IPv6 use doubled during 2015 and now carries around 10% of all Internet traffic with strong adoption in the US and some Asian countries. Other countries lag behind but promise to catch up. Facebook reports performance gains of 15% from it. So will 2016 be the year IPv6 becomes mainstream? And if it does, how will this affect deployment of cellular infrastructure?
The big benefit
IPv6 has several new features over IPv4, but the greatest justification is the much larger address range (128 bits vs 32) which allows every device to have its own unique IP address. With all IPv4 address blocks having been allocated and no new addresses available, there is growing pressure to migrate. Workarounds such as NAT (Network Address Translation) have been used for a long time to allow many users to share a single address, but these have limitations. It’s been inevitable that the world will shift, but despite strong warnings of doom for several years little seems to have happened. However, it looks like we’ve now reached a tipping point and 2016 will mark the start of visible take-up.
Google maintain and publish an online chart illustrating worldwide adoption which shows a doubling of total IPv6 traffic from 5 to 10% during 2015.
They break this down by country.
Surprisingly, South Korea is only 2% while Greece is 19% and Portugal is 22%
Some underlying reasons
Much of the US broadband internet service is provided by Cable TV companies. Comcast, the largest, has had an extensive IPv6 program in place for some years.
In the UK, Virgin and BT have a stockpile of IPv4 address space so haven’t been very motivated to evolve. BSkyB don’t have such a surplus and have been shipping a lot of IPv6 capable modems and running trials with over 1 million users. Other ISPs are running trials and preparing plans.
One reason for the delay is the need to update and replace huge numbers of home routers/hubs. BT recently announced plans to deploy IPv6 by end of 2016. Their latest Home Hub V5 supports this natively but they still have to announce a solution for the 80% of customers still on V4. Earlier products won’t be supported. There is quite a significant upgrade cost involved throughout the installed base.
Mobile network operators support IPv4 through large scale Carrier Grade NAT. Most modern smartphones (and tablets, laptops etc.) are dual stack capable, but this isn’t widely enabled. Some traffic shaping, security and other traffic management features may not yet be fully implemented within the operator's infrastructure, so it's not just a case of switching it on. Some operators are further ahead, with Heavy Reading reporting that Verizon Wireless already carry 50% of their LTE data over IPv6 (much of that would be to Facebook and Google sites).
You can easily check your own network status by browsing to test-ipv6.com and will be shown a screen similar to the one below. I've masked the network name on the second line.
An overlooked benefit
Facebook has adopted IPv6 whole-heartedly, and now over 50% of US users access it this way. In a presentation last year, they explained that large scale use shows a 15% speed increase for users who connect via IPv6. This translates to worthwhile commercial benefits for their operation, where users have to wait less and can spend more time actively engaged with their service.
Their presentation also explains that take-up in the Enterprise sector has been much lower than residential or mobile, with security concerns being a major factor.
Implications for cellular network infrastructure
End user IPv6 can be introduced by running this as an end-to-end service throughout the network. Most mobile devices are dual stack and can accommodate this. Those functions in the network which actively handle bearer traffic – counting packets towards usage totals, traffic shaping, transcoding video, blocking unsuitable websites etc. – would have to be upgraded and tested to operate in this mode. These are in the core network rather than in the radio access network, so there would be little or nothing to do in the RAN to enable this end user service. Today's small cells of all types would inherently support IPv6 as an end user service and are unlikely to be a bottleneck. This even includes residential femtocells connected through IPv4, because they can tunnel bearer traffic including IPv6 packets back to the core network.
For internal traffic between infrastructure, network operators have their own private IP network. That IP address range isn’t visible externally and doesn’t extend to end user mobile devices. Architects have told me they expect this to remain IPv4 only for quite some time – there is no address space issue to resolve and no new service to introduce, so no strong business case to justify the expense.
One aspect to consider is when small cells are introduced, especially large numbers of residential and enterprise products, whether these should use IPv4 or IPv6 (or be dual stacked) as their primary connection. Performance wise, there may well be a benefit to adopting IPv6. For compatibility, IPv4 support will be essential for quite some time.
For fixed/mobile networks considering major upgrade programs for their home hubs to handle IPv6, this may also be a time to consider if embedding a residential femtocell is a useful capability.
While I don’t think this will impact products immediately and would be more of a roadmap item, it is important for network operators to have a strategy which addresses this and to ensure it maps into their own IPv6 deployment plans.