Miser Internode is hoarding IPv4 addresses

39

Internode is ready to battle what is expected to be a global shortage of Internet Protocol version 4 addresses during 2011.

In the last 24 hours, global concern has arisen as the Internet Assigned Numbers Authority (IANA) allocated the last two free blocks of IP addresses to the Asia-Pacific area, triggering off apocalyptic rumours the like of which the global technology community has not seen since the year 2000 bug struck a decade ago.

However, national broadband provider Internode this morning said it held stocks of IPv4 addresses, and would be able to respond to projected customer demand for at least three to five years, adding that in the meantime the company will be working on the actual deployment of the new IPv6 protocol, which will resolve the problem.

In fact, Internode revealed it will progressively move its existing national IPv6 trial to a full production service in 2011 in order to face the depletion of the 32-bit IPv4 address space, predicted for the next few months. The company’s managing director Simon Hackett said Internode would be hoarding its remaining IPv4.

“The exhaustion of new IPv4 allocations from global internet registries is a little like ‘peak oil’,” Hackett said. “There won’t be any more large ‘discoveries’ of IPv4 addresses, so ISPs around the world will now mine their existing stocks to exhaustion point over the next few years,” he said.

The transition to IPv6 will be managed by allocating “two kinds of Internet addresses” to new customers; a practice which — Internode promises — will ensure customers a smooth pressure-less migration from IPv4 to IPv6. From a hardware point of view, Internode says it will make sure all the new routers are IPv6-compliant.

“We’ve ensured our customers have the best of both worlds, so they don’t need to worry about this issue during the transition period. Internode customers won’t run out of IP addresses,” Hackett said.

Although customers using IPv6 will find the new address very different from the familiar four numbers used by IPv4, Internode said most users wouldn’t notice any practical change; Access to web pages will still use domain names which computers will translate into IPv6 or IPv4 as appropiate.

““If the Internet engineering community gets this right, our customers won’t notice anything at all. Internode is getting this right, so its customers are future-proofed across the transition,” Hackett said.

During the current year, interested customers with Ipv6 compatible equipment could try out the new protocol by learning how to turni it on, through an Internode guide available online. By the end of the year new customers will be allocated to IPv6 automatically and already existing ones will be invited to turn on the new protocol.

Other ISPs such as Telstra, Optus and iiNet are being contacted for a statement on the IPv4 exhaustion issue.

Image credit: riptheskull, Creative Commons

39 COMMENTS

  1. In think the attitude to IPv6 from every other ISP in Australia has been pretty hilarious.

    • You mean the “she’ll be right” or… complete lack of response?

      He’s right, I haven’t noticed any changes, and neither have my family, ever since running IPv6 on my network.

      Which is a good sign. When the this hits mainstream media and they ask me about it “Oh yeah that, we’ve been running IPv6 for the past 18 months.”

  2. Sorry, but when did prudent forward planning turn into ‘hoarding’?

    When did reassuring our customers that we have planned ahead turn us into scrooge?

    C’mon Renai. Its obviously a slow news week when you have to find such a negative spin to a good news story.

    • Hi Simon,

      thanks for your comments!

      I am fully cognizant of the gravity of the IPv4 situation, and I understand that Internode is taking prudent action in this case to ensure no service interruption to its customers during the IPv4 to IPv6 changeover period.

      In addition, I am grateful that the company has publicised this matter — it has brought a level of debate to the issue, and we’re currently following up with your rivals to find out what they’re doing.

      However, I would disagree that we have put “negative spin” on the story. Delimiter isn’t a deathly serious publication like the Financial Review, and we often take a more light-hearted look at things than some of our fellow publications.

      A good example might be our “Oh Dear” series of articles; but we try and maintain that style of good humour throughout the entire publication. We also often run amusing pictures with straight news stories to inject a bit of colour. This sort of style is also used by a raft of other publications, but in Australia and globally — examples I can think of would be The Register, Gizmodo, TechCrunch, Engadget and so on.

      Cheers,

      Renai
      Publisher, Delimiter

      • Sorry Renai but Simon is spot here. Even your title is inflamatory: “Miser Internode is hoarding IPv4 addresses”. You use two seperate negative words: Miser and Hoarde. Not nice stuff to throw at one of the very few ISPs taking a proactive approach. If anything Internode should be lauded for taking the initiative but apparently not by Delimiter – which reflects poorly on both you as a journalist and Delimiter as a publication.

        • As an avid follower of Delimiter and a loyal customer to Internode, I did not find the title of this article offensive, nor inflamtory.

          The word “miser” in this context is used ironicly, because the fact that Internode is proactive enough to purchase a stockpile of IPv4 addresses indicates that Internode is far from cheap and it is actually well prepared.

          Also “hoarding” is not a negative word, oddly in modern English it seems to have been given a negative corelation in favour of “collection” althrough I cannot fathom why. It means only to amass a collection of vaulable objects.

          So, if we combine the word hoarding, with the word miser, and couple that with a resource that is short supply and thus high value both in terms of prestiege and practical use, I find this title to actually be a compliment to Internode.

          But maybe the fact I hang out with linguists has rubbed off on me and I should be offended, I don’t know.

          • Then we must agree to diasgree. I too am with Internode and read Delimiter and I remain unamused by Renai on this one. Simon was righ to call Renai out on it and Renai’s attempt to spin it just seem weak.

      • That may be ‘cognisant’ perhaps?

        If you want to see the real hoarders take a look at
        http://en.wikipedia.org/wiki/List_of_assigned_/8_IP_address_blocks
        American Defence holds 20% of the worlds ip addresses, Apple holds 16.7 Million addresses, Prudential Insurance of America holds 16.7 Million addresses.

        Internode has been prudent and is protecting its customers while being proactive in moving into IPV6, so as one of those customers I felt reassured by Mr Hackett’s announcement.

    • Good to see Internode planning ahead and looking after their customers. It appears like much of the rest of the world has been asleep on this issue, waiting until all IPv4 addresses are gone before bothering to act.

      A bit of a silly title though.

  3. Personally, I can’t wait for IPv6 to become ubiquitous. Good-bye NAT, non-static IP addresses and all those other kludges that we’ve had to live with up until now.

    Question for Simon, if he’s still reading the comments… how many IPv6 addresses does Internode assign to customers when they sign up for IPv6? I can imagine you’d be able to give everybody a nice big chunk of addresses: 32, 64, even 128 – one for every device they own.

    • Somewhat more then that- you get a /64 allocated to the PPP session plus a /60 for Prefix Delegation.

      So thats 18,446,744,073,709,551,616 address to the PPP session plus 16 times that for PD :)

    • Internode won’t be giving out statics IPv6 for non-SOHO plans, it doesn’t scale (routing table wise) given the way they load balance / failover DSL connections.

      They give out a dynamic /64 plus a static /60 at the moment. To use auto configuration (so everything Just Works(TM)) a /64 (18,446,744,073,709,551,616 addresses!) is split out from the /60 for your LAN interfaces.

      • Dynamic reallocation is an important part of IPv6. Provided that addresses are within the same /64 you can use the fe80::/64 series of addresses for routing, and any internal network traffic will transparently roll over when the prefix changes. This is by design, and quite well thought out. You can even have two prefixs assigned to the same network while you are phasing out the “legacy” prefix over a period of a few hours.

        Therefore the only time you need a static prefix is if you want stable external routing into your network, say for hosting a server, which is the same as IPv4. I’d go so far as to say that IPv6 handles dynamic allocation far more effectively than IPv4.

        Plus having a dynamic “pool” of addresses for an ISP is far more cost effective than having allocating a static address to each client.

  4. You should probably ask Adam internet what they are doing with their ipv6 rollout as well. I noticed that they have established ipv6 BGP sessions at various PIPE IXs. Not advertising any routes though when I checked a few days ago.

  5. Yeah, Sorry, No Dice. I agree with Simon’s statement. The article itself is a good news piece, the title is misleading and has a definate negative twist. I dont usually see this type of thing on here. I’m glad a read the article, because from a glance it appeared that Internode had made some indication that it would not release IPv4 addresses back into the wild. Dissapointing to say the least, I guess I’ll add this publication to the “Take with a grain of salt” list.

  6. APNIC might be interested to know that internode have been allocated enough IPv4 space to last for 5-6 years. Seems to fly in the face of current APNIC IPv4 allocation policy.
    Initial allocations are based on two year requirements, but subsequent allocations should only suffice for one years needs:
    http://www.apnic.net/policy/add-manage-policy#9.4
    Criteria for subsequent allocations
    “APNIC and NIRs will allocate enough address space to meet the LIR’s estimated needs for a period up to one year.”

    I assume that this also means Internode has excluded itself from applying for any further IPv4 address space from APNIC?
    http://www.apnic.net/policy/add-manage-policy#9.5
    “An LIR is not eligible to receive subsequent allocations until its current assignments account for at least eighty percent of the total address space from all allocations it holds. This is referred to as the “eighty percent rule”.”

  7. You might also be intersted to know that APNIC can only issue IP blocks, and tradiontally only gives an /8 or /16 to ISP. As there are about 250x more addresses between a /8 and /16, you kinda have to “overorder”.

    This is actually one of the primary reasons that IPv4 addresses are becoming so scare. They weren’t designed to be used on such a scale. Hence why the IPv6 standard is overkill in the extreme for IP address requirements at 2^128, because it is easier to allocate addresses in blocks than it is to issue a specific amount of IP addresses.

    So, it doesn’t fly in the face of APNIC’s policy at all.

    • Sorry to burst your bubble, but you might be interested to know that all RIRs allocate based on CIDR (classless inter-domain routing).

      http://www.apnic.net/services/services-apnic-provides/helpdesk/faqs/obtaining-resources—faqs#ClassC

      IPv4 address ranges are no longer referred to in terms of classes (class A, class B, class C). Classful addressing methods, which waste address space, are now made redundant by Classless Inter Domain Routing (CIDR). In CIDR, address ranges are referred to by their prefix (or subnet mask) length. For example, APNIC’s minimum allocation, /22 (“slash 22”), refers to 1024 addresses or the equivalent of 4 former class Cs.

      The use of CIDR is a fundamental requirement for eligibility to receive IP addresses.

      • See my below post where I pointed out that even with CIDR you will still overorder. And I also challenge you to find an ISP that will order as low as /22 considering their projected customer base. /22s are usually only ordered by businesses.

    • To clarify, althrough it is traditional to release in and /8 /16 and /24 some authorities and ISPs, including APANIC, also issue in other blocks, i.e. /10, /12, /18, /20, /22, etc, however, since those blocks are still 4x as big as the previous block size, the same logic still applies. Even if you break it down to /9 /10 / 11 etc… that’s still 2x.

      Let me put it this way. You project you need 5 apples for the next week, but you can only buy in powers of 2. That means you can order 4 apples, or 8 apples. So you order 8. This problem only gets worse as you require more apples. Say you need 18 apples, but you can only order 32.

      I’m simplifying a little bit, as you can order a collection of smaller blocks rather than the next largest block as well, however you will STILL tend to overorder.

      • Sorry, that’s incorrect.
        Let’s say you need 2000 more IP addresses over the next year than a /17 can provide.
        APNIC will allocate you a /17 plus a /21 (2048 addresses). They will NOT allocate you a /16, just because a /17 does not suffice.
        Please trust me, I have filled out the old AONIC IPv4 subsequent allocation form enough times over the years to know how it works :)
        It doesn’t really bother me that Internode have this much space (good luck to them), but I am surprised they would flaunt it like this.

        • In this example you overorded by 48 addresses. See. Thus proving my point. ;)

          You’re probably right, but it is entirely possible that Internode got a bit egoistical and ordered based upon demand initally and ended up overordering. Not that uncommon I think you’ll find.

          • Sorry NightKhaos, that’s not how it works at all.

            When you apply to APNIC, you provide them your expected customer numbers along with historical data, and they decide what they will give you (with a little negotiating room). You don’t decide yourself (directly) what address space you’ll order (oh, I think I’ll have a /16 this time).

            They also will allocate on any boundary (/17, /19, whatever), to match as closely as possible to the demonstrated need for the coming 12 months. No preference whatsoever for /8, /16 or /24. Additionally, over the last few years they’ve been getting more and more strict and detailed on their analysis.

            Anybody who’s saying they have a 5 year stock sounds like either they’ve been gaming the system or aren’t telling the full story.

          • “provide them your expected customer numbers”

            So if you over estimate your expected customer numbers and you esitmate seems sound by APANIC you can’t be overallocated? Oh wait… you can and will be.

            “Anybody who’s saying they have a 5 year stock sounds like either they’ve been gaming the system or aren’t telling the full story.”

            I would bet there is more to this than you’re saying. As I said, they could be saving addresses through other means, or overstating their stockpile, or there could have been a sharp drop in demand for services since they last got an allocation, or they recently brough a networking company and coupled their pools, or they brought some addresses from outside APANIC from say HP (unlikely by possible).

            Which is entirely my point from the start. It’s NOT THAT SIMPLE to say that because they have a stockpile bigger than they should that they have been gaming the system.

          • No, actually your point from the start was, (and I copy and paste as well as the next person):
            “You might also be intersted to know that APNIC can only issue IP blocks, and tradiontally only gives an /8 or /16 to ISP.”
            bbbbbzzzzttttttt wrong.

          • Dude, seriously, that was uncalled for.

            I do remember what I said, I can read, and I can copy and paste it. However, that wasn’t the point I was trying to make. Dude to the lack of the edit function, did you not my clarification? Did you not me going realise “oh shit, I forgot to include CIDR and this guy knows anything about IP allocation, which he probably does considering he quoted APANIC policy, he’s going to hound me for it.”

            The point I was trying to make is that, and look, I’ll copy it for you, from my first post:

            “So, it doesn’t fly in the face of APNIC’s policy at all.”

            In other words, I don’t think Internode have done anything wrong here when it comes to APNIC policy. And… if we look through my post there seems to be that theme doesn’t there?

            Thank you however for being an ass about it. I REALLY appericate it.

          • Sorry, that was a bit harsh, you’re right.
            My frustration is that you keep replying with incorrect statements. For example in your last post you say Internode may have bought space from someone else like HP. IP address space cannot be bought.
            APNIC and the other RIRs have policy that is black and while, you can read it yourself. There are also several mailing lists like sig-policy and apnic-talk where this stuff gets discussed if you would like to become familiar with APNIC policy. It is mostly very dry, but can be quite interesting.
            I like Internode as much as the next guy, but under current APNIC policy, they should not have enough address space to last for 5 years. End of story.

          • I know I haven’t be exactly… accurate in my replies here. I have been kinda stressed over the last few days and accuracy hasn’t exactly been my strong suit. Let me try and correct a few misunderstandings.

            “Internode may have bought space from someone else like HP. IP address space cannot be bought.”

            Well how else would you describe such a deal? It’s technically reallocting the address space for use by a provider, so it could be “renting” the address space.

            You do realise that deals where big companies who have stock piles of addresses, before CIDR came on board, will have to start selling/leasing/whatever-you-want-to call it there extra address space as it is the only way, once a particular authority have run of allocation to give, for new IPv4 addresses to enter the pool? If we just “leave” them with their allocations there is going to be way to ease ourselves into the transition to IPv6 properly.

            “it is entirely possible that Internode got a bit egoistical and ordered based upon demand initally and ended up overordering”

            I did not intend to mean here that they were intentionally overordering, just that they looked at the projections they had, ordered an allocation, and then something happened like say the GFC hit, or Telstra came up with some deals and undercut them, and their predictions for growth, reasonable according to them and APANIC, were unfortunately wrong and they ended up with a stockpile that they didn’t need.

            “I’m simplifying a little bit, as you can order a collection of smaller blocks rather than the next largest block as well, however you will STILL tend to overorder.”

            See I was quite aware that you could a series of smaller blocks to make a specifc size, but always within an error margin. Granted, I overestimated the error margin, and you did point out to me just how accruate they can get, and I did attempt to acknowledge that and point out that there are other ways that you can overover as well.

            I thought the fact that I said “thus proving my point ;)” was enough to indidicate this to you, however I’m sorry if it wasn’t.

            “/22s are usually only ordered by businesses.”

            This was stating that ISPs will hardly ever project such a low uptake of customers unless they are postively tiny. A /22, which consists of 1024 addresses, will be exhausted too quickly by an ISP. In practice I would expect their usage predictions to be rounded to the nearest 10K, not 1K, so they would rarely order low enough to get allocated a /22.

            It’s important to note, and this is kind of the premise of my entire agruement, that the number new of Internet Subscriptions in 2010 was way down compared to the previous year?

            According to the IWS website, there were an extra 2 million new Internet users in Australia between 2007 and 2009, however between 2009 and 2010, the number of new users was barely 100,000. It is not inconceivable, given this, that an over order took place by Internode, resulting in them having a stockpile. http://www.internetworldstats.com/sp/au.htm

            Maybe this could explain the over allocation? I don’t know. And I’m sorry for not being compeltely clear and concise with my statements. I will endevour to be more accurate from now on, however I didn’t think it mattered before, because it’s only an opionion comment, about a little discussed and not well known issue. Which was a foolish assumption to make, because obviously it is important to you, and you consider it a serious issue. I must apologise for this.

Comments are closed.