IPv4 addresses has been distributed to the registries today, making the global pool empty
• IPv4 will remain, but will not handle Internet growth
• You need to make IPv6 part of regular work, not a special project
• Use dual stack for public services, but do not assume it for every host on the Internet

Today is the beginning of the next phase of the Internet. Don’t miss this phase and lock yourself and your organization into the old network.

/Olle E. Johansson

Comments