10 Years Since World IPv6 Launch
The World IPv6 Launch event on June 6, 2012, was where Akamai, major web content providers, and internet service providers (ISPs) collaborated to overcome the inertia that had been holding back the widespread deployment of IPv6. IPv6 is the current generation of the Internet Protocol (IP) and it addresses the scalability issues inherent in legacy IPv4. Over the past decade, Akamai has witnessed massive growth in IPv6 adoption, bringing it from a narrowly used protocol to a default one that carries a majority of traffic for some customers and for some significant end-user populations.
In the spirit of the 10th anniversary, this blog post will highlight 10 things that have improved with IPv6 deployment over these 10 years, as well as list another 10 areas where work still remains.
As a general trend, the past decade has been about enabling IPv6 to live alongside IPv4 by dual-stacking networks and services, with some use of IPv6-centric networks that still have IPv4 access. Increasingly, we are seeing trends appear around enabling new capabilities and decreasing complexity through the use of IPv6-only networking.
10 areas of growth and evolution for IPv6 from the past decade
1. Akamai’s IPv6 traffic levels and client base
Akamai has seen a massive increase in IPv6 traffic levels over this past decade (Figure 1), due to increased end-user adoption (in large part because of more network provider adoption) and increased IPv4+IPv6 “dual-stacked” (IPv6-enabled) content on Akamai.
As of May 2022, Akamai’s peak IPv6 traffic stands at just over a whopping 41 Tbps (equal to 41 million Mbps!), which is almost double our 21 Tbps peak that we reported just two years ago in February 2020. Akamai’s content delivery network (CDN) is also serving more than 4 trillion requests per day over IPv6, which is a greater than 1,000 times increase from World IPv6 Launch when we delivered 3.9 billion IPv6 requests in a day. On “World IPv6 Day” in 2011, we served only 8 million IPv6 requests.
The number of unique IPv6 addresses Akamai sees in a given day has also gone up dramatically. On a typical day in April 2022, the Akamai CDN saw more than 2.2 billion unique IPv6 /64 prefixes and more than 7.5 billion unique IPv6 addresses (including over a billion in some single-hour windows). This is a few hundred times the 19 million unique IPv6 addresses we saw on World IPv6 Launch in 2012, and we see more unique IPv6 addresses than there are possible IPv4 addresses.
Fig. 1: How IPv6 traffic on the Akamai CDN has grown in the decade since World IPv6 Launch
2. IPv4 exhaustion and increasing prices
One of the primary motivations for IPv6 has been the challenges of scaling IPv4 (which only supports 4 billion possible addresses). Back in 2012, IPv4 addresses were readily available from the Regional Internet Registries to those who could demonstrate a need. Those supplies have run out and now anyone requiring significant numbers of IPv4 addresses for growing ISPs, cloud services, or other needs typically has to purchase them on the IPv4 transfer market.
Auction prices for IPv4 addresses keep going up: IPv4 addresses typically sold for around $20 in 2020 and often sell for upwards of $50 per IPv4 address today. This means that even a single IPv4 /16 (64K addresses) may now be worth more than $3 million. At small scales this isn’t a problem, but it quickly becomes costly for businesses looking to substantially grow their subscriber bases, as the number of IP addresses needed grows with the number of subscribers. Acceleration of the ways that IPv6 is being used can help alleviate these challenges.
3. Government IPv6 mandates
While Akamai helped our U.S. government customers meet a 2012 mandate to enable IPv6 on public-facing websites, additional IPv6 adoption mandates have been issued by various governments. In November 2020, the U.S. government’s OMB issued an additional mandate, M-21-07, which directs federal agencies to transition to IPv6-only networks over the next few years. In 2021, China also issued a plan for widespread adoption of IPv6 in the near term, with a requirement that new networks won’t be allowed to use IPv4 after 2023. Other countries around the globe have also issued (or are exploring) similar mandates, especially in the context of regulated industries.
4. Global adoption in many of the largest countries and ISPs
Akamai observes end-user adoption levels in the approximately 44% to 62% range for 6 of the top 10 global economies, as measured by looking at the percentage of requests to a subset of IPv4+IPv6 dual-stacked content on the Akamai CDN. In comparison, global IPv6 end-user adoption had barely started in 2012 (with Google’s measurements showing only approximately 0.6% of clients using IPv6).
Fig. 2: Percentage of requests over IPv6 to a subset of dual-stack sites on Akamai from July 2013 to May 2022 for top 10 global economies (by GDP in 2022, per IMF).
Akamai has been measuring IPv6 adoption by looking at the percentage of requests we observe over IPv6 compared with the total number of requests on a given set of IPv4+IPv6 dual-stacked hostnames (Figure 2). Note that the results have sensitivity to the shifting customer traffic mix (such as which devices and ISPs are included at any given point in time), and this causes both fluctuations in our measurements as well as differences relative to similar measurements performed by others. Details for additional countries and networks are available at https://www.akamai.com/ipv6.
Many countries worldwide have seen very substantial increases in IPv6 deployment over the past decade. China’s IPv6 deployment didn’t start at a significant scale until a few years ago, but they’ve now reached approximately 25% IPv6 adoption. A number of large countries now have IPv6 adoption levels exceeding 40%, including Belgium, Brazil, France, Germany, Greece, India, Japan, Malaysia, Mexico, Portugal, Saudi Arabia, Thailand, United Arab Emirates, United States, and Vietnam. About half of the world’s 100 largest networks and ISPs now also have IPv6 adoption levels of greater than 40%, as observed by Akamai.
5. IPv6-centric networks, especially on mobile
Many of the world’s largest mobile ISPs have migrated their networks to be IPv6-centric, and more than 90% of handsets on those networks have IPv6-only connectivity. Handsets that have only IPv6 addresses connect to legacy IPv4 resources via IPv4-as-a-service over IPv6 (IPv4aaS), typically using a combination of the NAT64+DNS64+464XLAT protocols. Apple’s AppStore even requires apps to support IPv6-only networks, and Android employs 464XLAT to help apps that don’t include native IPv6 support.
For these mobile networks, this means that IPv6 traffic natively accesses the internet while IPv4 traffic is forced to go through a Carrier Grade Network Address Translation (CGNAT) server. For some of the largest mobile networks, with many tens of millions of subscribers, even using a single private IPv4 address pool doesn’t scale (as even 10.0.0.0/8 only has 16.8 million addresses). So IPv6 becomes the only option without reusing the same IPv4 address in different parts of their network, along with associated operational headaches.
As many ISPs now see a majority of their traffic using IPv6, some may have limited incentives to keep investing in scaling their CGNAT infrastructure. This means that legacy IPv4 traffic is subject to another source of congestion and loss, especially during large events.
Residential broadband networks have less control over their client devices, including home gateways, so those ISPs have not been able to go to the same extreme, giving client devices only IPv6 addresses. However, many large residential ISPs are switching to IPv6-centric architectures that also use IPv4aaS through technologies such as MAP-T, lw4o6, and DS-Lite. This is a contrast to what we observed 10 years ago when some ISPs were running IPv6-over-IPv4 with 6RD.
There have even been anecdotes that some ISPs are or plan to break out IPv6 traffic at the edge of their network while bringing IPv4 traffic to more central CGNAT locations. As that continues to happen, not only will IPv6 have lower latency and better performance but it will also have better IP geolocation.
6. IPv6-only clients
A number of closed-system client device platforms have been deploying with a single-stack IPv6-only model. This can greatly simplify management at-scale, and can also enable new use cases. Truly IPv6-only general-purpose end-user clients (without ways for legacy IPv4 access) are still a ways out, and likely will stay that way until most ISPs and content are IPv6-enabled. However, within constrained environments with control over content and connectivity, it is becoming simplest and most scalable to just deploy IPv6.
For example, Comcast’s Xfinity X1 set-top box is typically deployed IPv6-only, meaning that content must be IPv6-enabled to reach it effectively (and Akamai sees nearly 100% of dual-stack content using IPv6 to reach X1 devices).
The highly anticipated new Matter Internet of Things (IoT) standard is also being developed to require IPv6. It is also becoming a somewhat common pattern to use local IPv6 addresses to pair-wise connect devices within closed environments.
7. IPv6-enabled relay services
Akamai is one of the partners delivering Apple’s iCloud Private Relay service. Because this decouples end users from origins for the purpose of improving user privacy, IPv6-enabled clients will use IPv6 to access Private Relay, even for IPv4-only origin content.
Similarly, the egress proxy will use IPv6 to contact IPv4+IPv6 dual-stack origins even for IPv4-only clients. Higher usage of IPv6 in the last mile, even when origins don’t directly offer IPv6, can result in improved end-user performance and more efficient routing. As IPv6 source addresses can have finer granularity than IPv4 on some egress proxy operators, server operators are encouraged to enable IPv4+IPv6 dual-stack to obtain a more accurate client location.
8. IPv4+IPv6 dual-stack content
Akamai has enabled IPv4+IPv6 dual-stack as the default for our CDN delivery products for many years, meaning that customers have needed to opt-out for content to be IPv4-only. We’ve also made it easy to switch IPv4-only sites to IPv4+IPv6 dual-stack, including making it scriptable through our application programming interfaces (APIs).
We’ve seen steady growth across much of our customer base of IPv6 getting enabled both through new configurations defaulting to IPv6 as well as customers enabling IPv6 on existing configurations. More than 40% of the tens of thousands of Akamai CDN customer hostnames that deliver over a million requests per day are now enabled for IPv6. This is a huge increase from the 5% that we saw back in 2015, and this has been steadily increasing (Figure 3).
Fig. 3: Growth of IP4+IPv6 dual-stacked hostnames on Akamai, serving more than 1 million requests per day
Beyond Akamai, NIST has been monitoring U.S. industry domains and now sees 20% of websites enabled for IPv6 and more than 65% of DNS enabled for IPv6. Other third-party monitoring shows that a majority of top sites in the U.S. and other countries are enabled for IPv6, and the Internet Society reports that 30% of the top 10,000 websites globally are enabled for IPv6.
9. IPv6 enables better performance
Over the past decade, IPv6 has shifted from potentially limiting performance (due to a limited deployment footprint and IPv6-over-IPv4 tunnels) to instead enabling substantially better performance. Akamai now has IPv6 deployed in the vast majority of its global CDN locations.
Studies over the past few years show dramatic IPv6 performance improvements:
Facebook performed a study in 2015 that showed that websites load 10%–15% faster over IPv6.
Akamai’s customer AbemaTV did a case study in 2019, which showed that IPv6 improved the throughput by 38% on average when compared with connections via IPv4.
Apple announced during the Worldwide Developers Conference 2020 that median connection setup times are 1.4x faster when IPv6 is in use when compared with IPv4.
Not all situations result in IPv6 performance improvements this extreme, but some of our recent measurements of customers who enabled IPv6 have shown valuable performance improvements, especially in some ISPs with widespread IPv6 deployments.
The reasons for better performance with IPv6 vary widely but may include:
ISPs that network address translate (NAT) IPv4 traffic but natively route IPv6 traffic, especially in cases where the NATs experience congestion or runs out of IPs/ports
IPv6-centric ISPs that deploy IPv4aaS over IPv6, especially when IPv4 traffic is routed through more central egress services while IPv6 can break out at the edge
Emerging cases where limited availability of IPv4 addresses mean that IPv6 services can be deployed with a broader global footprint
Akamai continues to encourage customers to enable IPv6 for their content and sites as an easy way that can often increase performance.
10. IPv6-only for cloud hosting and server deployments
The emerging use of IPv6-only (or IPv6-centric) environments for cloud hosting and data center server deployments unlocks new capabilities and enables new use cases. With the rapid growth of cloud deployments, and with the increasing prices of IPv4 addresses, provisioning all server instances with IPv4 addresses can be cost-prohibitive. Even using private IPv4 addresses (such as rfc1918 space) can have scaling and management limitations in large organizations.
Server environments often run more constrained software, and running both IPv4 and IPv6 in parallel is more operationally intensive. As such, a good model for new server and cloud deployments is to only use single-stack IPv6 internally, and to optionally use NAT64 for accessing external IPv4-only resources if needed.
Using IPv6 also means that entire subnets (e.g., a /64 contains 64 bits worth of individual addresses!) can be routed to individual machines. This means that individual containers, services, or even pieces of content can be given their own IP address. This can be valuable for container-based services, such as Kubernetes, whose network stacks sometimes give each container an IP address and also need additional IP addresses for ingress.
Linode: supporting IPv6 since 2011
Linode, a cloud computing service that is now a part of Akamai, has supported IPv6 since 2011. While Linode always provisions IPv4 addresses for customer virtual machines (VMs) today, providing IPv6-only VMs is being explored and will inevitably be offered at some point in the future. Some other cloud and hosting providers are already offering IPv6-only services with IPv4 addresses being an add-on.
Linode’s latest core network generation is already native IPv6-only, which has drastically reduced management overhead and complexity, and some features (such as BGP peering for elastic IP sharing) have launched requiring IPv6 support. Facebook has also moved their internal network to be IPv6-only. AWS recently launched IPv6-only VPC support. This trend is likely to continue.
Enable dual-stack delivery with the CDN
An interesting use case for the Akamai CDN is to enable dual-stack delivery at the edge (providing access to both IPv4-only and IPv6-only and dual-stack end users), but to connect to an IPv6-only content origin using only IPv6 (which can even enable a firewall to only allow a subset of IPv6 address space and no IPv4 address space). This can be configured today with the Akamai API Acceleration delivery product when using the Origin IP access control list feature, and we will be adding support to additional CDN products over the coming year.
10 opportunities for IPv6 growth over the coming decade
IPv6 adoption is a function of IPv6 availability and use across networks/ISPs, local network situations, client software, and service endpoints (such as content and websites). Each of these has come a long way over the past decade, but we still have at least another decade ahead of us before IPv6 is deployed as ubiquitously as IPv4 is today.
1. Broad deployment of IPv6 by more networks, especially at the long tail
While many of the largest ISPs/networks have very substantial IPv6 deployments today, likely because of a strategic need to deploy IPv6 to keep scaling and growing, there is considerably less extensive IPv6 deployment within smaller networks. There is a substantial risk of a two-tier internet where larger and/or faster-growing networks are IPv6-centric and increasingly focus their efforts on IPv6 while smaller networks remain primarily IPv4. This seems like an undesirable state, so we hope smaller networks will also be motivated to deploy IPv6 so that their users can keep up with the way the rest of the internet is evolving.
This difference in IPv6 deployment between the larger and smaller networks can be seen in Figure 4. Of the top 10 networks, only two lack IPv6 adoption over 40%, and both of those do have at least regional IPv6 deployments. Of the top 100 networks, approximately half have IPv6 adoption of over 40%. Looking at the top 800, this drops down to 20% with IPv6 adoption over 40%, and among the top 5,000 networks, only 9% have adopted IPv6 over this threshold.
It is possible that 5G deployments may also be a forcing function for further IPv6 adoption. Internet growth in the developing world could also potentially play into increasing IPv6 deployment. A major factor in India having some of the largest IPv6 deployments in the world is that they broadly deployed mobile 4G connectivity in an IPv6-centric manner.
Fig. 4: Percentage of the top-N networks (sorted by number of requests observed to a subset of dual-stack hostnames in May 2022) with IPv6 adoption over 40%
2. Broader country-level IPv6 deployment
The IPv6 adoption by individual countries is very much a function of the IPv6 deployment within the networks that operate in those countries. It is likely that a number of factors will drive broader country-level IPv6 adoption, including:
National drives, such as China’s plan to transition to IPv6 over the next few years.
Regional collaboration, such as “IPv6 Councils” that have been effective at helping local networks share best practices and drive adoption. In Europe, for example, the IPv6 Councils in Belgium, Switzerland, the United Kingdom, and some others helped drive early IPv6 adoption. The International Telecommunication Union and the IPv6 Forum both have lists of many country-level councils and forums, including some that have come into existence recently.
Developing countries that are expanding their internet presence are best off deploying IPv6-centric networks because of the lack of IPv4 needed to fuel substantial growth.
3. IPv6 for more content and services
As IPv6 becomes increasingly deployed across the globe, especially in IPv6-centric networks that give it a performance benefit over IPv4, content and service providers will be increasingly motivated to IPv4+IPv6 dual-stack.
As mentioned above, Akamai defaults to IPv6-enablement and makes it easy for our CDN customers to enable IPv6 and encourages customers to do so. For our customers, the primary constraints that prevent enabling IPv6 include back-end logging and third-party systems such as fraud prevention services that don’t yet support IPv6.
Tokens tied to IP addresses can also be a problem with IPv6 (and, increasingly, other scenarios). We’ve also worked proactively with many customers to help them overcome issues that prevent them from enabling IPv6.
As more IPv6-only deployments appear (such as in cloud/hosting environments and constrained environments such as ISP-managed set-top boxes), content and service providers that are IPv4-only will increasingly have problems reaching some of their end users. Government mandates, purchasing requirements, and guidance for regulated industries may also increasingly require that content and services are IPv6-enabled.
4. IPv6 in consumer networking devices
One of the impediments to networks/ISPs reaching 100% IPv6 adoption is that a subset of consumer networking devices (such as home internet gateways) lack full IPv6 support, have IPv6 disabled by default, or don’t implement the IPv4aaS functionalities utilized by the ISP. This explains why some of the highest IPv6 adoption levels are seen by ISPs that control which devices connect users to their networks — but even their users may have their own networking equipment downstream from the ISP’s equipment.
Especially in the case of ISPs who have deployed IPv4aaS over IPv6, users with IPv4-only networking devices may be subject to multiple layers of NAT and be at a disadvantage relative to users with end-to-end IPv6 support. Hopefully, as IPv6 deployment becomes the majority case, manufacturers of consumer networking devices will build in robust and default-on IPv6 support.
5. IPv6 in consumer electronics
We’ve observed that IPv6 support is lacking in some consumer electronics, especially some popular digital media players. This is a major contributor to depressing IPv6 adoption numbers within residential broadband ISPs, sometimes decreasing the adoption metrics we observe by 5% or more. (Note that many digital media players, smart TVs, and similar devices do have, and use, IPv6 support.)
As these consumer electronics devices gain IPv6 support over the coming years, this will both substantially reduce “residual IPv4” traffic, offload ISP CGNAT devices, and also improve end-user performance, especially in broadband ISPs that are deploying IPv4aaS technologies. It will be important, however, for vendors to provide their users with robust IPv6 stacks since sometimes having a bad IPv6 implementation can be worse than only supporting IPv4.
6. IPv6 in gaming and virtual reality
Gaming is a vertical that has quite a bit to gain from IPv6, especially for allowing low-latency point-to-point communication between players without relying on NAT. Microsoft Xbox even recommends IPv6 for the “best possible experience.” For some reason, however, some game consoles and game developers have lagged behind others in fully implementing IPv6 support. This may have made sense with tight time-to-market schedules when IPv6 adoption around the world was low, but now that IPv6 is widely deployed in many parts of the world we hope to see a shift to where the gaming and virtual reality industries start treating comprehensive IPv6 support as a must-have.
7. IPv6 in enterprises and universities
We’ve observed that enterprise and university IT organizations lag residential ISPs in deploying IPv6, even if some universities were some of the earliest adopters. This is significant enough that country-level IPv6 adoption metrics rise on weekends and holidays when more people are at home rather than in the office!
For smaller companies, there hasn’t yet been a huge motivator to deploy IPv6. However, many large enterprises are struggling to manage their IPv4 private address allocations as their networks grow and as they are involved in mergers and acquisitions. Managing overlapping and merging and splitting pools of IPv4 addresses is organizationally difficult and can make configuration management and forensics challenging.
Organizations would also rather operate single-stack networks than dual-stack networks, so some of the real value will be delivered when it is straightforward to build and operate IPv6-only enterprise networks that leverage DNS64+NAT64 for reaching legacy IPv4-only resources. It is not yet common for organizations to be building IPv6-centric enterprise networks, but that may evolve as a trend over the coming years. The U.S. government mandate for IPv6-only support within agencies may help drive the maturity of IPv6-only support in enterprise vendor products.
8. IPv6 in security products
It has always been critical that security products have robust IPv6 support as attackers haven’t waited to attack over IPv6. Even a decade ago we saw attacks over IPv6, as malware running on IPv6-enabled clients with malware using OS libraries would just follow IPv6 DNS records to attack servers, and the maturity of attackers has only grown since then. Note that IPv6 is not inherently more secure than IPv4, and neither is IPv4 NAT inherently secure or a replacement for a firewall and a robust security solution.
At the same time, the vastly larger IP address space of IPv6 opens up more opportunities for creative solutions that improve defense and management. The lack of IPv4 NAT may make it easier for some security products to evaluate client populations, as with IPv4 a single CGNAT address pool may hide tens of thousands of clients. Although there are ways attackers can sometimes scan portions of IPv6 address space, IPv6 is still much harder to comprehensively scan — using an IPv6-only CDN origin with a sufficiently random address and DNS name may be much harder for attackers to find, for example.
9. IPv6-only in cloud and hosting providers and data centers
As mentioned above, the rapid growth of the cloud is putting pressure on IPv4 address availability. Especially with many services being fronted by a CDN, load balancer, or other ingress layer, there are increasingly fewer reasons not to build out new infrastructure as IPv6-only, or at least IPv6-centric, from the start. (The major remaining reason is the limitations in IPv6 support in some commonly operated services, as discussed below).
As a simple example, relying on IPv6 Stateless Address Autoconfiguration (SLAAC) to bootstrap server provisioning can streamline deployments. As a broader example, building our IPv6-only Kubernetes clusters greatly simplifies cluster networking. Over the coming decade, relying on IPv6 for server infrastructure will be an increasingly common trend. Akamai (and Linode, which is now a part of Akamai) will be leveraging this trend over the next few years and we are already exploring offerings for services that rely on IPv6-only cloud origins for our CDN, as well as IPv6-centric offerings on Linode.
10. IPv6 in more open source software
One of the gating factors for running IPv6-only services in cloud environments has been gaps in open source software support. For example, dual-stack IPv6 support in the Kubernetes reached general availability only in December 2021 and the results of this are still percolating through the rest of its ecosystem. The widely used Hadoop is an example of an open source project that has not yet released production of IPv6.
The situation today is different than it was a decade ago, however. Back in 2012, there was IPv6 support in major operating systems and some applications, but IPv6 adoption wasn’t far enough along to motivate closing remaining gaps. For many of the reasons listed above, there is much broader support for IPv6 in open source software today. In a decade, I would expect to see many new open source projects written IPv6-first with IPv4 as a legacy feature or even as unsupported.
Conclusion
Akamai has seen massive growth in both IPv6 traffic and adoption in the past decade since the World IPv6 Launch. Although it seems likely that there will still be significant IPv4 traffic 10 years from now, we hope to be actively talking about paths to sunsetting IPv4. It is even possible that the need to upgrade and replace systems in the lead-up to 2038 (a looming Y2K-style event when “Unix Epoch Time” rolls over 32-bit signed integers) will force the retirement of some remaining IPv4-only systems and networks.
In the meantime, we continue to encourage our customers to enable IPv6 on their content and services, and we look forward to seeing what the next decade will bring for IPv6 adoption.
Thank you to the countless people at Akamai who have contributed to our IPv6 efforts over the past years (and beyond), as well as contributed to this blog post.
While precautions have been taken in the preparation of this document, Akamai Technologies, Inc. assumes no responsibility for errors, omissions, or for damages resulting from the use of the information herein. The information herein is subject to change without notice. Akamai and the Akamai wave logo are registered trademarks or service marks in the United States (Reg. U.S. Pat. & Tm. Off). Published June 6, 2022.
