3

I've been learning more about IPv6, and am getting to the point where I'm going to be implementing an IPv6 lab to test various technologies that our company relies on, so that I can re-engineer them now, if necessary, for a future IPv6 switchover.

My plan is that in a year and a half, we're able to run fully IPv6 inside the network, and that we'll be running dual-stacks for client access. I've decided a year and a half so that I have 6 full months of testing and planning, and run the "hot" side of our infrastructure as IPv6 for 6 months while the "warm" side is IPv4, and after 6 months, convert the "warm" to IPv6. That will give me a testing, go live, and fall back point.

I'm interested to hear how other people are solving this problem, and what your roll-out plan looks like.

Edit
@Evan: My business reason is foresight. Eventually it's going to be necessary to have IPv6 if you want new network blocks. Eventually, my clients will be on IPv6. Eventually, everyone is going to be on IPv6. I want to convert before we're forced to convert, and I want to be able to do it on my terms, rather than under pressure from some regulatory agency.

Edit 2
Gerald Combs makes a great point. Emerging markets are not going to be able to get IPv4 blocks in the volume that they'd need, so at a point in the near future, and far sooner than established infrastructures, they're going to be using IPv6 regularly. Anyone with international clients or in a market that shows growth potential in the world economy may want to step it up.

1
  • We might have a router that will route it...
    – squillman
    Aug 11, 2009 at 20:11

7 Answers 7

3

We have several web sites hosted externally and a business DSL connection at our office. Our hosting provider (SoftLayer) recently added support for IPv6 and our DSL provider (AT&T) is still IPv4-only.

Taking a lead from Google and Netflix, I'm in the process of adding IPv6-specific hostnames for our public-facing web sites. Once I'm satisfied that everything is working OK I'll add AAAA records for each standard (www) hostname.

I'm not sure when we'll add IPv6 to our office network. I'd rather have indoor plumbing (natively-routed IPv6) than have to resort to outdoor plumbing (some sort of tunnel). There's a visible need to support 6 on our public sites since we get visits from regions that may be IPv6-only in the next couple of years. The need for 6 is less apparent in our office.

1
  • You make a great point about the emerging markets, and how new IP blocks in those places are likely to be IPv6 only, and very soon. Thanks! Aug 11, 2009 at 20:09
3

Perhaps my head is stuck firmly in the sand (or someplace worse!) but I haven't seen any compelling business reasons, as of yet, to move any of my Customers over to IPv6.

What are the business reasons driving your desire to migrate, out of curiosity?

My Customer profile, SMB/SME companies who have public-facing servers typically hosted out of off-site data centers, is such that there isn't a driving push toward IPv6. Typically they don't need any quantity of public IP address allocation and, though I'm not a fan of the current NAT-everybody-behind-a-single-IPv4-address culture that has grown up, that's typically what they're doing.

BTW: I'm not one of those who thinks firewalling is NAT. It would be pretty cool to be able to give everything on a network a public-allocated IP address. It's not something that my SMB/SME Customers are going to see any business benefit from, however (at least, not yet).

Address-space exhaustion isn't going to drive IPv6 on the Internet, IMO. Consumers don't care about end-to-end and many ISPs and media companies actively want to see the end-to-end principle degraded by NAT. I'm not envisioning wide IPv6 adoption any time in the next 5 years unless some radical application can drive it.

Edit: To be clear on my position, I think great reasons that IPv6 is good for the Internet, but it may not make sense for many of the business end-users of the Internet.

If I were deploying public-facing services with any regularity, or developing products, I'd all over IPv6 support from day 1. New product and service offerings shouldn't be constrained to an IPv4-only world because you'll just be building in obselescence.

6
  • 1
    I think it's going to be that IPv4 is eventually going to be treated like a second class citizen. The v4 netblocks are almost expended. You can count on providers sitting on them, but eventually they'll get so expensive that everything new will be v6. It probably won't die completely for a long time, but it will be snickered at, and inaccessible for many. Heck, some people still use netware. Aug 11, 2009 at 20:08
  • I'm not sure 6 will catch on any time soon either, but adding IPv6 addresses, DNS, and firewall rules to our public-facing sites has been pretty trivial so far. Aug 11, 2009 at 20:12
  • 2
    @Matt: I'd think that until ISPs start offering IPv6 connectivity at the consumer level there's probably little chance of the "LAN on the end of a DSL line" (which most of my Customers are) type networks needing to be IPv6 internally. I'd have a hard time justifying the expense to my Customers at this point. Any transition is going to take years to happen, if it happens at all. Aug 11, 2009 at 20:17
  • I'm from the time where everyone on the internal network had a public facing fully routable v4 address... anyhow, features will drive v6 - nothing else. As soon as there's business valued features that need v6, we're good to go. Microsoft Direct Access might be one driving factor as it needs v6 internally - but it also needs Win7 and Server 2008 R2 which isn't widely adopted yet ;) Aug 11, 2009 at 21:14
  • @Evan: I'm not sure ISPs have to start offering IPv6. Microsoft and Apple are making it easy to completely bypass your ISP with Teredo and the Airport products. Also, why aren't you treating this as a source of billable hours? What's wrong with you, son? :) Aug 11, 2009 at 21:52
2

Due to being a larger higher-ed that was interested in this internet thingy at the right time, we have a class B. Yes, our allocation pre-dates CIDR. We have no shortage of IPv4 addresses and even use v4 addresses for all devices, even our core, run-about-scream-and-shout-if-they're-down servers (except those subject to PCI standards which have to be NATed).

Just because you can resolve the IP address of my administrative workstation doesn't mean you can actually GET to it. This is something that v6 emphasizes, even though most security professionals get squinty at the idea of defending a network with full visibility to the public internet. And by 'full visibility' I mean the IP space is mappable even if the addresses are not reachable.

We will move to v6 when the State mandates that we must, or the pain of staying pure v4 becomes too great. We're state-supported, so if the Office of Financial Management says we must move to v6 by a certain date, thus it shall be so.

2

I work at a medium sized web development company. Our intranet has been running dual-stacked for about a year now. Our internet provider doesn't supply IPv6 yet, so we're using a SixXs tunnel, which works fine.

Our public servers are hosted at a colo, and have native v6 connectivity since 6 months. All we had to do was ask, which kind of surprised me. Not all hosters are this far yet.

I had a few years experience with SixXs and IPv6 on my home LAN, which helped a lot when implementing it at work. Thankfully management understands that as a webdev company we need to support and experience many different technologies, even though the RoI may not be evident yet. Our reasons were the same as yours: foresight and doing it on our terms. Your rollout planning seems sensible.

Our experience has been mostly positive. Starting early has helped my programming colleagues to get used to it, and adapt our frameworks and apps to support it. Things like IPv6 address validation in PHP, e.g. for input validation or for access whitelisting, are harder than we first thought so we're glad to have that done. PHP's built-in or publicly available functionality were mostly flawed.

A problem we ran into is that if IPv6 connectivity is down most hosts will take ages to fallback to IPv4, and some will never fallback. Since they prefer v6 over v4, they simply won't connect. Some apps have a switch to force v4, but many don't. Make sure to test this in your lab.

1
  • Thanks. Great comment, I really appreciate your shared experience. Aug 11, 2009 at 23:13
0

I hope some folks are going to move... because we (SmoothWall) are going to be taking all our products IPv6 next year - maybe it won't payback straight away but I do think we'll get there.

1
  • Definitely. IPv6 firewalls are in short supply now. I use Shorewall6 on servers, which is one of the very few options available. Aug 11, 2009 at 21:27
0

I work for a small startup. We have no formal plans to include IPv6 support in our product, but I see that it would be prudent to do so sooner rather than later. I've quietly been coding support for it that I can unveil at the appropriate time. Presumably there are more like me at various firms.

0

Just glancing over the other replies I saw Evan's "Perhaps my head is stuck firmly in the sand (or someplace worse!) but I haven't seen any compelling business reasons, as of yet, to move any of my Customers over to IPv6." That is how I see it, and how I've seen it ever since the great "we're running out of IP addresses" issue was first raised. To me it's not too unlike the '85 "energy crisis".

Should the need (real need, not just an imagined one) to run IPv6 arise during my working life I'm not at all concerned because anything I work with that could conceivable need to deal directly with IPv6 can already do so. Other than email the only thing we have that needs to talk directly to other IPv6 systems is the firewall. The web sites are hosted by a third party, so that's their problem and I'm comfortable with their ability to support IPv6. Internally nothing needs IPv6. I don't need the extra hassle of supporting an unnecessary layer and the company doesn't need the cost of replacing those items that can't deal with it when there is absolutely no reason to do so. Let's face it, nobody's internal network is so large that IPv4 can't cope and the Internet facing devices shouldn't be in the internal network.

Not the answer you're looking for? Browse other questions tagged .