Re: IPv4 consumption statistics and extrapolations
On 7-nov-04, at 22:18, Tony Hain wrote: That said, I stand by the point that if the recent depletion rate of 9 /8s in 6 months holds, there are only 58 months left. That event may have been an anomaly, or it may be the precursor to an even more accelerated run rate. We won't know for several years which it was. Right. It's almost impossible to come up with good projections because there are many factors that are easily overlooked or too hard to quantify. A simple one is that /8s are reclaimed from time to time, which increases the number of available addresses and lowers the apparent burn rate. A much more complex issue is the practice of some ISPs to use old /8s they got way back when to give out addresses to their customers (4.0.0.0/8 is a good example). Once those private stashes are gone we'll see a higher burn rate. I think the fact that the whole broadband revolution doesn't show up in Geoff's statistics is a good indication that those statistics are incomplete. What I'm afraid of is that we may end up in a situation where many people have all the addresses they need and don't see any reason to adopt IPv6, while others who are late to the table can't get sufficient IPv4 address space and will have to adopt IPv6 out of necessity, resulting in a fragmented internet where large groups can only communicate with each other using hideous NAT hacks. -Original Message- Ietf mailing list Ietf mailing list Ietf mailing list Ietf mailing list Would you guys please stop quoting previous messages verbatim? At this rate the SMTP protocol will break before the end of this debate. ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
Re: IPv4 consumption statistics and extrapolations
What I'm afraid of is that we may end up in a situation where many people have all the addresses they need and don't see any reason to adopt IPv6, while others who are late to the table can't get sufficient IPv4 address space and will have to adopt IPv6 out of necessity, resulting in a fragmented internet where large groups can only communicate with each other using hideous NAT hacks. I can echo every part of this concern except for the use of future tense! And, if people are already on opposite sides of NATs, it's easier for ideas like IPv9 to pop up (the Great Firewall of China discussion earlier this year, etc.)... Spencer ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
RE: IPv4 consumption statistics and extrapolations
--On 6. november 2004 17:15 -0800 Christian Huitema [EMAIL PROTECTED] wrote: By the way, I must apologize. I had a bug in the spreadsheet that I used, and the numbers that I quoted are goofy. The consumption did increase during the last year, but not quite that fast. Let's say that, depending on which number you pick, the consumption per year is between 4 and 10 blocks, which means an exhaustion in 10 to 15 years if the rates continue. But it also mean that we are de facto in the upper size of the S curve, because we have passed the 50% capacity point. your remark reminds me of the red-and-black chart of allocated and routed addresses that someone (KC?) showed at an IETF plenary some years back. Other people called it proof that there is no problem. I called it halfway to Hell. ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
Re: IPv4 consumption statistics and extrapolations
From: Geoff Huston [EMAIL PROTECTED] I would like to correct a few numbers in Tony's comments based on my work in this area that Tony has referred to. ... At this rate the central pool will exhaust in 2018, some 14 years hence. The incredibly rich irony, for those with long memories, is that *IPv6 only exists because of a previous round of FUD about IPv4 address exhaustion* - one spread by the proponents of yet another protocol that was going to replace IPv4 - i.e. CLNP. It's deja vu all over again. Of course, there's a wondeful recursive flavour to this go-round (as noted above) that was missing from the last one. (Of course you should consult your favourite oracle, mystic, soothsayer or whatever for your own preferred version of the future.) At one point, Frank Solensky had a very interesting graph - which was of the forward motion of the IPv4 addresses will run out date. (I.e. the X axis was date prediction was made, and the Y axis was predicted date of exhaustion). It would be very interesting to see what it would look like today, if updated. Noel ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
Re: IPv4 consumption statistics and extrapolations
On Sun, 7 Nov 2004 09:30:12 -0500 (EST), Noel Chiappa wrote: The incredibly rich irony, for those with long memories, is that * IPv6 only exists because of a previous round of FUD about IPv4 address exhaustion* - one spread by the proponents of yet another protocol that was going to replace IPv4 - i.e. CLNP. Noel, this assertion is just plain wrong. The IPv6 effort began because, 10 years ago, we ALREADY had legitimate requests for address blocks being refused. Had those requests been honored, we would have run out of addresses almost instantly. And therein lies the key, strategic error we made. We looked for when we would no longer be able to satisfy ANY requests -- and the estimate I remember was 2020 -- rather than when we had to start refusing legitimate ones. So, the market found a way to route around our non-responsiveness. Now we need to learn to live with it. (We would probably have had to learn to live with NATs in any case, since there is a different line of analysis that says that NATs serve more purposes than countering address space scarcity and that they were essentially inevitable because of the leaf-network conveniences they provide.) d/ -- Dave Crocker Brandenburg InternetWorking +1.408.246.8253 dcrocker a t ... www.brandenburg.com ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
Re: IPv4 consumption statistics and extrapolations
From: Dave Crocker [EMAIL PROTECTED] *IPv6 only exists because of a previous round of FUD about IPv4 address exhaustion* - one spread by the proponents of yet another protocol that was going to replace IPv4 - i.e. CLNP. Noel, this assertion is just plain wrong. So what was Kobe and the ensuing Boston Tea party about, then? Look, I'm not saying there wasn't concern about address space usage rates, and eventual exhaustion - clearly there was. (And - and how ironic is this - one of the *earliest* references to comlete address space exhaustion was in a presentation *I* gave at the 19th IETF, in December 1990, at Boulder, Colorado - up until then we had mostly been worried about the usage rate of class B's.) However, my perception was that the IPng effort was started in response to concerns raised by backers of CLNP, who did so in an attempt to push adoption of CLNP. Would we have started work on IPng without those efforts? I don't think so, but YMMV. the market found a way to route around our non-responsiveness. ... We would probably have had to learn to live with NATs in any case This I actually agree with, but with a slightly different spin: I think that even if we had written an IPng spec overnight, the market would almost certainly still have gone with IPv4+NAT; just less overall hassle, plus the *other* reasons people deploy NAT (which you list). Noel ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
RE: IPv4 consumption statistics and extrapolations
Geoff Huston wrote: I would like to correct a few numbers in Tony's comments based on my work in this area that Tony has referred to. The least squares best fit of advertised address space in the IPv4 domain over the past 5 years is a consumption rate of 4 /8s per year, slightly less than half of Tony's number To continue the refinement; as you frequently point out as well, there is a lag between request/assignment from the RIR and advertisement. Given that, the advertisement growth you are quoting does not account for the recent slope change in the IANA pool depletion I am referencing, and won't even start to do so for another 6-12 months. Even over the past 10 months the least squares best fit of data is a consumption rate of 5.5/8's per year Let's be clear, consumption rate from the pool is not the same as advertisement rate you are basing your measurements on. The size of the advertised pool has absolutely no bearing on the size of the remaining stock held by IANA or the RIR's. The slopes may be on a time delayed track, but there is always the opportunity for addresses to be pulled from the pool yet never advertised. As my I-D on 1918bis points out there are organizations that have outgrown the available private space, so there only current option is to acquire public space they never intend to route. At this rate the central pool will exhaust in 2018, some 14 years hence. No, by your measure this is the date the advertised pool will be receiving all possible prefixes. As we have discussed before, there are two problems with these numbers, the first is that it assumes all currently reserved /8 prefixes can and will be used (by my count there were really only 78 useful ones left in August), and second that it assumes that someone will find and reclaim the ~13% of the space currently 'lost in the system' (the difference between the IANA reserved and RIR-assigned). i.e. some 168 months hence. Allowing for an accelerating consumption rate at an exponential rate brings this forward to 10 years, or 120 months. (details of the analysis are at http://bgp.potaroo.net/ipv4/) (Of course you should consult your favourite oracle, mystic, soothsayer or whatever for your own preferred version of the future.) We can probably all agree that the 'last IPv4 address' will never be acquired. Policies will become stricter until the price is so high that nobody can afford it; or nobody will care once the replacement is deployed. Tony regards, Geoff At 07:38 AM 6/11/2004, Tony Hain wrote: Harald, I would like to congratulate you on your successes, and suggest you have the opportunity to be the last chair to preside over active work related to version 4 of the IP protocol suite. With the publication of the tunneling drafts that v6ops has been sitting on, there is no further need to discuss 32 bit address objects. At the same time, there is really no further justification for any other IETF working group to be discussing 32 bit addresses in current work. With all due respect to Geoff's efforts to document the address growth rate in the routing system, even he acknowledges that measure lags the allocation timeframe and assumes the RIRs will recover all space currently considered lost. Given that IANA allocated 9 /8's over a 6 month period this year, coupled with the fact that only 78 /8's remain in the useful part of the pool (ie: 52 month burn out), it should be clear to everyone that products that rely on current standards activities will appear in the market place after the central pool of 32 bit values has run dry. As such I would recommend your legacy include an active review of all working group discussions next week for items related to IPv4, followed by closure of all 32 bit address related work items before your departure in March. Tony -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Harald Tveit Alvestrand Sent: Friday, November 05, 2004 1:20 AM To: [EMAIL PROTECTED] Subject: Stepping down as IETF chair in March Thomas' note reminded me that there are probably some people who haven't heard this yet I'm stepping down as IETF chair in March, and I am not a candidate for reappointment. It's been a great four years, containing lots of learning experience, lots of hard work and lots of joy - but after four years as IETF chair, and ten years total on the IESG/IAB, March seems an appropriate time for me to leave this stage of my life behind. The IETF is a great organization. I will enjoy watching it continue to grow and prosper under new leadership. Thank you! Harald ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf ___ Ietf mailing list [EMAIL PROTECTED]
Re: IPv4 consumption statistics and extrapolations
On Sun, 7 Nov 2004 12:00:09 -0500 (EST), Noel Chiappa wrote: *IPv6 only exists because of a previous round of FUD about IPv4 address exhaustion* - one spread by the proponents of yet another protocol that was going to replace IPv4 - i.e. CLNP. Noel, this assertion is just plain wrong. So what was Kobe and the ensuing Boston Tea party about, then? You are confusing a mis-handling of making a decision with the instigating cause for the effort that involved that decision. Kobe came roughly a year after work was done on considering the problem of rapidly depleting address space availability. There was plenty of basis for the concern about address space. And CLNP was not proposed until well into that process. Kobe was the culmination of a protracted process of disconnection between the IAB and the rest of the community. (And for what it's worth, the parallels with the current disconnect between the community and the iesg/iab are pretty striking.) That CLNP advocates controlled that later decision process is significant but not in terms of the core technical problem that was being addressed. Look, I'm not saying there wasn't concern about address space usage rates, and eventual exhaustion - clearly there was. you attributed the effrort to a CLNP consipracy. That's a pretty silly assessment, no matter how serious and damaging the CLNP error actually was. However, my perception was that the IPng effort was started in response to concerns raised by backers of CLNP, who did so in an we had a lot of OSI types involved with the IAB at that point, so there is some guilt-by-association that accounts for your erroneous perception, in my opinion. But there were plenty of non-OSI folk also concerned about address depletion. the market found a way to route around our non-responsiveness. ... We would probably have had to learn to live with NATs in any case This I actually agree with, but with a slightly different spin: I think that even if we had written an IPng spec overnight, the market would almost certainly still have gone with IPv4+NAT; just less overall hassle, plus the *other* reasons people deploy NAT (which you list). wow. i think we agree about that, too. I think that the leaf-network admin conveniences permitted by NATs has long been ignored by the IETF community. In general, we have architecturally ignored too many of the boundary issues between independent environments. And I really like the use of the term tussle points to refer to these; it strikes me as a really key perspective on their role. d/ -- Dave Crocker Brandenburg InternetWorking +1.408.246.8253 dcrocker a t ... www.brandenburg.com ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
Re: IPv4 consumption statistics and extrapolations
On Sun, Nov 07, 2004 12:00:09PM -0500, Noel Chiappa allegedly wrote: From: Dave Crocker [EMAIL PROTECTED] *IPv6 only exists because of a previous round of FUD about IPv4 address exhaustion* - one spread by the proponents of yet another protocol that was going to replace IPv4 - i.e. CLNP. Noel, this assertion is just plain wrong. So what was Kobe and the ensuing Boston Tea party about, then? Look, I'm not saying there wasn't concern about address space usage rates, and eventual exhaustion - clearly there was. (And - and how ironic is this - one of the *earliest* references to comlete address space exhaustion was in a presentation *I* gave at the 19th IETF, in December 1990, at Boulder, Colorado - up until then we had mostly been worried about the usage rate of class B's.) However, my perception was that the IPng effort was started in response to concerns raised by backers of CLNP, who did so in an attempt to push adoption of CLNP. Would we have started work on IPng without those efforts? I don't think so, but YMMV. My recollection is that CLNP was not a motivator, it was recommended as a bandaid, in reaction to the perceived problems. After we did that, the CLNP proponents ran with it (and we got Kobe). Remember the ROAD meeting where you said CLNP is only slightly less paleolithic than IP? ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
RE: IPv4 consumption statistics and extrapolations
And for further clarification... I put this response together based on the data I saw from Geoff a couple of months ago, and couldn't check the URL in the air. Everyone should check the site because he has included further evaluation of the data. I apologize for any perception or inference that Geoff may not have been presenting valid data. The data he has been presenting is valid for what it measures, our differences of opinion have been over what and where to measure. That said, I stand by the point that if the recent depletion rate of 9 /8s in 6 months holds, there are only 58 months left. That event may have been an anomaly, or it may be the precursor to an even more accelerated run rate. We won't know for several years which it was. Tony -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Tony Hain Sent: Sunday, November 07, 2004 12:37 PM To: 'Geoff Huston'; 'Harald Tveit Alvestrand' Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]; 'Pekka Savola' Subject: RE: IPv4 consumption statistics and extrapolations Geoff Huston wrote: I would like to correct a few numbers in Tony's comments based on my work in this area that Tony has referred to. The least squares best fit of advertised address space in the IPv4 domain over the past 5 years is a consumption rate of 4 /8s per year, slightly less than half of Tony's number To continue the refinement; as you frequently point out as well, there is a lag between request/assignment from the RIR and advertisement. Given that, the advertisement growth you are quoting does not account for the recent slope change in the IANA pool depletion I am referencing, and won't even start to do so for another 6-12 months. Even over the past 10 months the least squares best fit of data is a consumption rate of 5.5/8's per year Let's be clear, consumption rate from the pool is not the same as advertisement rate you are basing your measurements on. The size of the advertised pool has absolutely no bearing on the size of the remaining stock held by IANA or the RIR's. The slopes may be on a time delayed track, but there is always the opportunity for addresses to be pulled from the pool yet never advertised. As my I-D on 1918bis points out there are organizations that have outgrown the available private space, so there only current option is to acquire public space they never intend to route. At this rate the central pool will exhaust in 2018, some 14 years hence. No, by your measure this is the date the advertised pool will be receiving all possible prefixes. As we have discussed before, there are two problems with these numbers, the first is that it assumes all currently reserved /8 prefixes can and will be used (by my count there were really only 78 useful ones left in August), and second that it assumes that someone will find and reclaim the ~13% of the space currently 'lost in the system' (the difference between the IANA reserved and RIR-assigned). i.e. some 168 months hence. Allowing for an accelerating consumption rate at an exponential rate brings this forward to 10 years, or 120 months. (details of the analysis are at http://bgp.potaroo.net/ipv4/) (Of course you should consult your favourite oracle, mystic, soothsayer or whatever for your own preferred version of the future.) We can probably all agree that the 'last IPv4 address' will never be acquired. Policies will become stricter until the price is so high that nobody can afford it; or nobody will care once the replacement is deployed. Tony regards, Geoff At 07:38 AM 6/11/2004, Tony Hain wrote: Harald, I would like to congratulate you on your successes, and suggest you have the opportunity to be the last chair to preside over active work related to version 4 of the IP protocol suite. With the publication of the tunneling drafts that v6ops has been sitting on, there is no further need to discuss 32 bit address objects. At the same time, there is really no further justification for any other IETF working group to be discussing 32 bit addresses in current work. With all due respect to Geoff's efforts to document the address growth rate in the routing system, even he acknowledges that measure lags the allocation timeframe and assumes the RIRs will recover all space currently considered lost. Given that IANA allocated 9 /8's over a 6 month period this year, coupled with the fact that only 78 /8's remain in the useful part of the pool (ie: 52 month burn out), it should be clear to everyone that products that rely on current standards activities will appear in the market place after the central pool of 32 bit values has run dry. As such I would recommend your legacy include an active review of all working group discussions next week for items related to IPv4, followed by closure of all 32 bit address related work items
Re: IPv4 consumption statistics and extrapolations
* * At this rate the central pool will exhaust in 2018, some 14 years hence. * i.e. some 168 months hence. Allowing for an accelerating consumption rate * at an exponential rate brings this forward to 10 years, or 120 months. * (details of the analysis are at http://bgp.potaroo.net/ipv4/) * * (Of course you should consult your favourite oracle, mystic, soothsayer or * whatever for your own preferred version of the future.) * * regards, * * Geoff * * Some 10 years ago, every IETF plenary meeting had a soothsayer session, projecting how soon we would run out of IPv4 addresses. Has anyone looked to see how today's data extrapolates from the predictions then? Was it as S curve, after all?? Just wondering... Bob Braden ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
RE: IPv4 consumption statistics and extrapolations
To complement Geoff's numbers, I will add this: there is a phenomenon that can't be measured by numbers, and this phenomenon is stockpiling just in case. What I see a lot today is enterprise administrators requesting way more addresses than they could use, just in case. It's just like when there is a hurricane warning: even if the hurricane is forecast to landfall two hundred miles away with a 20 miles error margin, everyone rushes to the supermarket to stockpile mineral water and rice. We have collectively said so many times that IPv4 addresses were going to be rationed that people stockpile them, especially when is costs nothing. The typical scenario being: a 1,000 person organization, 700 desktops and some server is going to request a /20 to their ISP because they can justify it. It costs nothing more than requesting a /24. The reality is that the organization conveniently forgets to mention that 99% of the desktops and 75% of the servers are behind NAT with a RFC1918 address and that they barely use two /25s. It's a matter of money, as always: when IPv4 addresses begin to be difficult to find, people that have excess of them will sell them, rent them, or ask for financial incentives to release them. IPv4 addresses will never run out, they will simply come at a cost. For the next 25 years, as long as I am willing to pay $5/month for an IPv4 address, I have one. Michel. ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
RE: IPv4 consumption statistics and extrapolations
Title: Re: IPv4 consumption statistics and extrapolations Some 10 years ago, every IETF plenary meeting had a soothsayer session, projecting how soon we would run out of IPv4 addresses. Has anyone looked to see how today's data extrapolates from the predictions then? Was it as "S" curve, after all?? There are two kinds of S curves, depending on what creates the asymptote. You may have an S curve that flattens when everybody is served (e.g. everybody on earth has a TV set), and another that flattens when the resource is exhausted (e.g. the lastcod has been fished). Whether the address allocation falls in one or the other category will certainly be debated... As for extrapolating IANA assignment of /8 addresses, it is an interesting game. The data is available for everybody to look at http://www.iana.org/assignments/ipv4-address-space. If you sort allocations by date, you see three phases: - an initial allocation phase that ends in May 1993 when addresses start to be allocated by RIR using the CIDR policy. At the end of May 93, 94 prefixes are allocated or otherwise reserved. - a relatively slow growth from May 93 to April 04, during which 50 new prefixes are allocated - a recent spurt of activity causing 20 allocations between April and November 04. Depending over which period you average, we can argue that the allocation rate is: - 6.8 per year between 1981 and 2004 (163 blocks divided by 24 years) - 4.5 per year between May 1993 and April 2004 (50 blocks divided by 11) - 6 per year between May 1993 and November 2003 (70 divided by 11.5) - 34 per year lately (20 blocks over the course of 7 months) I can assume that different soothsayers will pick different values, depending on whether they want to tell us that the sky is falling, or on the contrary that we should not worry. Another point of debate is how many blocks are actually available. Right now, 163 are in use, out of a total of 256, so we may assume that 93 are available. However, 16 of these blocks fall in the former "class E" category, and may or may not be easy to use... -- Christian Huitema ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
Re: IPv4 consumption statistics and extrapolations
Bob Braden wrote: * * At this rate the central pool will exhaust in 2018, some 14 years hence. * i.e. some 168 months hence. Allowing for an accelerating consumption rate * at an exponential rate brings this forward to 10 years, or 120 months. * (details of the analysis are at http://bgp.potaroo.net/ipv4/) * * (Of course you should consult your favourite oracle, mystic, soothsayer or * whatever for your own preferred version of the future.) * * regards, * * Geoff * * Some 10 years ago, every IETF plenary meeting had a soothsayer session, projecting how soon we would run out of IPv4 addresses. Has anyone looked to see how today's data extrapolates from the predictions then? Was it as S curve, after all?? Bob, if you constrain a resource, it will inevitably follow an S curve. The fact that we have collectively strongly constrained the supply of IPv4 addresses for the last ten years automatically produces the results Geoff observes. Tony Hain makes the real point - if we don't remove that constraint, we will (continue to) constrain innovation and expansion of the Internet. I think that would be immoral. Yes, immoral - we should grow the Internet to be big enough for the whole world population; anything less is selfishness. Brian (maybe a bit tired with jet lag - I don't normally get so steamed up about this) ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
Re: IPv4 consumption statistics and extrapolations
That's the point, IPv6 is no longer about (only) more addresses, but about innovation. And innovation means freedom as well. I even think that it will be possible to recover, slowly, some IPv4 pools when IPv6 is extensively deployed, but who will care then ? Regards, Jordi De: Brian E Carpenter [EMAIL PROTECTED] Organización: IBM Responder a: [EMAIL PROTECTED] Fecha: Sat, 06 Nov 2004 23:10:58 +0100 Para: Bob Braden [EMAIL PROTECTED] CC: [EMAIL PROTECTED] Asunto: Re: IPv4 consumption statistics and extrapolations Bob Braden wrote: * * At this rate the central pool will exhaust in 2018, some 14 years hence. * i.e. some 168 months hence. Allowing for an accelerating consumption rate * at an exponential rate brings this forward to 10 years, or 120 months. * (details of the analysis are at http://bgp.potaroo.net/ipv4/) * * (Of course you should consult your favourite oracle, mystic, soothsayer or * whatever for your own preferred version of the future.) * * regards, * * Geoff * * Some 10 years ago, every IETF plenary meeting had a soothsayer session, projecting how soon we would run out of IPv4 addresses. Has anyone looked to see how today's data extrapolates from the predictions then? Was it as S curve, after all?? Bob, if you constrain a resource, it will inevitably follow an S curve. The fact that we have collectively strongly constrained the supply of IPv4 addresses for the last ten years automatically produces the results Geoff observes. Tony Hain makes the real point - if we don't remove that constraint, we will (continue to) constrain innovation and expansion of the Internet. I think that would be immoral. Yes, immoral - we should grow the Internet to be big enough for the whole world population; anything less is selfishness. Brian (maybe a bit tired with jet lag - I don't normally get so steamed up about this) ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
RE: IPv4 consumption statistics and extrapolations
Title: Re: IPv4 consumption statistics and extrapolations I think Christian made very important points. I'd like to add one point that I'm sure will sound like a broken record to some people. Mobile mobile mobile! There are more mobile devices today than IPv4 can handle, period. Everyone is using NATs to build their mobile networks. The question here is will NATs survive the types of services that operators want to provide? From where I'm looking, the problems are profound and extremely difficult to solve. I think the reason this hasn't been more widely understood is that we're still not close to seeing those mobile peer to peer services over IP. But when that happens (no speculation from me here), I think operators will realise what they're in for. This is especially true for large operators. So, there is more to this than simply looking at the remaining address space and the rates of allocation. Hesham -Original Message-From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Christian HuitemaSent: Saturday, November 06, 2004 5:06 PMTo: Bob Braden; [EMAIL PROTECTED]; [EMAIL PROTECTED]Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]Subject: RE: IPv4 consumption statistics and extrapolations Some 10 years ago, every IETF plenary meeting had a soothsayer session, projecting how soon we would run out of IPv4 addresses. Has anyone looked to see how today's data extrapolates from the predictions then? Was it as "S" curve, after all?? There are two kinds of S curves, depending on what creates the asymptote. You may have an S curve that flattens when everybody is served (e.g. everybody on earth has a TV set), and another that flattens when the resource is exhausted (e.g. the lastcod has been fished). Whether the address allocation falls in one or the other category will certainly be debated... As for extrapolating IANA assignment of /8 addresses, it is an interesting game. The data is available for everybody to look at http://www.iana.org/assignments/ipv4-address-space. If you sort allocations by date, you see three phases: - an initial allocation phase that ends in May 1993 when addresses start to be allocated by RIR using the CIDR policy. At the end of May 93, 94 prefixes are allocated or otherwise reserved. - a relatively slow growth from May 93 to April 04, during which 50 new prefixes are allocated - a recent spurt of activity causing 20 allocations between April and November 04. Depending over which period you average, we can argue that the allocation rate is: - 6.8 per year between 1981 and 2004 (163 blocks divided by 24 years) - 4.5 per year between May 1993 and April 2004 (50 blocks divided by 11) - 6 per year between May 1993 and November 2003 (70 divided by 11.5) - 34 per year lately (20 blocks over the course of 7 months) I can assume that different soothsayers will pick different values, depending on whether they want to tell us that the sky is falling, or on the contrary that we should not worry. Another point of debate is how many blocks are actually available. Right now, 163 are in use, out of a total of 256, so we may assume that 93 are available. However, 16 of these blocks fall in the former "class E" category, and may or may not be easy to use... -- Christian HuitemaThis email may contain confidential and privileged material for the soleuse of the intended recipient. Any review or distribution by others is strictly prohibited. If you are not the intended recipient please contactthe sender and delete all copies. ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf
RE: IPv4 consumption statistics and extrapolations
By the way, I must apologize. I had a bug in the spreadsheet that I used, and the numbers that I quoted are goofy. The consumption did increase during the last year, but not quite that fast. Let's say that, depending on which number you pick, the consumption per year is between 4 and 10 blocks, which means an exhaustion in 10 to 15 years if the rates continue. But it also mean that we are de facto in the upper size of the S curve, because we have passed the 50% capacity point. From: Soliman, Hesham [mailto:[EMAIL PROTECTED] Sent: Saturday, November 06, 2004 3:11 PM To: Christian Huitema; Bob Braden; [EMAIL PROTECTED]; [EMAIL PROTECTED] Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: RE: IPv4 consumption statistics and extrapolations I think Christian made very important points. I'd like to add one point that I'm sure will sound like a broken record to some people. Mobile mobile mobile! There are more mobile devices today than IPv4 can handle, period. Everyone is using NATs to build their mobile networks. The question here is will NATs survive the types of services that operators want to provide? From where I'm looking, the problems are profound and extremely difficult to solve. I think the reason this hasn't been more widely understood is that we're still not close to seeing those mobile peer to peer services over IP. But when that happens (no speculation from me here), I think operators will realise what they're in for. This is especially true for large operators. So, there is more to this than simply looking at the remaining address space and the rates of allocation. Hesham -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of Christian Huitema Sent: Saturday, November 06, 2004 5:06 PM To: Bob Braden; [EMAIL PROTECTED]; [EMAIL PROTECTED] Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: RE: IPv4 consumption statistics and extrapolations Some 10 years ago, every IETF plenary meeting had a soothsayer session, projecting how soon we would run out of IPv4 addresses. Has anyone looked to see how today's data extrapolates from the predictions then? Was it as S curve, after all?? There are two kinds of S curves, depending on what creates the asymptote. You may have an S curve that flattens when everybody is served (e.g. everybody on earth has a TV set), and another that flattens when the resource is exhausted (e.g. the last cod has been fished). Whether the address allocation falls in one or the other category will certainly be debated... As for extrapolating IANA assignment of /8 addresses, it is an interesting game. The data is available for everybody to look at http://www.iana.org/assignments/ipv4-address-space. If you sort allocations by date, you see three phases: - an initial allocation phase that ends in May 1993 when addresses start to be allocated by RIR using the CIDR policy. At the end of May 93, 94 prefixes are allocated or otherwise reserved. - a relatively slow growth from May 93 to April 04, during which 50 new prefixes are allocated - a recent spurt of activity causing 20 allocations between April and November 04. Depending over which period you average, we can argue that the allocation rate is: - 6.8 per year between 1981 and 2004 (163 blocks divided by 24 years) - 4.5 per year between May 1993 and April 2004 (50 blocks divided by 11) - 6 per year between May 1993 and November 2003 (70 divided by 11.5) - 34 per year lately (20 blocks over the course of 7 months) I can assume that different soothsayers will pick different values, depending on whether they want to tell us that the sky is falling, or on the contrary that we should not worry. Another point of debate is how many blocks are actually available. Right now, 163 are in use, out of a total of 256, so we may assume that 93 are available. However, 16 of these blocks fall in the former class E category, and may or may not be easy to use... -- Christian Huitema ___ Ietf mailing list [EMAIL PROTECTED] https://www1.ietf.org/mailman/listinfo/ietf