Martin Levy

Enabling IPv6, one step at a time

Internet technologist Martin Levy recommends a steady approach. But don’t wait to get started.

Martin Levy, director of IPv6 strategy for Internet carrier Hurricane Electric, doesn’t pretend that implementing the new Internet Protocols on a network is easy, but it will be necessary and does not need to be impossibly difficult.

Levy is an Internet technologist with extensive experience in international IP networking. For more than 16 years he has built regional ISP offerings in California, worked with European telecom companies and done business and technology planning for IP networks in Latin America. At Hurricane Electric, which began implementing IPv6 in 2001, he focuses on global IPv6 backbone services.

He spoke recently with GCN’s William Jackson.

GCN: What is Hurricane Electric?

LEVY: Hurricane Electric is an Internet service provider, wholesale IP backbone. The company is based in California, in Silicon Valley, and we take large IP connections from anywhere in the world and, with our extensive peering, have the ability to transport and swap traffic. We are not an ISP with consumer offerings. We aren’t normally known by the general public. The company also operates large data centers for Web hosting and co-location in California.

Why did Hurricane begin deploying IPv6 in 2001?

We have a focus on technology, we look at capabilities; we don’t have an interest in staying still. In 2001, we had the ability to enable IPv6 in a very rudimentary way on our IP backbone. This is a classic technology issue. If you go back and look at what you were doing nine years ago, you’d laugh. But if you hadn’t done it, you wouldn’t be where you are today. So we started in a simple manner to move some IPv6 traffic on our backbone. There was very little of it, there was very little support for it in routers or operating systems or applications. But we were able to get that early experience.

You have to come forward to about 2006 or 2007 before we could truly say in any commercial sense we could move IPv6. This was fundamentally because we had done a technology upgrade, and we had acquired hardware that ran IPv6 properly. That was part of the [request for proposals] process for buying it.

The most important takeaway from that is that around 2006, vendors started delivering both the hardware and the appropriate software or firmware that runs on it that could be called operationally acceptable IPv6 technology. If someone wants to do IPv6 today, the question is: When did you last do a technology update on your network? If you’ve done it in the last three to four years, you may already have what it takes to run IPv6.

Why should IPv6 be adopted?

Do you want to still be in business in two to three years time? That is the question. If you start looking at the issues of IPv4 address allocation, you see that if a network is not IPv6 enabled now or very soon, it will start having issues connecting to parts of the Internet that are starved for IPv4 addresses.

The biggest reason why this is so fundamentally important is that there are going to be resources on the Internet that are not going to be able to get IPv4 space. There is going to come a time when services that people want to deploy or access will become impossible to reach on a v4 platform. And that time isn’t tens of years down the road. It is in the next couple of years or less. That is going to be a fundamental driving force.


Related stories:

Agency IT plans could hinge on IPv6 implementation

Security will not come naturally with IPv6


How much IPv6 traffic are you handling?

It’s still miniscule compared to the amount of v4 traffic flowing on our backbone and other backbones. There is argument over what that percentage is — is it a hundredth of a percent, or one percent of traffic? There are a few cases, such as Google’s efforts in the v6 world, and Netflix has done work as well. Obviously, on the government side, there are networks out there using it. But the reality is that we haven’t yet hit that ubiquity such that the traffic measurements truly make sense. We find that the traffic is becoming more generic, it is becoming similar to v4. We see Web access, we see [Domain Name System].

Who is using it, and how is it being used?

We see some enterprises that are forward-thinking that are using it. We definitely see research and education networks. We see some government usage. We also see efforts by large IP users, such as the work that YouTube and Yahoo have done, that are great examples of companies that have taken the effort to understand how to implement v6 in the real world. We see lots of smaller companies and telecom providers around the globe that are the dominant part of our customer base who are getting into v6, whether it is by company mandate or simply by the realization that this is now the right time to do this. But without focus on the robustness, reliability and ubiquity of these things, you can’t consider real usage beyond taking it as a science project.

Where is the United States in the adoption of IPv6?

It depends on which metric you want to use. If you use the number of IP backbones that are enabled for v6, we’re getting there. It’s possible that Europe is slightly ahead of us. There has always been a perception that Japan has been IPv6 ready far longer than anyone else, and that may be the case for certain networks in Japan. But I think we’re doing pretty good in the U.S. If you measure how many end-users have access to v6, then we’re not as far along as other countries. But it’s very spotty, and it is also a case that the statistics may be too small to use for an accurate trend.

But I think we’re going to see major changes as we see consumer networks like Comcast, Verizon, AT&T start to adopt IPv6 not just for trial but for general usage by end-users. I also think that as a country, we are fairly far ahead in research and education networks in the university and the higher education environment.

What lessons have you learned in using IPv6?

Nothing is easy. Don’t get me wrong; I will always paint a positive picture. But I am fully aware that sometimes things take a long time.

There are two major lessons. The first one is that there is a fundamental difference between enabling v6 inside the core of the network [and] enabling it in all parts of the network in a ubiquitous manner. The second lesson has to do with the need to ensure that the network is as robust in the v6 as the v4. Is traffic routed with the same quality and robustness that is exercised in the classic IPv4 Internet? Is it usable? Is there support from vendors and backbones with v6 offerings that is as good as their v4 offering?

We got to review for a large telecom a 50- or 60-page IPv6 enablement document. The problem was that the company hadn’t turned anything on with IPv6. They had only written a document. What they focused on was getting v6 to every corner of their network, getting it enabled for every type of user, whether it was easy or hard.

We rewrote that document for them as a three-page document, and it said this: Inside your test lab, enable v6. Secondly, look at only the core of the network, the major routing elements, and work out how to enable v6 inside the core. Then, look at a couple of simple services to enable. The simplest one we came up with was their DNS services because DNS has been unbelievably well tested for v6. None of this was customer-facing, but it had the ability to get the key thing: experience with IPv6. What we told them was, take the first step and then stop. Think about what you’ve done. I’m not saying that it’s simple, but it also is not blindingly complicated.

Will IPv4 ever be abandoned?

No. No, no, no. Let me repeat, no.

What are the challenges to operating both IPv4 and IPv6 on networks?

Experience will show that this is not that hard. As an analogy, you can remember what it was like 10 or 15 years ago trying to find that gas station that actually had diesel. Now you go to the gas station, and it’s the same pump. The basics of v4 are different protocols. But by the time you are finished, you will see that they do essentially the same job and need to be operated on the same platform. I am politely ignoring transition protocols here, which are more in the realm of how cable and DSL and connectivity companies have to operate. But if we look at the way operators of networks will use it, v4 and v6 will become identical items.

Hopefully, IPv6 will become ubiquitous in the environment where v4 will still exist. It is the copious number of end-users who will probably be connected only with externally working v6 addresses, and it is quite possible that some services may not be reachable with IPv4.

Reader Comments

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above