Today's ARIN estimated depletion date:


Looking back at 2011

2012-01-02

The top stories for 2010 was that some major sites and ISPs turned on IPv6. What was the top 10 IPv6 and IPv4 depletion stories for 2011? Did I miss any significant story? Please comment.


1. IANA runs out of IPv4 addresses

In February 2011, IANA runs out of addresses. This was much earlier than most people predicted. People start to realize that they must start looking into IPv6.

2. APNIC runs out of IPv4 addresses
In April, just two months after IANA runs out of IPv4 addresses it is the first RIRs turn. APNIC serving Asia Pacific and Asia is down to their last /8 and the more restricted policy stats to take effect. They can no longer allocate IPv4 addresses the conventional way.

3. IPv4 addresses for sale
Microsoft aquires a large chunk of addresses from what is left of Nortel. They pay $11.25 per address. People are scratching their heads woundering why Microsoft just didn’t go to ARIN and request the addresses.

4.World IPv6 day
On 8 June, 2011, ISP’s and content providers joined together and turned on IPv6 for a day. The event was a success and very fee issues were found. This attest that IPv6 is ready for prime time.

5. Decrease in IPv4 allocation rate
The allocation rate of IPv4 slows down after APNIC’s depletion. More restrictive policies are in place at the RIR’s and we have not seen and really large allocations in either ARIN or RIPE.

6. Godaddy enables IPv6 DNS
Godaddy enables IPv6 for their DNS servers. 25% of the domains in the world are all of the sudden accessible over IPv6. (Note, that this relates to DNS servers, the web hosting service at Godaddy is still on “Internet classic”)

7. The realization that dual stack networks brokenness is minimal
World IPv6 day and other measurement efforts around the web concludes that the worries about IPv6 brokenness are exaggerated. It can be concluded that turning on IPv6 will not cause any significant loss in website viewers for most companies.

8. Vendor support
With the depletion of IANA and APNIC a lot of vendors jump on the IPv6 bandwagon. Most Firewall, Load balancer, DNS, etc vendors are issuing press releases about how they now support IPv6. They interesting questions will be if the support is as good as they claim.

9. Lack of IPv6 strategies from some large organizations
Just like in 2010. Some large organizations still seems to lack an Ipv6 strategy. Did they read the news at all?

10. Nokia ditches Symbian
Nokia changes strategy and ditches SymbianOS. This creates problems for the Wireless Carriers with Ipv6 plans, since symbian actually had the best IPv6 support of all phone platforms.

Tags: , ,

39 Responses to “Looking back at 2011”

  1. jmv2009 says on :

    Looking at

    http://www.potaroo.net/tools/ipv4/index.html,

    It looks like arin and afrinic slowed down. Ripe and especially lacnic largely continued with business as usual.

    Ripe is a bit slow, but could be accelerating again. Basically the change in policy caused a 9 months(12 months – 3 months) delay in exhaustion.

  2. Karen says on :

    There appears to be an issue with that site since some of the stats hasn’t updated since the 15th December 2011. RIPE’s estimated date has stated the 22nd July all through Christmas etc.

  3. ipv4depletion says on :

    Hi Karen,

    Thanks for checking. I had some issues with the site for a while but they should all be fixed a couple of weeks ago. Can you be more specific on where you see 22nd July? I see 2012-02-24 everywhere I see?

    Thanks, S

  4. Karen says on :

    Hi

    I was refering to an issue with potaroo.net rather than your site. On http://www.potaroo.net/tools/ipv4/index.html I am seeing the RIPENCC date at the top as 22-Jul-2012 and the report genterated date of 15-Dec-2011 21:49 UTC

    Karen

  5. JHB says on :

    And I sent a mail to CS at he.net, asking to check the update of statistics. Got feedback it is repaired now

  6. ipv4depletion says on :

    The Potaroo site is still showing as being updated on Dec-15 for me. I know HE pulls their stats from Potaroo.

    /S

  7. Michiel says on :

    Today the Potaroo site, says it is updated jan 19. It now predicts 27-Jul-2012 as the RIPE exhaustion date.

    Still a big discrepancy with your 2013-03-06…

  8. ipv4depletion says on :

    Hi Michiel,

    There are three main reasons for the difference:

    1) If you look at Potaroo’s graph http://www.potaroo.net/tools/ipv4/plotend.png you will notice a big discrepancy between the thick (historical) line and the thinner (prediction line). For both RIPE and ARIN the difference is huge. When you do curve fitting you might end up with that problem, especially if your historical data doesn’t match the most current trend. What Potaroo is doing is that he is using a very large set of historical data for his prediction (you can see the thin lines showing up all the way to the left in the graph). I always move my prediction up to the point where it matches with the history to avoid such gaps. If you estimate what such similar move would do for potaroo, it would advance the RIPE depletion date with 3-6 months or so.

    2) Since the allocation pattern changed recently when all RIRs implemented a more restricted way of allocating v4 addresses this will in my opinion not give you accurate results. I only use 1 year of historical data so that I mostly get post-depletion historical data. Historically we burned 1 x /8 a month. The last year, we burn half of that per month. I don’t feel that it is relevant to use such a large set of data.

    3) Additionally, potaroo is using an ‘accelerating’ polynomial function for RIPE. If you look at the last year of allocations it is clear to me that there is no acceleration in the allocation rate in Europe. Instead I’m using a linear growth model (still a growth, but not accelerating).

    I’m willing to bet $100 that RIPE will not be depleted before July. Any takers?

    /Stephan

  9. Michiel says on :

    Stephan,

    Thanks for the explanation. I mostly follow your line of reasoning, except that due to the economic recession we may see some more slow down in the European burn rate.

    So, forgive me for not taking your bet, I expect to lose. ;-)

  10. Mike Dee says on :

    Well, I am not willing to put my money where my mouth is, but my current prediction is that RIPE exhaustion will happen on 21 May. I use a linear model starting at IANA exhaustion on 7 February 2010 using the weekly data published by RIPE.

    I then assume that at some point a run will begin on the last addresses. Using APNIC as a proxy, I determined that this run began 3 months before exhaustion, and it increased the allocation rate by a factor of 3.3.

    The average allocation rate over the last year has been 0.77m IP addresses per week (i.e. 0.2 /8s per month). My model assumes this allocation rate will continue until around 21 February, at which time allocation rate will increase to 0.6 /8s per month, leading to exhaustion on 21 May.

    One further reason to believe that there will be an increase in the allocation rate is that the last year was a time of burning the reserves, since the policy was changed from allocations sufficient for 12 months to allocations sufficient for 3 months. Once the buffer is used up, allocation rates should increase to the old level again.

  11. JMV2009 says on :

    Mike Dee,

    If ripe would have followed the pattern of APNIC, it would already be at the accelerated level: When apnic only had 4/8 left, twice what ripe has now, it was moving very fast. Historically, apnic was going through addresses twice as fast as ripe.

  12. ipv4depletion says on :

    Potaroo now seems to be stuck at 27-Jul-2012 for Ripe. It has not changed in a couple of weeks.

  13. Karen says on :

    The amount of data is still going down – so perhaps the data rate RIPE is currently going through matches the expected depletion rate.

  14. Andrew Bower says on :

    There’s an article on RIPE labs about IPv4 allocation statistics for 2011: https://labs.ripe.net/Members/mirjam/ipv4-allocation-statistics-2011

  15. Karen says on :

    Andrew – this no longer should stand true. There should now be a 3 month limit on IPv4 assignments. So it shouldn’t be such a big increase early in the year.

  16. Mindbuilder says on :

    In areas where people have been getting ipv6 addresses, have the addresses been nearly static, or do they change often? With the exception of mobile devices, are ipv6 addresses in the future expected to be nearly static, or will they be expected to change often?

  17. Karen says on :

    Mindbuilder

    Here are the stats that I have for IPv6 uptake per region. I have done two queries, 91 days (13 weeks) and 365 days. The item in parenthesis is the average over that period

    91 days:
    AFRINIC: 21 (0.23)
    APNIC: 125 (1.37)
    ARIN: 143 (1.57)
    LACNIC: 50 (0.55)
    RIPE: 279 (3.07)

    365 days:
    AFRINIC: 125 (0.34)
    APNIC: 608 (1.67)
    ARIN: 963 (2.64)
    LACNIC: 148 (0.41)
    RIPE: 1546 (4.24)

  18. MindBuilder says on :

    @Karen – I don’t think your numbers are the answer to the question I meant to ask, but they look interesting nonetheless. Unfortunately I don’t know what they represent. Do they represent uptake? If they’re uptake, what units are they in?

    What I meant to ask is whether the ipv6 address of typical home users have been more like a static address, or like the dynamic ipv4 addresses we get by dhcp from todays typical cable modem or dsl providers. I saw a post by one Australian internet customer who said that his ipv6 address has stayed the same for a couple years regardless of modem resets. My ipv4 address seems to stay the same accross most quick modem resets, but it has changed several times over the years. I know that ipv6 has an auto configuration feature to assign an IP address, but I’m wondering if service providers are commonly setting up their systems to auto configure their customers addresses the same nearly every time. I would think they would.

  19. Mike Dee says on :

    OK, according to my assumptions – which is that RIPE follows the same pattern as APNIC (i.e. usage increased by a factor of 3.3 three months before exhaustion), we have now reached this point. I would expect to see accelerating allocations starting around now, culminating in exhaustions around 6 May.

  20. JMV2009 says on :

    Apnic ran through 20% of the last 28 months in 45 days or so. Ripe also spent 80% of its last stock in the last ~26 months. So if Ripe follows Apnic, it should be done in 45 days or so.

    However, Ripe is currently not moving that fast currently.

  21. Grahm says on :

    > Apnic ran through 20% of the last 28 months in 45 days or
    > so. Ripe also spent 80% of its last stock in the last ~26
    > months. So if Ripe follows Apnic, it should be done in 45
    > days or so.

    Try removing China when looking at APNIC data?

  22. Mike Dee says on :

    In the last 3 weeks, RIPE went through 4 million IP addresses. The remaining stock (excluding last /8) is 32 million addresses. At current speed, they would be exhausted in 24 weeks, which is in August 2012. I still expect an acceleration as it is the rational thing (for the individual firms!) to wait until close to the end and then claim as much as possible. I think May or June 2012 is likely.

  23. Michiel says on :

    So Mike why don’t you take Stephan’s bet that RIPE won’t deplete before July? ;-)

  24. Michiel says on :

    “Kruidvat”, a large drug store chain in The Netherlands is now offering the Sitecom 300N wireless router from €39.95 for €24.95. That sounds like a very good deal. … For those who have never heard of IPv6, because that’s what is missing.

    Could this be the start of router vendors dumping their IPv4 only stuff?

  25. Karen says on :

    ARIN make some minor adjustments to block 47.

    Bell Labs have always had block 47. It looks like they found just over 500,000 they didn’t neeed and have been assigned in two blocks to Optimum Online. So the whole block gets reset, squewing todays count.

  26. rb12345 says on :

    I noticed today that the RIPE chart (http://www.ripe.net/internet-coordination/ipv4-exhaustion/ipv4-available-pool-graph) shows that 0.75 /8s were allocated in the past 11 weeks. That’s a lot faster than I remember it being in the past year, and if it continues we may well see it running out in November.

    As an aside, it would be interesting to see if this increase is linked to economic improvements compared to January…

  27. plasmoid says on :

    The recent speed up in both ARIN(http://www.potaroo.net/tools/ipv4/fig27i.png) and RIPE(http://www.potaroo.net/tools/ipv4/fig27j.png) could be due the change in reserve policy that was implemented.

    I remember when the global pool was depleted ARIN switched to a 3 month supply guarantee to ISPs from the 12 months they were providing before. So now ISPs have exhausted their old 12-month buffer supply and are requested smaller blocks more frequently.

    I could see ISPs in the RIPE region requesting somewhat buffing their numbers slightly just to make sure then have more IPs when D-Day comes.

  28. Grahm says on :

    :rb12345

    http://www.ripe.net/internet-coordination/ipv4-exhaustion/ipv4-available-pool-graph now shows decline from 53.66 to 41.16 M IPs from Jan 23 and Apr 9 2012.
    That’s 12.5 M IPs difference.

    Looking at ftp://ftp.ripe.net/pub/stats/ripencc/delegated-ripencc-latest I see that in the past year,
    from Jan 23 and Apr 9 2011, 14.1 M IPs were issued.

    So it’s not “a lot faster”, but “a bit slower”, if I’m correct.

  29. Karen says on :

    Yes and no. Things fluctuate.

    Four months of
    Dec 2010 – March 2011 – 18,498,288
    Dec 2011 – March 2012 – 19,228,600

    There isn’t much difference there. However, when you consider that the earlier figures had assignments that were supposed to last up to 12 months, and now they are 3 months.

  30. JMV2009 says on :

    Ripe exhaustion still seems about 20 weeks out, if allocation rates stay this slow. Within three months after that, most regular growing companies will have run out, unless they are able to take further address conservation measures.

    In the APNIC region, most regular growing companies should be running out about now, a year after APNIC exhaustion.

    I have not heard of any companies which have been limited in the APNIC region?

  31. jmv2009 says on :

    It seems like ARIN and IETF are making a /10 of private address space available.

    http://www.lightreading.com/document.asp?doc_id=220006&f_src=lightreading_gnews

    This is good news, as this reduces the pressures on the likes of comcast to apply for or buy public address space. Furthermore, there is less conflict between private address assignments, such as with VPN.

  32. jmv2009 says on :

    40-55% ipv4 trading growth … per month

    http://www.seattlepi.com/business/press-releases/article/Accuro-Consulting-Posts-Record-IPv4-Transaction-3470491.php

  33. jmv2009 says on :

    Shared “transition” space:
    100.64.0.0/10

    http://chrisgrundemann.com/index.php/2012/100640010/

  34. rb12345 says on :

    jmv2009: 20 weeks out is about mid-September, isn’t it? That seems a little bit too fast to me.

  35. Karen says on :

    rb,

    Don’t forget that RIPE have a /16 reserve, so they only have just over 1 /16 to get through.

  36. Karen says on :

    Sorry a /8. Been a long day

  37. rb12345 says on :

    I was assuming that the 20 weeks was “time to last /8″, but was also assuming that the current rate of about 0.7 /8s per 11-week-period was going to end, and it would fall back to nearer 0.5 as it was for much of the past year.

  38. jmv2009 says on :

    Ripe exhaustion about 17 weeks from now. Slipped a week further, but still mid-september.

    http://www.ripe.net/internet-coordination/ipv4-exhaustion/ipv4-available-pool-graph

    The summer might be quiet (due to hollidays), or hectic (due to pending exhaustion).

  39. jmv2009 says on :

    According to akamai, “The state of the Internet, Q4 2011″(free after supplying E-mail)., South Korea is using 14% less public IPv4 addresses with respect to a year earlier. This is probably not due to IPv6 usage, but due to NAT deployments where previously public addresses where used.

Leave a Reply