Around 15 years ago, a small hotel in Alexandria, VA, would become the birthplace of an idea that inspired two of the most globally significant IPv6 events in history. While the standards work was complete for IPv6 production, caliber support in commercial products, namely hardware and software, lagged. Furthermore, there was still much industry discussion of “if” IPv6 would be deployed, not “when.” The groundwork for the World IPv6 events was set in motion several years before planning for the World IPv6 Day, and the Launch began. Organized by the Internet Society, where co-author Leslie Daigle worked at the time, our collective hope was that these would be the only actual and last ever industry events required to make large-scale adoption of IPv6 a reality.
The Network Operator Perspective
My personal IPv6 history dates back to the late 1990s. Broadband (specifically of the cable variety) was starting to take off. DOCSIS was being developed and standardized, DSL was the dominant media to access the Internet, and Google had just been born and was getting attention. The startup I worked for specialized in enterprise network management software. With one of the largest cable companies in the US in our backyard, we dipped our toe into the cable broadband provisioning market. The journey that would follow was breakneck and chock full of industry-first scaling challenges for the Internet at the time, pre-dating the web-scale of Google, Facebook, and others. One area that we took notice of was the need and scarcity of publicly routable IPv4 address blocks. IPv6 was intended to address this challenge. However, it was not scale tested then, and most (if not all) implementations were experimental and not remotely close to production grade. The inability for IPv4 to scale to meet the needs of the networks experiencing explosive growth forced large network operators to evaluate the deployment of IPv6 seriously.
The challenge was then to transform IPv6 to production-grade from the experimental implementations that were available at the time spanning everything from network equipment to operating systems and, most dauntingly, the plethora of back office applications that required proper support for IPv6 to achieve true operationalization.
The Internet Growth Perspective
Development of the IPv6 standard began in the era when academic and research networks still constituted a significant chunk of the global network. In that era, it was common to work out refinements of protocols at standards meetings and then go home and implement them in the network. By the time the IPv4 address crunch was upon the global network and IPv6 was the only real answer, the Internet was a different place. Most network operators (large and small) were commercial endeavors, stability was prized over constant evolution, equipment was deployed with an expected lifetime of years (not months), and changes had to be properly vetted before being committed. Further, the beauty of the Internet is that it is about network interconnection – each network operator is free to do whatever they see fit within their networks as long as they follow the standards for interconnection. This allowed for many disparate networks to be connected into the global whole – but it still means that there is no “one size fits all” deployment plan for new technology.
For individual network operators, deploying IPv6 was often a priority – but never the priority. There was always a business-critical feature deployment or a customer requirement higher up the list. And, because few customers actively and engagedly demanded IPv6 support, no single network operator could “go to IPv6” on their own. To get there, a collaboration between industry competitors would be necessary.
Reliably, Scaling IPv6
A few of the many examples of fundamental IPv6 implementation issues that needed to be addressed once production adoption became the objective were software-based (not hardware accelerated) IPv6 implementations to the tune of 19K packets per second (PPS), barely the equivalent of dial-up Internet speeds from pre-broadband, incompatibilities between networking equipment vendors for point to point IPv6 interface configurations (/126 versus /127), and operating systems that became unstable when multiple router advertisements are used on a single link or virtual LAN (VLAN). This was followed by DHCPv6 support for address and delegated prefix assignments, including IPv6 support for many access network technologies like DOCSIS and FTTH. This circa 2005 preceded World IPv6 Day and Launch by six and seven years, respectively.
Due to the vintage and diversity of their infrastructure, many service providers began their IPv6 journeys earlier than the budding Internet firms of the time. However, once some of the largest broadband networks had a clear path, the content giants followed quickly. Google, YouTube, Netflix, and Facebook impressively deployed Internet-scale support for IPv6 to complement the growing number of potential broadband eyeballs that would follow. The fundamental ingredients for two, not just one, global IPv6 events were in place. A collaboration between some of the world’s largest broadband and content networks, as if we all worked for the same company, laid the groundwork for the most significant, global IPv6 events the industry had ever seen. Beyond being poster children for Internet style, industry collaboration, deep friendships that will last a lifetime were made.
Mountains were moved leading up to World IPv6 Day, where content providers enabled IPv6 (dual stack) by default for a single 24-hour period. The result was some much-needed data to assess what has been affectionately referred to as “IPv6 brokenness”. It turns out the results were positive but not perfect. With necessity being the mother of invention, “Happy Eyeballs” was born but would not pervasively be deployed for what would follow one short year later - World IPv6 Launch. One percent was the goal for participating broadband providers. Many who participated achieved the goal, and while others may not have reached the 1% goal, they also laid the groundwork to deploy IPv6 in the months following the World IPv6 Launch pervasively. Some of the most outstanding IPv6 achievements took place in the years that followed - TELUS’ impressive rollout across their mobile and fixed line broadband network along with Facebook and their use of IPv6 are notable examples.
No Other Alternatives for an open Internet
At the time of the Alexandria, VA meeting, the end of IPv4 address allocations was within industry sight, and most technical discussions were about how networks could cope with fewer addresses than active customers at any given time. The proposed answer was “Large Scale NAT” or “Carrier Grade NAT,” (though some argue that the results of such NATs are not carrier-grade). Some discussion focused on how many ports would be assigned to each active user, as means of multiplexing a single IPv4 address at any given moment in time. However, at that point in history, content and application websites were also coming into their own. It wasn’t clear that general web access would be satisfied by the number of ports network operators could offer. The needs of two industries were colliding.
Even though large-scale NATs are a factor in today’s Internet, the reality remains that they can only ever be a stop-gap measure if we expect to see continued Internet application development and evolution.
The years following the World IPv6 Launch yielded continued growth with some notable adoption spikes. Generally, all who participated advanced their deployments, and some newcomers even deployed IPv6 at impressive rates.
Interestingly while the industry focused on continued, incremental deployment, a shift was not so quietly unfolding that could and would shape the rate of IPv6 adoption in the years to come - the cloud. Circa 2005, “the cloud” meant many things to many organizations. In many cases, it meant a purpose-built cloud infrastructure created by an organization for an organization - not necessarily for public use. Each organization, if sufficiently motivated, could (and, in some cases, did) enable IPv6 (dual stack) support. Sadly, in many scenarios, the underlying technology or platform did not support IPv6, resulting in roadblock after roadblock. It also turns out that developing, building, and deploying a cloud infrastructure is not for everyone. This, in turn, produced a massive shift to the public cloud.
Over the past decade, extraordinary shifts have moved workloads to “the cloud.” Innovation, open source or otherwise, has fueled a wide range of adoption. The confluence of this growth, innovation, and absence of IPv6 support has led the industry to some new challenges and opportunities. The challenges, in retrospect, are not so new, being the scarcity of IPv4 and the absence of IPv6. The possibilities, however, are new and quite different - the kind of “different” that could complete what the World IPv6 series of events started over a decade ago. Public cloud infrastructures serve a heterogeneous and growing population of customers worldwide. The shift continued and expanded with the use of CDNs and the introduction of new players like Cloudflare, which is far more than a CDN. Cloudflare paved the way for “IPv6, on by default” for CDNs and, in many ways, today leads by example.
On June 6th, 2022, in Montreal, Canada, NANOG hosted its 85th event. It just so happens that the first day of the conference was on the exact anniversary of the World IPv6 Launch. NANOG dedicated the day (June 6th) to IPv6 content to tribute to World IPv6 Launch. The conference kicked off with a keynote from the Father of the Internet, Dr. Vinton G. Cerf, several IPv6-oriented presentations, and, of course, a panel discussion highlighting the World IPv6 Launch. The call to action by Vint Cerf while addressing the NANOG audience was - “IPv6, on by default”. The audience learned that many of the cloud and CDN heavy hitters, namely Amazon Web Services (AWS) and Akamai, have begun enabling IPv6 by default. Unsurprisingly, Cisco Meraki is enabling IPv6 by default for their enterprise WiFi management and connectivity platform.
Has the industry shift to “IPv6, on by default” happened? If so, why does every data source measuring the adoption of IPv6 still suggest that the transition may take upwards of 10 years from today (2022)? The nagging question is, what remains?
“IPv6, on by default” is not to be mistaken with IPv6 only. Outside of specific use cases, the latter will likely take a decade, probably more, to materialize for consumer and/or enterprise-facing Internet connectivity. The topic of IPv6 only is for another day, another blog post, and another panel - hopefully, moderated by a new set of faces from a new set of companies.
So what is next? Many of the firms that started their journey or began scaling IPv6 since the World IPv6 Launch is enabling IPv6 - “IPv6, on by default”. While the “IPv6, on by default” journey has begun, there is still a long road ahead. Several weeks after the anniversary of the World IPv6 Launch, as planning continues for more content around the globe, specifically for APAC (APNIC) and EMEA (RIPE), the following points still resonate as unaddressed and unresolved:
- Operational use of IPv6 fuels usage
- Enterprise networks and IPv6
- The cost of IPv4
An in-depth analysis of IPv6 adoption suggests that IPv6 use typically increases after standard corporate operating hours, effectively when people leave work and go home - where IPv6 is primarily enabled by default and has been for many since before the World IPv6 Launch in 2012. There is tremendous pent-up demand for IPv6 across some of the world’s largest enterprise networks. The cost of IPv4, coupled with another topic from NANOG 85, “Buying and Selling IPv4 Addresses,” highlights another fascinating intersection of issues. Can the “cash” from the sale of IPv4 addresses fuel the operational use of IPv6 and the deployment across enterprise networks? The leveraging of IPv6 to operate and deliver revenue-generating products and services could very well be a catalyst for adopting IPv6 across enterprise networks globally. Awareness of the enterprise and IPv6 was discussed in depth approaching World IPv6 Day/Launch. However, the task at hand, launching IPv6 for broadband and content networks, was a greater priority to ensure we could achieve scale and reliability.
Now that we have all the critical building blocks in place, are we now approaching (or are in the era)of IPv6 for the enterprise? Are the economic incentives and opportunities finally aligning with technological readiness?
2022, the tenth anniversary of the World IPv6 Launch, must not be a celebration of past successes but instead a reminder of what lies ahead. The work that so many started so long ago is not complete. The growth of IPv6, while still increasing, has slowed in recent years. Support for IPv6 is production-grade; quoting Dr. Vint Cerf from his keynote at NANOG 85, it is ready for “IPv6, on by default”. The convergence of the same with the economic incentives related to IPv4 introduces substantive motivation for a new era of IPv6 adoption that collaborative spans the cloud, enterprises, and content delivery networks.