Randall Munroe (xkcd)'s map of the Internet IPv4 space in 2006

The Sky Still Hasn’t Fallen — Further Thoughts On IPv6

Randall Munroe (xkcd)'s map of the Internet IPv4 space in 2006Just over a year ago, I wrote a piece on why IPv6 has not taken off. 12 months on, has much changed?

To a greater extent, the urgency behind why IPv6 has to be adopted has grown – as mentioned in this piece pulling out the difference between the IoT and the IoE, the growth of devices that are not under your direct control will mean that it becomes increasingly difficult to use IPv4 and network address translation (NAT) to meet the needs of an end-to-end mass device network.

State of the nations…

This interactive table on the RIPE NCC website shows that fewer than 20% of all networks with a registered autonomous system number (carriers plus private) are using IPv6 globally. In the UK, it is over 26% (unfortunately, although low, it is a rather good score) – the US is languishing at under 14%. The number of Alexa Top 1000 web sites (those measured by Amazon’s Alexa division to have the most internet traffic) accepting IPv6 connections is running at less than 15%.

This is not promising – and the rate of growth in the use of IPv6 seems to be slowing, rather than accelerating. Could these low levels of adoption create a major problem for the workings of the IoE?

The IoE and IPv6

Depending on which sources are used, stats show the IoE will result in anywhere between 50 and 500 billion devices being added to the global network by 2020. Driven by three main markets – mobility, wearables and industrial devices – this level of device growth clearly cannot be met by just pushing the envelope of IPv4 use. Although a large chunk of industrial usage can be managed by NAT, mobility and wearables require a global address – these devices need to connect and share data via public networks. No matter how ‘clean’ the IPv4 namespace is (i.e. how many unused addresses are recovered for reuse), no matter how careful everyone is in how they use these addresses, the available pool will be depleted very rapidly.

Just maybe, the use of a software abstraction capability, as embodied in the software defined wide area network (SD-WAN) could be the most elegant way to deal with the problem.

Driven by three main markets – mobility, wearables and industrial devices – this level of device growth clearly cannot be met by just pushing the envelope of IPv4 use.

Carriers, organizations and individuals

According to the World IPv6 Launch website, the majority of carriers are struggling to implement more than 20% of their own networks to support IPv6. If the carriers cannot move to provide full IPv6 support, then there is little impetus for those further down the chain to do so. The carriers need to see how IPv6 investment will help their bottom line: at the bottom, there is little there for them to see any monetization at the end of the IPv6 rainbow.

For organizations using the carriers’ networks, they also have to see something that is immediate and useful to them. For example, why should an organization go to all the trouble of putting in place an IPv6 network – which may entail hardware, firmware and software updates – when they will have to continue operating in dual-stack IPv4/IPv6 mode for the foreseeable future anyway, and not see any real benefits from their investment? The positive impact of IPv6 and the negative of IPv4 needs to be better spelled out so that such investment can justified.

For the individual user, they need to be sold on what IPv6 means for them. This may be easier – if their several hundred dollar wearable doesn’t work because they cannot gain an IP address on the global network, then they will pretty soon understand what is needed.

 Dual lanes?

It may be that the carriers have to take a pragmatic view of this. Putting in place intelligent access points where software defined wide area network (SD-WAN) constructs can more easily direct traffic may be the only way to deal with the problem in the short term.

When a device tries to connect to the network, it sends out various pieces of data that the network acts on before it provides it with an actual address. As part of this, an intelligent access point could find out if the device has an IPv6 address – and send its traffic through an IPv6 only network. Wherever the device is talking native IPv6 (which in many cases with new wearables could be the case), only this network would be required – no IPv4 address space needs to be taken.

Where the device needs to talk to an IPv4-only environment, then the carrier can provide an IPv4 address, but only for the period of that interaction. These IPv4 address can be provided through highly dynamic, short-life DHCP tales, so that the devices are only using the address while active.

The issue then becomes one of concurrency on IPv4 addresses, rather than on actual device numbers. If only 5% of devices are using IPv4 at any time, then the problem becomes more manageable – for a time.

It is still sticking a band-aid over an arterial bleed, though: IPv4 will be a constraint on the global IoE network unless something is done at a very basic level.

As carriers move more traffic over to IPv6, it should be possible to turn off more and more IPv4 addresses. Hopefully, this will tide us over until the global internet becomes predominantly IPv6 – and we can all breathe easily again.

Image credit: Randall Munroe (xkcd.com) / CC-BY-NC