In my opinion there are actually two IPv4 long tails, the first is the plethora of home devices ( PCs, gaming consoles, home routers, web cams, wifi cameras, Internet TVs ) which are, for the most part, IPv4-only. The second long tail is on the content side, only 0.15% of the top 1 million web sites are available over IPv6. Of course this will change but it will be some time, possibly decades, before all content is available over IPv6. It is abundantly clear that IPv4 and IPv6 must coexist as the Internet expands.
At Juniper, we have a pragmatic approach to address this diversity. At the end of the day, this is not about making a choice between IPv6 or IPv4 but the coexistence of IPv6 and IPv4. This is about growing the Internet.
For this reason Juniper is focusing on a continuum of solutions to help our customers evolve their networks. For today’s blog entry I’ll spend some time discussing the following three options:
1. IPv4 reclamation
2. IPv4 address sharing
3. IPv6 as a substrate layer
IPv4 reclamation: Network administrators can squeeze more efficiency out of their assigned IPv4 address pools. This might be as simple as re-numbering some internal devices with private addresses instead of public addresses, or migrating Internal spaces to IPv6 thereby freeing-up IPv4 addresses. It is much easier to do this when we are dealing with a closed network using a very limited set of applications. A variation of this is to leverage the ‘transfer’ policies enacted by all the RIRs over the last two years. If you know a network that is not using some of their address space, you can get this space ‘transferred’ to you. You might have to ‘compensate’ that network for their efforts in this process. It will be interesting to see how the free market values IP addresses!
IPv4 address sharing: The technology to accomplish this is very well known; NAT. It has been around and deployed in small and large scale environments. Most households for example are connected to the Internet through a home gateway using NAT. Most wireless service providers have NAT in place. Try this little app “Free IP address” on your iPhone and you can observe this in action. Using the application you will see two addresses, an IPv4 local address and an IPv4 global address. Carrier Grade NATs’ or ‘large scale NATs’ or ‘CGNat’ (many names for the same thing, a box sharing IPv4 addresses) will be deployed by wireline residential service providers as well. Resulting in a solution called double NAT, NAT at the edge of the customer network and within the service provider network.
On the commercial service provider side, new business customers will receive smaller pools of IPv4 addresses. Instead of a /24 or /26, they will now get a /28 or /29, even a single /32. In turn those business customers will leverage NAT at their exit point more aggressively than they already do today.
What scaling numbers can we get for these NAT boxes? Studies on port consumptions have shown that on average a residential customer is using only a few ports at peak time. This contrasts heavily with the maximum number of ports a user can use at any given time, which can be in the order of several thousand! If you dynamically allocate ports, you can set NAT for average and not for the peaks, and essentially have thousands of users sharing the same IP address! If you want headroom, make this only a hundred. So what does this mean in practice? Today’s residential service provider can turn a /16 class B address, (65,000 IPv4 addresses) into a resource that can be shared by 6.5 million users.
IPv6 as a substrate layer: By creating a foundation network in IPv6, service providers can build a much larger network that is manageable end to end.
Now, this ‘foundation layer’ can be used as a substrate to overlay IPv4 services. This is akin IPv6 providing a layer 2.5 network that can be leveraged to position IPv4 services and infrastructure in different locations.
For example, in residential networks, the NAT function of the home gateway and the CGNat can be co-located with a technology known as DS-Lite. This consolidates multiple layers of NAT at one point, simplifying the whole system and increasing reliability.
In commercial networks, virtual aggregation of the IPv4 address space can now be achieved because the overlayed IPv4 address plan does not have to strictly follow the underlying topology.
By decoupling the deployment of IPv6 in the network from the deployment of IPv4 in the services, a service provider can solve the chicken and egg problem that has faced IPv6 for the last twenty years.
Using this methodology, the costs and benefits of deploying IPv6 can be aligned. With this viewpoint, IPv6 alone is NOT the solution to IPv4 address exhaust (NAT is), but IPv6 is a solution to reduce the OPEX of delivering IPv4 services. This would be a logical first step. As a second step, once this IPv6 layer 2.5 is available, content and eyeballs will naturally follow. Remember, any packet that flows over IPv6 is a packet that will not go through the NAT infrastructure. This will enable a cap & grow strategy: cap the IPv4 NAT investment and at the same time grow IPv6 services.
Exploring the vision for the networking industry and the issues shaping its future.