On Tue, 2006-11-28 at 17:37 +1300, Nathan Ward wrote:

jamie baddeley wrote:

...

APNIC might disagree.

http://www.apnic.net/news/hot-topics/index.html#ip-addressing

"The Sky is Falling" emails like this don't help. However, much of that content is 3 years old.

It'd be good if Geoff were able to revise his findings with updated numbers.

http://www.potaroo.net/tools/ipv4/

-- Nathan Ward

_______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog

It'd be good if Geoff were able to revise his findings with updated numbers.

Geoff no longer subscribes to NZNOG, so he can't post directly. He asked me to forward his response:

Date: Tue, 28 Nov 2006 16:05:36 +1100 To: nznog(a)list.waikato.ac.nz From: Geoff Huston Subject: Re: [nznog] IPv4 Exhaustion

I perform a daily analysis of the consumption rates of IPv4 addresses - they are updated on a daily basis at http://ipv4.potaroo.net

If tomorrow continues much like today, and this trend continues uniformly then a reasonable estimate of the date at which the first RIR will be unable to meet a request for IPv4 addresses is 4th August 2012, or slight under 6 years from now.

Frankly, its a VERY rubbery date, and while the statistical calculation is probably reasonable, the underlying assumptions are very flawed. The major assumption behind this projection is theat industry players will not panic, and there will be no last minute run on the remaining resource pool at any point prior to exhaustion. Given the relative paucity of Ipv6 deployment today and the continued reliance on IPv4 in many mainstream product deployments it would seem that this assumption is highly unlikely. A rush is more likely, but the exact timing of any such rush and the volume of such a rush is something I've been unable to model with any level of satisfaction. Tony Hain could be right - it could be in a 2 - 3 year time frame. Or maybe not. It is just not possible to tell from conventional statistical techniques, and at this stage you can get out any of your favourite forms of divination and have a play for yourself - within some bounds any guess is perhaps equally likely as any other!

Now if this was an agile industry with large pools of available capital to make infrastructure investments in advance of need then I suspect that this would not be a very interesting problem - we'd all head into a dual stack world and then phase out IPv4 and there would be no drama. But thats not this industry. Ipv6 today represents a considerable investment, an investment that does not raise any additional revenue and does not produce any additional efficiencies in supply. In short, in the short term business ledgers any such investment is all loss. And in a cutthroat world of mass deployment of low margin broadband access services life is tough enough without adding one more cost element to the business.

So instead we live in a world that appears to be largely in a state of denial, where we'd prefer to spend more time wondering precisely when the falling brick will hit the pavement rather than figuring out what to do afterwards, and then understanding what needs to happen now.

regards,

Geoff

-- Robert Loomans Email: robertl(a)apnic.net Programmer/Analyst, APNIC Phone: +61 7 3858 3100 http://www.apnic.net Fax: +61 7 3858 3199

I attended a very interesting presentation today by Tony Bain, one of the key IETF IPv6 chairs. He says that the rate of /8's being allocated is increasing exponentially at that moment (22 were allocated in 18 months [till July 2005], leaving just 64 available worldwide), and the prior estimate of when IPv4 address was going to be exhausted of 2020 has been revised to 2009 - as in three years away, and that IPv4 address space is likely to take on a much more commercially tradable value as it becomes very rare (like oil).

For existing players in the market this doesn't mean much. However for expanding networks this is going to put a serious crimp in their

The APNIC article you reference only uses data up to January 2003. The data doesn't even show the exponential growth in /8 allocations that have occurred in the last two years. The article makes the erroneous assumption that allocations will continue to be linear. This has no been proven untrue. At the time the article was written there where 221 /8's available. In the two years since then the number has dropped to 64. See the problem with the sudden rapid growth? -----Original Message----- From: jamie baddeley [mailto:jamie.baddeley(a)vpc.co.nz] Sent: Tuesday, 28 November 2006 5:34 p.m. To: Philip D'Ath Cc: nznog(a)list.waikato.ac.nz Subject: Re: [nznog] IPv4 Exhaustion APNIC might disagree. http://www.apnic.net/news/hot-topics/index.html#ip-addressing "The Sky is Falling" emails like this don't help. jamie On Tue, 2006-11-28 at 17:23 +1300, Philip D'Ath wrote: plans.

I polled a lot of players in the NZ market about 6 months ago and

plans

for IPv6 deployment were almost non-existent.

There is an older article he wrote here:

http://www.cisco.com/web/about/ac123/ac147/archived_issues/ipj_8-3/ipv4.

html

We really need large commercial transit players like Global Gateway, TelstraClear, AT&T, UUNet, AsiaNetCom, and Sprint to start offering dual stack IPv6 connections to downstream ISPs, and for ISPs to start allowing clients to connection via IPv6, including via DSL.

We really need to get critical infrastructure, like DNS, fully IPv6 enabled. The .nz domain isn't even reachable via IPv6 addresses yet. Now I know the shared registry system has IPv6 support now, but it is limiting in that you must still supply IPv4 addresses. You can't supply only IPv6 glue records. What happens when you can't get IPv4 addresses to use for glue?

Also the majority of the registries out their offer very poor IPv6 support - as in no IPv6 connected DNS servers, and an inability to easily add IPv6 AAAA records.

_______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog

Where did those numbers come from, because they disagree with the numbers quoted here: http://www.cisco.com/web/about/ac123/ac147/archived_issues/ipj_8-3/ipv4. html The figures you have given look like the monthly allocation rates, as opposed to the annual rates. This is the chart of /8 allocates per month: http://www.cisco.com/web/about/ac123/ac147/images/ipj/ipj_8-3/83_ipv4_fi gure_02_lg.jpg -----Original Message----- From: leo vegoda [mailto:leo(a)ripe.net] Sent: Tuesday, 28 November 2006 9:18 p.m. To: Philip D'Ath Cc: jamie.baddeley(a)vpc.co.nz; nznog(a)list.waikato.ac.nz Subject: Re: [nznog] IPv4 Exhaustion On 28 Nov 2006, at 8:55:am, Philip D'Ath wrote:

The APNIC article you reference only uses data up to January 2003. The data doesn't even show the exponential growth in /8 allocations that have occurred in the last two years. The article makes the erroneous assumption that allocations will continue to be linear. This has no been proven untrue.

At the time the article was written there where 221 /8's available.

The 221 /8s are the whole unicast space and that number hasn't changed. The significant number is the number of those /8s still available. If you want to look at the allocation rate in recent years it is: 2002 - 4 2003 - 5 2004 - 9 2005 - 13 2006 (part) - 10 And that list doesn't take account of the 16 /8s added to the IANA Reserved pool in April 2003. Regards, -- leo vegoda Registration Services Manager RIPE NCC

Hi Philip, On 28 Nov 2006, at 9:37GMT+01:00, Philip D'Ath wrote:

Where did those numbers come from, because they disagree with the numbers quoted here: http://www.cisco.com/web/about/ac123/ac147/archived_issues/ipj_8-3/ ipv4. html

I took the data from the IANA IPv4 registry: http://www.iana.org/assignments/ipv4-address-space

The figures you have given look like the monthly allocation rates, as opposed to the annual rates.

Fortunately, they are the annual rates.

This is the chart of /8 allocates per month: http://www.cisco.com/web/about/ac123/ac147/images/ipj/ ipj_8-3/83_ipv4_fi gure_02_lg.jpg

It's worth noting that the author of the article writes that that chart makes it "difficult to discern the trend, or develop an expectation about the overall lifetime of the remaining pool." Regards, -- leo vegoda Registration Services Manager RIPE NCC

On Mon, Dec 11, 2006 at 11:39:55PM +0000, Andy Davidson wrote:

...

If you want to look at the allocation rate in recent years it is:

Slightly OT, but topical :- A map of IPv4 address space http://xkcd.com/c195.html -jim

I disagree. The driver wont be by customers at all - they don't care abut transit, only the applications that use the transit. Lets say in two years time the cost to get new IPv4 address space rises to US$10k per /24 per month. Is the service provider going to want to pay this, or migrate to free IPv6 space? The cost of not migrating is going to be huge. Service providers need to make sure they are deploying kit now that is IPv6 capable, and making implementation decisions that will assist with the IPv6 migration. Potentially the issue of multihoming and mobile IPv6 is going to be very taxing on current equipment. -----Original Message----- Here is an interesting presentation by Robert Rockell of Sprint about IPv6: http://www.nanog.org/mtg-0405/pdf/rockell.pdf . Slides 5, 7, 9, and 10 are very valid as to why this is unlikely for a few more years to come. The major driver will be customer demand (read: $$$$), or the network operators themselves finding resource shortages are impacting on their operations, thus moving to a v4-over-v6 core. That said, Sprint do offer semi-native v6 transit, and I believe UUNET also.

We'll have to agree to disagree on this one. I still don't believe the demand from customers will drive the adoption. Lets say you're the technical manager at an ISP, its 2010, and the board of directors instructs you to plan for growth of 500 broad band customers per month. You go to APNIC, and they decline to give you ANY IPv4 address space what so ever. You report back to the board that you are no longer able to accept any more customers. I think the board of directors will flip, and tell you to find a solution, as not growing the business will not be considered an option. Now you'll find that you'll be suddenly rolling out IPv6 real fast as the cost to not do it will be extreme. I disagree about NAT-PT as well. NAT-PT has been deprecated as a standard. Cisco routers still support it (and Cisco has no plans to stop supporting it - other platforms will probably support it as well, I have no experience with them). However it is not a solution. It works if you have an IPv4 only server and want to give IPv6 people access to it. I have found it to be very poor going the other way - giving IPv6 only people access to the IPv4 Internet as a whole (I tried this deployment in our own office and had to go back to using dual stacks - but this could be a reflection of my abilities ...). We need the wholesale provides to start selling (I have no problems with it being sold) IPv6 transit, as using tunnels to get everywhere just doesn't scale well. I use tunnelling now, and while usable I would not call it snappy. As the IPv4 address space runs out I can't see any other option but for people like APNIC to drastically increase the cost of supply IPv4 address space, otherwise their own members will go at it like a bull at a gate, and it'll suddenly go in a mad rush towards the end. -----Original Message----- From: Alastair Johnson [mailto:aj(a)sneep.net] Sent: Wednesday, 29 November 2006 12:00 a.m. To: Philip D'Ath Cc: nznog(a)list.waikato.ac.nz Subject: Re: [nznog] IPv4 Exhaustion Philip D'Ath wrote:

I disagree. The driver wont be by customers at all - they don't care abut transit, only the applications that use the transit.

Of course it will be driven by customers. The carriers you cited (ANC, UUNET, Sprint, AT&T, Global Gateway, TCL) are primarily wholesale service providers. Their customers are ISPs, who are driven in turn by their subscribers. If the subscriber is saying "omg, my BitTorrentv6 doesn't work", the ISP probably isn't going to care. If the subscriber is saying, "I can no longer get to (ASB|ANZ|AmEx|whatever)", the ISP will care. As Sprint stated: the customers aren't prepared to pay for v6, and the carriers aren't prepared to invest in it, the enterprises don't run it, and the vendors don't support it thrillingly well yet.

Lets say in two years time the cost to get new IPv4 address space rises to US$10k per /24 per month. Is the service provider going to want to pay this, or migrate to free IPv6 space?

Then the cost of service rises to cover that, because at that time the majority of services are still going to be v4 based. The ISP may encourage v6 adoption, but if subscribers pay... hey, why not. and v6 isn't free. Presumably, given that APNIC/ARIN/RIPE etc are governed ultimately by their own 'customers', if the cost rises to $10k per month then it's the service provider's own fault!

The cost of not migrating is going to be huge. Service providers need to make sure they are deploying kit now that is IPv6 capable, and making implementation decisions that will assist with the IPv6 migration.

The cost of migrating is also huge. I do agree with your latter points, though. However, I'm reasonably convinced at this stage that the exhaustion, or near exhaustion of IPv4 space will lead to: 1) As predicted, a short term rush on space; 2) A response of much more restrictive policies; 3) A longer term response of investigating existing allocations and assignments to see what are in use. Certainly I can cite examples of ISPs in NZ and abroad which are assigning subnets much larger than they should be, or could be engineering to support smaller more efficient utilisation; and can also cite examples of large multinational corporates burning multiple public /16s internally instead of using RFC1918 space or more efficient addressing. 4) A significant amount of R&D in NAPT technologies, and technologies in NAPT traversal. COnsidering most applications these days happily work without a public IP address, is there really a huge demand for everything to be uniquely addressed? Certainly my home networks don't need it.

Potentially the issue of multihoming and mobile IPv6 is going to be very taxing on current equipment.

Multihoming is a disaster under v6. -- i still say ipv4 and atm for ever. yeah.

I think the board of directors will flip, and tell you to find a solution, as not growing the business will not be considered an

The whole world doesn't have to instantly change to IPv6 - they it does need to start running dual stacks now (IPv4 and IPv6), so that in 3 to 6 years (and that is the only time frame debated in this thread) it wont be so harsh when IPv4 stops becoming available for new customers. -----Original Message----- From: Robert Gray [mailto:bob(a)brockhurst.co.nz] Sent: Wednesday, 29 November 2006 8:12 a.m. To: nznog(a)list.waikato.ac.nz Subject: Re: [nznog] IPv4 Exhaustion Philip D'Ath wrote: option.

Now you'll find that you'll be suddenly rolling out IPv6 real fast as the cost to not do it will be extreme.

I'd check that the board were happy that your customers won't be able to host web sites or email servers that can be seen in v4 space or even run p2p apps with v4 customers. When I looked at v6 some 10 years ago (I bought a book) it seemed to me that the world all had to change together and that's not a cutover I can see working too well. Is v6 really the best/only solution to v4 exhaustion? -- Robert Gray _______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog