burnrate > 3 x /8 per month
2011-03-31
A few recent allocations in the APNIC region to Pakistan and South Korea has pushed the global burn rate of IPv4 addresses over 3 x /8 per month. With only about 16 blocks of free IPv4 left in all pools combined, it is not hard realize that IPv6 is needed sooner rather than later.
Last time the burn rate was this high was in May 1995 when Ford, US defense agency DISA and Prudential all got one /8 allocated. At that point, IPv4 addresses was of course handed out in a very wasteful way.
Most of the IPv4 allocations are being made in the APNIC region. I must confess that the tool doesn’t do a good job of modeling the allocation spike that we are seeing in APNIC currently. A more realistic prediction is probably that APNIC will be depleted in May 2011. Some readers have suggested April. I do not believe in April, because it will be hard for members that recently got large address space allocated to justify yet another allocation prior to the depletion (A rush to concert tickets will be most severe the first couple of hours/days, then the rate will slow down even if tickets still are available)
Another interesting thing is that we clearly can see that the ARIN policy that caps the maximum allocation to a 3 month demand has kicked in. In fact, the ARIN burn rate is quite high, but there are no allocations larger than 250,000 IPv4 addresses the last 30 days.
Loading ...
If depletion means zeroing the address pool then once APNIC hits 1.00 it will only handout 1024 addresses per LIR which means it will take a long time to do 16,384 allocations.
Some weeks ago I also thought the rush in allocations might slow, but it has gone on so long that I think now it can easily stay above the one /8 per month that it will take to deplete APNIC in April. Hasn’t APNIC been restricting allocations to 3 months supply for more than 3 months? APNIC has only one /8 left (or maybe 1.2 or 1.3 if you go by Huston’s numbers instead of APNIC’s own exhaustion graphic). To make it to May would require an allocation rate below what has been sustained for a few months now. The lull that I thought might happen a few weeks ago doesn’t seem to have materialized. I wouldn’t be surprised if APNIC’s rate drops some, but I can’t see it dropping below 1.3/month in the last month of availability. I sure wish now that I could see a six month graph of APNIC’s allocations.
Apnic members can still get a 12 months supply.
I’m also not sure how many applications go through two or three 5 work-day application cycles. Because at the current rate, the jig is up in three weeks.
Anyway, a graph would be nice indeed. One should first be guided by historic data. Then we can speculate whether it’s going to increase further or is going to drop again.
But indeed, it is hard to imagine all the big ones have not made their applications yet. But they are already cutting it awfully close. Just two more request from apnic for additional information could get them losing out.
Everyone is interested to plot the course, but really if you just wait about 3 weeks you’ll get all the answers you want
.
Let’s put it this way: what was allocated in the last 13 days is 106% of the currently available stock.
… And last week was considered “weak”
There are a couple of compounding reasons why the concert ticket rush is not fully equivalent:
1) In this case, when applying later, you get more, since it will be your last application.
2) As recent as a month ago, the predictions were showing ~september as exhaustion date. Now, exhaustion is possible in as little as two large allocations.
These two together can cause a real frenzy.
Remaining APNIC “various” blocks now at 0.11 /8s (down from 0.23 yesterday). As I said before, I am not sure whether these are included in the general IPv4 “remaining blocks” figure or not. It now does not really matter anymore as this is just supply for one day anyway.
As jmv2009, I see no real reason why allocation rates should go down significantly in the next weeks. One would only be able to conclude this by checking that every single large ISP in the APNIC region has received a large allocation in the last few months. If there are just two or three large ones that still have the right to request more address space, this would be enough to bring down stocks to the last, reserved /8 block.
I am sticking to my estimate of 10 April plus/minus 5 days (admittedly, this range is quite broad for a date 10 days away from today).
I too see no reason for the APNIC burn rate to slow down. We did not see a last minute slow down when the IANA pool neared depletion either. On the contrary, the intensity of the last minute rush surprised everyone.
Given the large uncertainties inherent to this phase of the process, I prefer not to give a date with a spread, but to just give an upper limit. I predict APNIC will quasi deplete (drop below 1.0 /8) BEFORE April 20th.
Strangely, it seems like there were net returns of 0.03 /8 blocks today. 1.05 /8 blocks remaining.
April’s fool day?
-0.03/0.05-7/5=-2.0
with 0.05 estimated usage rate
So ~2 days later exhaustion
Possibly this will bring the stock of Huston and Lagerholm to the same value of that claimed by Apnic.
Hi JMV,
My stock includes the Japan Inet leftovers in the 43.0.0.0/8 block. This block was previously tagged as Japan Internet by IANA, but the administration was handed over to APNIC in 2007. There are still a couple of million addresses available in this block, for example 43.224.0.0/16. Maybe APNIC started to count some of those addresses?
Right, something like that.
Haine’s predictions changed today. Curiously, not the exhaustion date, but the current stock now makes sense, and starts at 2.3 at the beginning of april, where a couple of days it was ~1.7, which didn’t make sense.
Haine’s exhaustion date: April 18.
Huston’s exhaustion date: April 22.
Lagerholm’s exhaustion date: June 2.
“Lagerholm’s exhaustion date: June 2.”
So clearly Stephan is the dissident.
43.224.0.0/16 is not the reason for the increase. Both yesterday and today, it showed as “reserved” in the delegation file.
(Although it is unclear why it is reserved in the first place.)
Interestingly, the “inventory status” pie chart on the APNIC website now shows “0″ for “reserved”. This still was a small number yesterday. I think it was 0.04 – can anyone remember?
It seems the IPv4 addresses marked as “reserved” (0.28 /8s) in the delegation file are not reflected in the pie chart numbers. The 49.53 /8s marked as “allocated” or “assigned” in the delegation file is equal to the “delegated” figure in the pie chart. 2.05 /8s are marked as “available” in both the delegation file and the pie chart.
So what are the “reserved” addresses in the delegation file and why are they not shown in the pie chart?
Bad time to start an ISP in asia?
I think there is an unlevel playing field now for incumbents and newcomers.
How are you going to be able to acquire a million or so IPv4 addresses any time soon any cheap?
How long before this doesn’t matter anymore?
No slow-down today – 0.27 /8 blocks allocated (26% of remaining stock).
0.78 /8 blocks available for normal assignment.
0.28 /8 blocks marked as reserved (unclear whether still available for allocation)
1.00 /8 block earmarked for small future allocations
One of the biggest issue with the remaining space is going to be continuous blocks. Say someone wants a block of 8 million addresses, they are going to be very disappointed.
Hi All,
I made some changes to the tool. It is now possible to select 1/2 year of historical data. This will also be the default of my calculations. This moves the predicted APNIC date forward with a few days.
Despite the change, the automated prediction is probably not going to that accurate for APNIC. The rush we are seeing is hard to model with statistical tools. Personally, I think May 1st could be the date.
/S
/S
I cannot really see the APNIC pool lasting out until May, simply because the average allocation rate for APNIC is about 1 /8 per fortnight. There are 1.78 /8s remaining today, including the final /8, so based on that, I’d expect (quasi-)depletion to be around the 12th April. You could possibly add a couple of days to that estimate to allow for inactivity over the weekend, but overall that shouldn’t matter much.
Now, if APNIC can last out until the 21st April (Good Friday), we can argue that the Easter holiday and associated bank holidays will hold off depletion for a week or so. For that to happen, we’d need the allocation rate to be unusually low (<0.3 /8s per week versus the usual ~0.5). If allocation is as slow as it was two weeks ago (21st-27th March was very slow at 0.23 /8s per week), depletion day would be the 25th April, ignoring weekends and the Easter break. If that happens, then I'd agree that the 2nd or 3rd of May makes sense. Personally, it does not seem likely that we'd see three weeks at such a low rate, but if providers managed to get exceptionally large allocations this week, maybe it could happen.
I should probably include the case where we maintain this week's rate of 0.83 /8s per week, just for completeness. If that happens again, APNIC will start their final /8 policy on the 8th – next Friday!
In the last week APNIC allocated more space than is now available in multiple independent allocations. There is absolutely no sign of a slowdown. Four weeks requires allocation rates of last year.
I realized that making the graph I wanted of APNIC consumption was as easy as spending two minutes typing in the remaining numbers from the APNIC page into a spreadsheet. So here it is.
http://cid-6ff72de971be8ef0.photos.live.com/embedphoto.aspx/General/ApnicMar17toApr3.png
Unfortunately I only have data back to March 17, which is only almost three weeks back.
Huston seems to think APNIC has .2 /8s more than APNIC thinks it has, and since I think Huston works for APNIC and puts a lot of effort into this, I decided to add .2 to the APNIC numbers to match Huston.
Today Huston is predicting exhaustion of APNIC on 19 April, which looks consistent with this short graph.
I don’t know what happened to the previous copy of this (posted on the 2nd), since it seems to be stuck in moderation. Expired captchas, maybe?
Anyway, the following is all speculation on my part, just based on linear fits of the pool graphs (http://www.apnic.net/community/ipv4-exhaustion/graphical-information). The idea is similar to Mindbuilder’s graph, but based on a shorter date range.
The average allocation rate for APNIC is about 1 /8 per fortnight. There are 1.78 /8s remaining today, including the final /8, so based on that, I’d expect (quasi-)depletion to be around the 12th April. You could possibly add a couple of days to that estimate to allow for inactivity over the weekend, but overall that shouldn’t matter much.
I think we can safely say that APNIC should deplete before Easter. For depletion on the 22nd, we’d need the allocation rate to be unusually low (<0.29 /8s per week versus the usual ~0.5).
In the worst case, we get last week's rate of 0.83 /8s again, and APNIC will start their final /8 policy on the 8th (this Friday).
And so the problems start:
http://www.techedbackstage.net/2011/04/01/teched-and-the-ipocalypse-what-you-need-to-know/
Mike, that actually provides a new possible use for the Nortel addresses Microsoft were rumoured to be buying, assuming they are legacy blocks. If you have RIR-independent legacy addresses, these can probably be used world-wide without any objections from RIRs, assuming you can get providers to route your traffic. Now, for certain types of events, you might actually need >1024 public addresses, or temporary addresses may be unavailable post-depletion. It would be possible to save most of the Nortel address space for cloud use, and still keep a few sizeable blocks as temporary addresses for conferences, which would be unaffected by local RIR depletion. I suppose Microsoft could even rent these blocks out for other conferences to use if they felt like it.
Still no slowdown (but no speed-up either). 0.12 /8s allocated today, meaning 0.63 /8s remaining.
Mike,
Are you excluding the various element assigned to APNIC?
Karen
Karen, the 0.63 /8 figure is simply from the APNIC graphs (subtracting the final /8). In APNIC’s extended statistics file, these are the “available” blocks.
On the depletion date: I was thinking yesterday that this week would be quiet, but from today’s figures, depletion around next Wednesday still looks pretty likely. It’ll be interesting to find out!
If somebody asked for a /8 five work-days ago, will they get a /9? Or will they get nothing? Or will they get everything remaining?
@Karen,
0.63 /8s are still normally allocatable
0.30 /8s are marked as “reserved” (not included in 0.63)
0.08 /8s are “various” blocks (I am not sure whether they are included in the 0.63 above, can someone check?)
This potentially makes 1.01 /8s still to be allocated before the final /8; although some of the reserved blocks will definitely stay reserved.
At Apnic, the pie chart now says 1.56 available. I’m not sure what it said yesterday, but the bar chart is not updated yet from yesterday.
Ah, there we go: 0.54 still to allocate. Exhaustion around next weekend…
Mike,
The extended list doesn’t show the availability of the various. You have to work that out. Since the allocations could have been made by different RIR’s, you have to look at all 5.
For instance, if you remove the 196,608 reserved for block 144, there is (if my maths add up) 114,688 left. I just hope I have the correct blocks that are now counted at APNIC.
139 : 196,608
140 : 0
144 : 114,688
150 : 65,536
153 : 0
157 : 65,536
163 : 32,768
171 : 72,704
According to Huston, 4.42% of address space is used by allocating a /9, or 0.5 /8. There is a chance this will happen to the currently available stock. Exhaustion could therefore happen any day now with a significant probability.
To be clear: On a second by second basis, depending on the allocation policy, the chance of getting depleted by trying to allocate a /9 should be around 4.42%. The stock is then between 0 and 0.5 /8.
The chance of the stock being between 0.25 and 0.5 is then 4.41/2=2.21%.
APNIC could drop to below 0.5 any day now. When this happens, at a usage rate of 0.05 per day, there is a 4.42%/(0.5/0.05)=0.442% daily chance of exhaustion.
Karen,
yes, I know that the various blocks are not shown in the APNIC file. I actually use Huston’s summary here:
http://bgp.potaroo.net/stats/nro/various.html
I parse this automatically every 24 hours.
The numbers there (in the column “free /32s”) are somewhat different to your numbers (totalling around 1.4 million addresses). According to the report from Huston, for APNIC, just 0.083 /8s are remaining from the various pool. So irrespective of how this will get used, it is pretty insignificant.
JMV2009: With today’s allocations (106.104/14, 1.200/16, 49.51/16, 49.127/16, and 101.97/16), we’re already under 0.5 /8s. Also, APNIC told me that if they approve a request for space larger than what’s available as a single contigous prefix, they would offer to satisfy the request using multiple smaller prefixes (this is already happening, for example recently CHINANET received a bunch /16s from ERX space on the same day), limited upwards to the total free space of course.
Mike Dee: The various/erx space *is* included in the APNIC delegated-extended file, but it’s simply nothing free for allocation remaining there anymore. Huston’s numbers doesn’t take into account reserved (recently returned and quarantined) blocks, which is why he still lists remaining free various space.
So if somebody gets approved a request for a /9 tomorrow, they will be asked if they want to receive all the remaining space as a few large prefixes (1 /10, 1 /12, 1 /13, 1 /12, and 1 /16, as far as I can tell) and over 1000+ smaller individual ones (all adding up to almost a full /9). My guess is that the answer would be “yes”, because even though they feel all those small prefixes are to much hassle with to be actually useful, the APNIC post-exhaustion transfer policy allows them to sell them on to the highest bidder.
Tore
Apologies, we’re not quite under 0.5 /8s yet – math error on my part. Also there’s a typo in the list of large remainig prefixes – it should have read: 1 /10, 1 /12, 1 /13, 1 /**14**, and 1 /16
Tore
@Tore:
“the APNIC post-exhaustion transfer policy allows them to sell them on to the highest bidder.”
I wonder. Yes, transfer of unneeded space is allowed. As I understand it however, a request for space is only granted when the requester demonstrates need. Saying one needs space and then later sell it, saying “we don’t need it”, may raise nasty questions…
I’m not sure APNIC would want to start taking back blocks just because someone was trying to sell them.. technically they’d be in the right to do so but the politics of it would be messy.
The other question is how long will the final /8 last.. 16,384 individual allocations doesn’t sound a lot in an area the size of asia – especially when the LIRs realize they won’t get another chance.
Indeed,that final /8 can not last long. It won’t be long now before the first “sorry we are out” will be heard by an ens user. Matter of weeks, I’d say…
I think the final /8 will last quite long. 16300 allocation is quite a lot given that there are only around 3000 APNIC members who will each only get 1 allocation. This means that APNIC membership needs to grow to 16000. I think the last /8 will last at least for a decade and probably for more (unless they change their policies).
Most interesting will be to check the burn rate of the other RIRs once APNIC runs out. If it goes up significantly, RIR shopping is probably taking place and IPv4 will run out very quickly in the other regions as well.
So, standing this policy for the last /8, an “official” exhaustion could come not before years from now, or never. In fact the triggering of the last /8 policy will be the only exhaustion date we’ll have.
So we’ll have (0.43-0.25)/0.07=2.5 more work-days where the chance of depletion is 4.42%/(0.5/0.07)=0.6% while the stock is above ~0.25 by allocation of a /9.
After that, the chance increases by 6.63%/(0.25/0.07)=1.9% to 2.5% per day for (0.25-0.125)/0.07=1.8 days while the stock is above ~0.125 by allocation of a /9 or /10.
After that, allocation rises to near certainty within days or hours.
Needless to say that the pool is very soon de facto exhausted: Within about a work-day, submission of an application or clarification of an application is not expected to be filled from the free pool with 50% probability, due to a five work-day processing time.
Here’s a graph of the last three weeks of APNIC remaining. The trend line shows depletion a week from this Monday on Apr 18. But the trend line should probably be steeper because there were some used on Apr 1 that I can’t account for. So it looks more like Friday. Of course from the jittery nature of the graph it is obvious that it could go several days either way, or even deplete Monday. I added .2 /8s to APNIC’s numbers as Huston does.
http://cid-6ff72de971be8ef0.photos.live.com/self.aspx/General/ApnicRemainingApr11.png
@Mike RIR shopping is unlikely, or at least will be limited, because the big allocations in Asia go to National ISPs like China Telecom, NTT, Indonesia PTT; I wonder how these could justify getting IPs from RIPE for example. Unless they buy a European incumbent.
6.405.888 addresses allocated today, 525.056 remaining for allocation. From the remaining ones there’s one /18 and one /19, the rest are /21 and smaller. And the APNIC office hasn’t even closed fpr the day yet.
Significant allocations today is 39.192.0.0/10 to APNIC’s Debogon project, and slightly more than 2M addresses to CHINANET Guangdong province spread out over 241 individual allocations ranging between /13 and /21 in size.
The End Is Near, folks.
Tore
Oh, and another thing you might find interesting: 303.872 of the remaining 525.056 addresses are in blocks sized /24 or /23, which are smaller than APNIC’s minimum allocation size (cf. http://www.apnic.net/publications/research-and-insights/ip-address-trends/min-prefix).
So I wonder what happens next. Do they…
1) Dial down the minimum allocation size to /24?
2) Implement the final /8 policy when they’re out of allocateable prefixes? (That would mean declaring declaring the remaining /23 and /24s unusable, and that there’s only 221.184 addresses left to allocate.)
3) Not implement the final /8 policy until all the /23 and /24s have been assigned to end users, refusing any allocation requests that come in during that time?
Tore
Interesting observation indeed. While option 2 seems the most logical to me, I note that the link you provided return an error 404. So maybe thay have just dropped the minimumm requirement.
@Michiel: Tore’s link returns a 404 because it wrongly includes the closing parenthesis. http://www.apnic.net/publications/research-and-insights/ip-address-trends/min-prefix still works.
Michiel, try http://www.apnic.net/publications/research-and-insights/ip-address-trends/min-prefix . (Apparantly the ending parenthesis accidentally became part of the URL.)
Tore
Aha. And what do I read there?
“APNIC’s current minimum IPv4 allocation size is a /22. As APNIC’s IPv4 transfer policy permits transfers of /24 blocks and larger, it is now possible that routable prefixes within a /8 are now smaller than the minimum allocation size of a /22. Therefore, the minimum prefix size for all APNIC IPv4 /8 address blocks is now a /24.”
So it looks like it will be option 1.
Wouldn’t the allocations to the Debogon project only be temporary?
Michiel, perhaps you are right – but the way I read it, is that the minimum allocation size that APNIC can make to its members is still /22, but that smaller allocations might exist in the records due to transfers between members. FWIW not a single /23 or /24 allocation has been made this year, but there’s been over a hundred /22s allocated.
Mike Dee, they’re probably temporary, but they do count as unavailable space according to APNIC’s own records and exhaustion status web site. It might of course happen that they’ll all be released prior to the implementing the last /8 policy, but to me that does not agree with making a new assignment *today* when that particular assignment takes up *almost all* the remaining space. I mean, if they have to release it for normal allocations tomorrow anyway, then they can’t expect to be able to do any useful testing with it, in turn making the whole assignment utterly pointless. But the assignment *has* been made…
Tore
Tore
Tore:
Yes, when I read it again, I think your interpretation is correct. How about option 4?
4) When a /22 is no longer available from the unallocated space, a party requesting space, will be issued a combination of /23′s and /24′s that adds up to the same space.
That makes more sense to me then letting them goto waste. We will probably have the answer in a couple of days.
And at the end of the day, there are only 458240 IPv4 addresses left at APNIC, corresponding to less than 0.03 /8s.
Pls note the differences between RIPE and APNIC’s “final /8″ policies.
RIPE’s says “allocations from the final /8 it receives from the IANA”
APNIC’s says “allocation from the remaining /8 worth of space” when “pool reaches a threshold of a total of one /8″
It is not the same.
APNIC’s is about the total number. RIPE’s about certain /8 block.
So… Xday tomorrow?
Michiel,
I hope it will be at 15 of April as its my birthday
Can’t they just save up all the tiny blocks for use towards the end of their final eight policy in a few years? Or is the final eight made up entirely of tiny blocks? Likewise can’t they make temporary allocations out of the final eight, since they can presumably take them back when needed? And where are the .2 /8s that Huston seems to think are available above what the APNIC page reports?
I almost thought it was stupid to graph the remaining when there is basically none left, but then I went ahead anyway. I was surprised when after adding the .2 /8s that Huston thinks are available, the linear regression line of the last four weeks still doesn’t predict exhaustion until next Monday April 18.
http://cid-6ff72de971be8ef0.photos.live.com/embedphoto.aspx/General/ApnicRemainingApr12.png
@mindbuilder:
Yeah, linear regression on the stock is not smart. You should average the allocations, and use the current stock to see when depletion will happpen.
With your method, the depletion date can get earlier over the next days, even when there are no allocations, while logically the depletion date should get later.
Michiel, your #4 = my #1
That APNIC will hand out many small blocks if they don’t have a big one that can be used is already known (e.g. the 241 ones to CHINANET), but if either of those small blocks are smaller than a /22, then they’ve effectively decreased their minimum allocation size.
Mindbuilder, the minimum (and maximum) allocation size under the final /8 policy is still /22, so the small blocks are not usable under that policy, no. For your other question about the 0.2 /8s that Huston adds – those are, as far as I can tell reserved blocks (recently returned and quarantined), blocks transferred to other RIRs, and returned space in the APNIC-designated ERX blocks that was previously allocated in another regions (and therefore returned to that region). APNIC didn’t use to publish data that said explicitly what was available before, so he had to infer it from various other sources. But now they do (delegated-apnic-extended-latest), and I see no reson to doubt the accuracy of that file (and their own bar graphs).
BTW, only delegation so far to day is 203.7.160.0/21, as far as I can tell.
Tore
Tore:
“Michiel, your #4 = my #1″
Not exactly. With #1, one could show need and ask for a /24 and get it. With #4, when one asks for a /24, it will be denied. One has to ask and show need for a /22 to get anything at all.
Mindbuilder & JMV2009:
Now that we are so near the end, the graph has become useless for prediction no matter what extrapolation model is used. One moderately large allocation will end it and that is basicly unpredictable. All that is left is statistics. We could say there is an X% chance it will end today, an Y% chance that it will end tomorrow, etc. Where X is probably in the order of 50 and Y 25.
Michiel, point taken.
Anyway – the chance of X is 0%… APNIC offices are closed, and the only allocation visible so far is 203.7.160.0/21.
Tore
Below the command… and the result. I hope the command doesn’t break the comment system
$ grep ipv4 delegated-apnic-extended-latest | grep available | perl -ne ‘/\|([\d\.]+)\|(\d+)\|/; printf “%s/%s\n”, $1, 32-(log($2)/log(2)) if $2 > 2**(32-21);’ | grep -v “^103″
106.0.32.0/19
116.90.0.0/18
meaning that there are only these two blocks “available”, bigger than /21, if we remove the last /8, that is 103/8. If you want to see smaller blocks, just change the “21″ in the command.
Oh and btw you have to download http://ftp.apnic.net/apnic/stats/apnic/delegated-apnic-extended-latest
If I messed up with the command please tell me.
Tore:
“Anyway – the chance of X is 0%… APNIC offices are closed”
Argghh those time differences! APNIC updates the IPv4 usage graph when I start prepairing for dinner. I am on UTC-2. In my mind there was still a whole day to go when I wrote “50%”.
I am really bad at predicting, especially the future.
Guillame, that matches my script (which excludes everything in 103/8):
$ ./sizes.pl < delegated-apnic-extended-latest
/18 (16384) 1
/19 (8192) 1
/21 (2048) 36
/22 (1024) 120
/23 (512) 240
/24 (256) 707
total available: 525056 (0.031 /8s)
Michiel, the delegated file is updated at around 16:15 UTC and the graph a bit later. However, the APNIC office close 7:30 UTC and then it's just a matter of checking their whois database to figure out how much has been delegated that day, if you're as impatient as I am anyway.
Tore
Huh, no allocations today?
Are they up to something?
“the delegated file is updated at around 16:15 UTC”
Actually, it is 15:15 UTC.
91.6% of space is allocated in batches of larger than a /18 (In early 2010).
Maybe they are rejecting these already?
This would cut the allocation rate by 91.6%.
They should at least be rejecting batches larger than a /13, which reduces the allocation rate by 46.95%.
Okay, so a couple of hours before APNIC closes for the day, the status is:
Allocated/assigned today: 498432
Remaining outside of 103/8: 24576
Remaining total: 16734976 – 0.997 /8s
(This is assuming there have been no resource returns today.)
So, if APNIC interprets its own policy text literally, i.e., the «a threshold of a total of one /8» part, they have already passed the threshold and should be implementing the final /8 policy as I write this.
If however they simply disregard the 103/8 block, which is specifically set aside for the implementation of the final /8 policy, there’s still a /18 and a /19 left to delegate from the normal pool before they reach the threshold. Perhaps they’ll hold off doing so until tomorrow so that they can time the final two allocations with a press release or something.
In any case I think it’s safe to say that it’s now way too late to submit any new requests for IPv4 space under the normal policies.
Tore
How is this possible?
We were at 525056 available.
With 498432 allocated, that should leave 26624?
Or was some of 103/8 already allocated before today?
JMV2009, precisely:
$ whois -h whois.apnic.net — -T inetnum -r -m 103/8 | egrep ‘(inetnum|netname)’
inetnum: 103.1.4.0 – 103.1.4.255
netname: Debogon-prefix
inetnum: 103.1.0.0 – 103.1.3.255
netname: Debogon-prefix
inetnum: 103.0.0.0 – 103.0.255.255
netname: Debogon-prefix
Tore
By the way, all of the 1102(!) individual allocations made so far today are made to CHINANET Fujian Province Network.
Tore
There we go, 106.0.32.0/19 and 116.90.0.0/18 was just assigned to the debogon project. So there’s now 0 addresses left outside of 103/8, unless if something has returned to the free pool today.
Tore
According to delegated-apnic-extended-latest, there are now only 0.996 /8s available.
Tore/Mike, don’t forget that apnic extended does not include availability information for what was various.
If I take of the reserved block information, I still show:
Block 139: 196,608 (after taking off 0 reserved)
Block 144: 114,688 (after taking off 196,608 reserved)
Block 150: 65,536 (after taking off 65,536 reserved)
Block 157: 65,536 (after taking off 196,608 reserved)
Block 163: 32,768 (after taking off 327,608 reserved)
Block 171: 72,704 (after taking off 0 reserved)
There are a lot of addresses that are also used for debogon which probably will come back to the free pool. Some of the reserved items will also be freed when advertising issues from returned blocks are resolved.
Karen
Karen, why do you say that? To me it appears like it does indeed include the ERX space, for example:
apnic|ID|ipv4|139.0.0.0|65536|20110330|allocated|A92CC4B4
Also worth noting is that it used to not include it, and then there was a prominent notice on the http://www.apnic.net/community/ipv4-exhaustion/graphical-information page that the bars did not include ERX space. However, a while ago the notice was removed, and allocations from the blocks started.
Tore
Tore,
For various ut includes allocations and reservations, but not available
Take block 139 the allocations by region are:
afrinic 65,536
apnic 4,194,304
arin 6,750,208
lacnic 65,536
ripe 5,505,024
Total 16,580,608
Left 196,608
APNIC has officially announced the Final /8 policy starting now.
http://mailman.apnic.net/mailing-lists/apnic-announce/archive/2011/04/msg00002.html
Karen, I don’t think you can assume that all free space in block 139/8 belongs to APNIC – only the space that were free when the unused space were divided between the RIRs back in 2008. Same goes for all the other blocks.
For example, 139.5.0.0/16 and 139.28.0.0/16 are unallocated, but they belong to the RIPE region – they are present in delegated-ripencc-extended, but not in delegated-apnic-extended. Presumably they have been returned to the RIPE NCC sometime after the 2008 split.
So I still believe delegated-apnic-extended file includes all ERX space under APNIC control. If you check the 20110329 version, you’ll see that the 139.0.0.0/16 example I mentioned earlier is indeed included and marked available.
Tore
Well, we have to hand it to Tony Haine, he has predicted this date for about two months or so (plus minus two or three days)…
What still puzzles me are the blocks allocated to the APNIC debogon project. Assuming no major problems found that can’t be solved, they should be returned relatively soon, bringing the pool back up to more than one /8…