IPv4/IPv6 dual stack better, but more expensive and less future-proof than DS-Lite and 464XLAT
IPv6-only is slowly coming closer
IPv6-only is slowly coming closer
Various options exist for enabling both IPv4 and IPv6 for your users. The best transition mechanism to use depends largely on the infrastructure you've currently got, and whether the network is fixed-line or mobile. Of the four solutions we see most often, a complete IPv4/IPv6 dual stack (usually supported by NAT44 or NAT444) is currently likely to create the fewest headaches. The main drawback with this solution is the infrastructure cost: it effectively means operating two entire networks in parallel. And, if NAT444 is used, there is the additional cost of double NAT translation. DS-Lite and 464XLAT, both of which are based on an IPv6 network that can also handle IPv4 traffic, offer more future-proof infrastructures and more straightforward transition. One key difference between them is that 464XLAT doesn't support DNSSEC for IPv4.
Let's begin by outlining the basics of the four technologies mentioned above:
Dual stack Separate IPv4 and IPv6 networks operating alongside one another. With a few exceptions (Freedom Internet and the old XS4ALL), the IPv4 network operators use NAT (NAT44) to translate a private IPv4 address (RFC 1918) into a public IPv4 address.
NAT444 A CGNAT technology developed specifically to allow NAT44 to be scaled up for an even greater number of end users. It involves two NAT translation layers based on the specially reserved address block 100.64.0.0/10 (RFC 6598). For a detailed explanation, see this article.
464XLAT A technology that uses a DNS64 DNS server to map (embed) connections to IPv4-only servers to the dedicated IPv6 address block 64:ff9b::/96. Traffic can then be routed over an IPv6-only network to a NAT64 gateway operated by the access provider (RFC 6877). For a detailed explanation, see this article.
DS-Lite A CGNAT technology that leans heavily on the end user's 'customer-premises equipment' (CPE), which tunnels traffic from a private IPv4 address over an IPv6-only network to a CGNAT gateway operated by the access provider (RFC 6333). The Lightweight 4-over-6 extension also shifts NAT translation to the CPE (RFC 7596), so that the CGNAT gateway has to maintain only a very small amount of state information.
The advantage of the last two technologies (464XLAT and DS-Lite) is that they are based on an IPv6 network but handle connections to IPv4-only servers as well (albeit not always perfectly). That removes the need to operate two IP networks in parallel, as with a dual-stack set-up (with or without NAT44/NAT444). 464XLAT in particular is designed with a view to removing the IPv4 legacy from your network as fully as possible.
AIS and 3BB – the two Thai access providers discussed below – decided to use dual-stack set-ups for their fixed-line networks because of the density of the IPv4 legacy in their edge networks and end-user environments. Both of these early adopters have identified the flawed IPv6 support provided by old equipment as a significant drawback of their chosen approach. Nevertheless, the high cost of (CG)NAT systems (for IPv4-only) was felt to warrant the creation of a complete IPv6 network alongside the IPv4 network. In the Netherlands, meanwhile, Ziggo is using DS-Lite for its IPv6 infrastructure. However, one often sees posts in the company's user forum highlighting the fact that the technology can't be combined with port forwarding or with use of a CPE router. The only solution to that problem that the access provider can offer is to put the user back on IPv4-only. 464XLAT was pioneered by T-Mobile USA. More recently, the technology was also chosen by South Korean mobile provider SK Telecom for its new LTE network. Meanwhile, Australian telecom service provider Telstra transferred its mobile users from dual stack with NAT444 to 464XLAT-based IPv6-only at the start of the year. That move was prompted by the high cost of acquiring IPv4 addresses and new CGNAT hardware. As previously highlighted on this site, tens – and in some cases even hundreds – of millions of (mobile) internet users have been connected to IPv6-only networks in Asia and South America. It's therefore simply a matter of time before IPv6-only services – in other words, services that IPv4-only users (here) can't access – are launched in those regions.
Over the last three years, use of IPv6 in Thailand has gone from almost nothing to 35 per cent. The rapid growth is mainly down to two major access providers: AIS and 3BB. In the last couple of years, the first of those companies enabled IPv6 for cable customers and replaced its native IPv4 connections with a CGNAT set-up. For AIS, IPv6 is the default option, both for users and in the development of new services. 3BB has also adopted a dual-stack configuration, but based on IPv4 with NAT. Again, though, the emphasis is on IPv6, mainly with a view to reducing expenditure on NAT systems.
Both Thai access providers were early adopters and faced the challenges associated with an existing customer infrastructure. That was why they chose dual-stack configurations. In terms of both extent and speed, adoption was held back by lack of adequate IPv6 support in some customer equipment, coupled with the need to ensure service continuity.
A company that has the luxury of being able to roll out an entirely new mobile infrastructure is more likely to choose 464XLAT. The was the case with SK Telecom, South Korea's biggest telecoms operator, which decided to provide its 23 million LTE users with IPv6-only connections by default. In Vietnam, telecom firms, government organisations and service providers have linked up to promote IPv6 over the last decade. As a result, the country has achieved 45 per cent adoption, a figure bettered in Asia only by India, Malaysia and Taiwan (where the rates are 70, 53 and 50 per cent, respectively).
India's very high IPv6 adoption level owes much to mobile communications provider Reliance Jio, which has transferred several hundred million Android users to IPv6-only services. The migration was relatively straightforward, because the operator's national mobile network was a modern 4G network rolled out as recently as 2016. "We had an IPv4 address shortage when we only had computers connected to the network. Now there are half a billion smartphones in India, and before long air conditioners, fridges, wearables and countless other devices will be connected to the net as well," said Akamai's Product Marketing Manager Vaishali Sangtani at the time. "IPv6 is the future: all upcoming improvements to internet protocols, standards and security arrangements will be based on IPv6, not on IPv4. So it's really important that we don't miss the boat." In Brazil, meanwhile, access providers have been wrestling with the extreme local shortage of IPv4 addresses. Because every available IPv4 address has to be assigned to a business user, end users are almost always behind CGNAT systems, making gaming and other peer-to-peer applications very difficult.
In China, IPv6 users now account for 20 per cent of the user population. That percentage may not be impressive in absolute terms, but it's double what the figure was just eighteen months ago. The rapid recent rollout is part of an industry policy of moving all China's internet users to IPv6 by 2025.
China Telecom's engineers have produced an Internet Draft describing how the state-driven 2017 IP Action Plan has led to IPv6 (dual stack) now being implemented almost throughout the network infrastructure (including data centres and cloud infrastructure). Universal implementation within a network is a prerequisite for user-side IPv6 adoption, whereas most IPv4 users are currently connected to NAT44/NAT444 networks. According to the authors, (other) transitional IPv6 technologies such as DS-Lite, Lightweight 6over4 and IVI Translation were evaluated, but the standards were still under development at the time and very few commercial CPE devices supported them. Therefore, starting in 2015, dual-stack IPv6 was rolled out province by province as a standard feature of LTE (pre-4G). Although the number of users to whom IPv6 is available has increased greatly over the last two years, active IPv6 users account for 79 per cent of all users on the mobile network, but only 30 per cent of those on the fixed-line network. According to the authors, the reason is that a lot of end-user equipment doesn't yet support IPv6. In that context, it's worth noting that Chinese users can buy and connect their own modems, many of which don't offer IPv6 support. What's more, a significant proportion of commercial content providers don't enable IPv6 access, generally because their infrastructures are dependent for IPv6 support on data centres and content delivery networks (CDNs).
"IPv4 address shortages exist even in countries whose fixed-line and mobile internet infrastructures have been in place for some time," says Chongfeng Xie, Senior Network Engineer at China Telecom and one of the authors of the Internet Draft. "China is the most populous country in the world, and it simply doesn't have enough addresses to set up new services, such as cloud and IoT services. What's more, CGNAT makes networks more complex and more expensive. IPv6 has the advantage that all network nodes have end-to-end access, which keeps down the cost. So there's an urgent need to migrate networks and services to IPv6." Xie sees huge potential for IPv6-based IoT innovation. "With IPv6, IoT devices can communicate directly with one another and with their service platforms. It's therefore increasingly common for NB-IoT chips and terminals to support IPv6. Only IoT solutions that operate within closed environments can manage without IPv6." "Major Chinese content providers, including Alibaba and Tencent, are now using IPv6 for their service provisioning. And parent company Alibaba reports that more users are now accessing the Youku video service and Gaode Maps (AutoNavi) with IPv6 than with IPv4. Whether we'll actually see the introduction of IPv6-only services any time soon I'm not so sure, but it's certainly possible."