It’s World IPv6 Launch day, but will your organization need an intervention to move to IPv6? IPv6 adoption has been remarkably slow—the Internet has not yet hiccupped from a lack of IPv4 addresses. Still, it’s never too early to start planning for an IPv6 migration.
No IPv6 migration discussion is complete without a comparison of the IPv6 address space to other, irrelevant populations. The address space is simply huge, and it's pretty safe to say it won't be exhausted in 30 years. Having that much address space does, of course, have some interesting side effects. What size will your ISP give you at home? A /64 will net you 1.84467E+19 addresses for your home use. That will be sufficient for a while.
Regarding the last column of results? "Total market dollars predicted by research firms for new products" is tongue in cheek. Sort of.
Despite the notification that IPv4 is nearing the end of its useful life, few organizations are actually moving to IPv6. Only 13% of respondents to an InformationWeek survey are running IPv6 at all, and 23% are looking at another year or two before embarking on an IPv6 migration.
The bulk of enterprises, 38%, have no IPv6 deployment plans. None. It makes you wonder what they'll do when forced to move to IPv6. Granted, there are a number of Henny Pennys pointing to the sky but, as yet, nothing has fallen. If you fall into that group, I'd suggest you start figuring out what to do now so when your IPv6 migration won't be a fire drill.
Moving to IPv6 need not be that big of a deal. The basics are very much like IPv4, with some new twists. Our IPv6 Networking Resources page is a good place to start learning about IPv6.
The good folks at the University of California, San Diego Cooperative Association for Internet Data Analysis (CAIDA, for most of us), created the IPv4 and IPv6 AS Core AS-Level Internet Graph in 2010, showing, on the left, the interconnections between autonomous systems (AS) via IPv4, and IPv6 on the right. Autonomous systems in this graph correspond, roughly, to ISPs. The IPv6 graph looks sparse, but that's because of the relatively few networks running IPv6.
Researchers found an 84% growth from 2009 to 2010 in IPv6 AS, compared with 22% growth in IPv4 during the same period. The growth is relative: The IPv6 AS grew by 433 new IPv6 autonomous systems in 2010, compared with about 5,000 for IPv4--but growth is growth.
Growth in autonomous systems is one thing, but what's running on IPv6? According to Hurricane Electric, the number of IPv6 AAAA DNS records in the top-level domains is small, making up less than 1% of all domains (with the noted exception of .de in Germany). Of those AAAA records, an even smaller number of IPv6-enabled DNS glue records exist. That means that some domains have IPv6 records added, but there's no way to reach them. It's parking, of a sort.
According to Alexa, many popular sites (including Facebook, Google, Vonage and our own InformationWeek and XKCD) have IPv6 addresses. Network Computing is partially IPv6-enabled. The main site is reachable via IPv6, but the media and ad elements are still IPv4 only.
Of the organizations running IPv6, the majority of the deployments are in the network core and for publicly available resources. ARIN CEO John Curran has said, "A request for a public Web service is a request for an IPv6 service." And that makes sense. If your organization wants to be relevant on the Internet, you might as well start adding IPv6 to your public-facing services. You can do this using a dual-stack technology where your servers, load balancers, firewalls or whatever the public connects to runs both IPv4 and IPv6 simultaneously.
If you're using a load balancer, you can run dual-stack IPv4 and IPv6 on it, and use address translation for your existing IPv4 resources. If you start now, you can iron out any transitional problems before they become mission critical.
Of the 255 respondents who told InformationWeek that they aren't deploying IPv6, a lack of business need was cited as the top reason for the delay. It's true. There's no need to drop everything right now and move to IPv6. There is no hard deadline, real or imagined, looming that will force a mass IPv6 migration. This is the No. 1 issue facing IPv6 adoption. Everything is working under IPv4, but let's see how that goes when end users can no longer connect to Internet hosts.
Do we get away from network address translation (NAT)? Not on your life. Like it or not, NAT is here to stay, and in IPv6 that means NAT66--which just translates the IPv6 address prefix and doesn't do port translation. Blogger Tom Hollingsworth has a long and detailed post, "IPv6, NAT, and the SME - A Response" on why NAT66 will be with us--despite his objections that NAT masks underlying design problems. Hollingsworth is a practical guy, and while he'd like NAT to disappear, he says he does see value in some of the use cases.
IPv6 has no notion of RFC 1918-type addresses, so each one is globally routable. The Internet Engineering Task Force quietly killed off site-local addresses, which were meant to provide address space independent of the global address space in 2004. The reasons are listed in "RFC 3879: Deprecating Site Local Addresses." That doesn't mean you're handcuffed to your carrier because of the cost of an IP renumbering. Hollingsworth points out that organizations can get provider-independent address space from the regional Internet registries like RIPE and ARIN, but it also means you need to know more about how to use them.
Luckily, Hollingsworth says, renumbering IPv6 isn't as difficult as with IPv4. "RFC 4192: Procedures for Renumbering an IPv6 Network without a Flag Day" provides a method for doing so.
Of course, you can just use NAT66 and be done with it.