Date: March 1, 2014
As more service and content providers deploy IPv6, the Internet engineering community is grappling with how to establish milestones and metrics to measure progress in the rollout of next-generation Internet services.
To reflect on the headway that’s been made in IPv6 deployment and to consider what steps need to be taken next, the Internet Society hosted a panel discussion entitled “IPv6: What Does Success Look Like?” in conjunction with IETF 88.
Leslie Daigle, the Internet Society’s chief Internet technology officer, moderated the panel, which included John Brzozowski, fellow and chief IPv6 architect at Comcast Cable Communications; Erik Nordmark, an Internet Architecture Board member and software developer for Arista Networks; and Chris Palmer, program manager for IPv6 at Microsoft.
Daigle kicked off the discussion with a recap of the significant progress that’s been made in IPv6 deployment during the past year. For example, Google reports that the percentage of its users accessing its sites over IPv6 doubled between August 2012 and November 2013 from 1 percent to 2 percent.
“In June of 2012, we had World IPv6 Launch, which was an opportunity for access providers, content providers and CPE vendors to step up to deploy IPv6,” Daigle said. “A lot of major ISPs stepped up to that particular challenge. Sixty-nine networks achieved a measurable amount of IPv6 traffic—at least 0.1 percent—in June of 2012, and now that number is up to 197 networks in October of 2013.”
Daigle said that for these ISPs, IPv6 is “a regular part of their business now, with new subscribers getting IPv6 by default without user configuration.” She added that the deployment of IPv6 by Google, Facebook, and Yahoo is driving most of the IPv6 traffic.
The major ISPs that are deploying IPv6 for their users are finding a significant amount of their traffic immediately migrates from IPv4 to IPv6, Daigle added. “Verizon Wireless sends 40 percent of their traffic to Google, Facebook, and Yahoo over IPv6,” she said. “When IPv6 gets turned on, pretty serious things happen.”
Daigle posed a series of questions to the panelists about progress being made in IPv6 deployments. In particular, she wondered how the IETF community will know if it is progressing both in the right direction and quickly enough. “As a community, what are the metrics and milestones of importance?” she asked.
Brzozowski argued that the best metrics are IPv6 traffic volumes and the numbers of users with IPv6 access to the Internet. He said that Comcast has IPv6 deployed across 75 percent of its broadband network, with 25 percent of its customers currently using IPv6. He said Comcast will complete its deployment of IPv6 across its broadband network in early 2014.
“By this time next year, we do expect to have double the penetration,” Brzozowski said. “We hope to have 50 percent to 60 percent of our customers using IPv6.”
The biggest hurdle to IPv6 deployment is the lack of support from consumer electronics vendors and content providers, Brzozowski said. “Our statistic when we light up a house with IPv6 is that 20 percent of the traffic almost immediately happens over IPv6,” he said. “That’s not bad, but that’s not great either.”
Brzozowski said one milestone that will indicate successful deployment of IPv6 is when it’s possible for users to operate in an IPv6-only fashion. “We hope to announce IPv6-only trials and, of course, they will be opt-in,” he said. “We really want to go down that path so we can better assess what the IPv6-only experience is and is not.”
Nordmark pointed out that IPv6 is not making inroads in enterprise networks, which he says will hamper interoperability if it is a long-term trend. However, one possibility is that the Bring Your Own Device (BYOD) shift in corporate IT environments will prompt companies to adopt IPv6.
“The trend that people want to bring their smart phone or tablet to work and have the same type of services when they use the Wi-Fi of the enterprise as they have at home means that there will be some pressure on the enterprise to adopt IPv6,” Nordmark said. “Also, companies are using more virtual networks, and for some virtual networks done at very large scales, IPv6 might be a natural fit for simplicity and flexibility reasons.”
A milestone that Nordmark says would indicate continued progress for IPv6 is when software developers automatically support both IPv4 and IPv6. “Something that we can collectively work towards is IPv6 version agnostic software development, which will help over a very long time period,” he added. “But I think to get rid of IPv4 legacy software will be about 20 years from now.”
Palmer pointed out that IPv6 has picked up momentum in three categories: in-home hardware and operating systems; in-home routing infrastructure; and network access. “In all three of those dimensions, there has been significant qualitative and quantitative improvement and forward velocity when it comes to IPv6 transition. So that’s good news,’’ he said.
Palmer prefers a simple metric for measuring the success of IPv6 deployment between now and 2020: a working Internet.
“The simplest thing that I would like to have is the Internet continue to work and work well for the purposes of using it every day, despite the fact that the number of devices is going to increase by 30 times between now and 2020,” Palmer said. “We need to be ready for that.”
When discussing specific metrics that indicate IPv6 progress, Brzozowski said Comcast will reach an important milestone when its Internet-facing traffic is 51 percent running on IPv6. Nordmark would like to see corporations running VPNs over an IPv6 backbone. Palmer said he would like to see less reliance on transition mechanisms that allow IPv6-only devices to reach IPv4 sites.
“Now that we’re getting close enough to IPv6 being operational by default, how do we avoid or sunset the transition technologies… that leave one foot left in the IPv4 doorway because they make applications and services that are not on IPv6 very fragile,” Palmer asked.
One concern for Daigle is that ISPs in North America are making great strides in IPv6 deployment, but that their rivals in Asia are lagging.
“Asia Pacific ran out of IPv4 addresses a couple of years ago,” Daigle said. “There are no network build-outs with IPv4, but they are not deploying IPv6. Instead, they are exploring vast workarounds of large-scale network address translation (NAT). Can we have a path for success of IPv6 if it’s not uniformly deployed across the globe?”
Brzozowski admitted that the user’s Internet experience may vary around the globe, with better and faster service in countries where ISPs deploy native IPv6 and slower service in countries where ISPs adopt NATs.
“It will be hard for [Asia] to avoid the tidal wave that is IPv6 transition without creating some definition between the Internet here and the Internet there because a plethora of services work in a degraded or broken state there,” Palmer said.
During the question and answer session, Erik Kline, an IPv6 software engineer with Google, pointed out that some Asian countries such as Singapore are making inroads in IPv6 deployment because regulators require ISPs to adopt it.
“The Singapore regulator put together a policy called ‘no islanding’ and used this to get IPv6 going,” Kline said. “This has not been without issue, but Singapore is in the top 12 countries for IPv6 deployment because of that regulatory foresight.”
Another audience question prompted a discussion among the panelists about whether or not it makes sense to market IPv6 directly to end users so they can ask for it from their ISPs and hardware vendors. Panelists disagreed about whether that was a good idea.
“If we waited for people to ask for IPv6, people’s dream of deploying it by 2020 would be 3020,” Brzozowski said. “Early on, we said we must have IPv6 on by default.”
However, Palmer pointed out that IPv6 offers bottom-line benefits for gamers due to reduced latency. “There are a couple benefits of IPv6 because the complexity and fragility of STUN [Simple Traversal of UDP through NAT] is a real thing,” he said. “UDP is not universally implemented and it’s not implemented well. The complications of STUN have a real impact on people’s experience.”
Cisco Fellow Mark Townsley suggested that instead of marketing IPv6 to end users, it might be more appealing to market the idea of globally unique addresses to both end users and regulators.