RFC 4692
Considerations on the IPv6 Host Density Metric
RFC 4692: Considerations on the IPv6 Host Density Metric ✨
Okay bestie, let's start with the header stuff 💅 This is RFC 4692, called Considerations on the IPv6 Host Density Metric 🦋 It was written by G. Huston from APNIC and came out in October 2006. It's labelled Informational, which means it gives the Internet community info and does not specify an Internet standard of any kind. Its header doesn't list anything it obsoletes or updates, so it's a total solo queen 👑 It's part of the Network Working Group series 🌸
Status of This Memo
This memo is info for the Internet community, not a standard 🫶 It doesn't specify an Internet standard of any kind, and it can be distributed without limits. 💜
Copyright Notice
Copyright (C) The Internet Society (2006). 🌈
Abstract
Okay so this memo digs into the Host Density metric, which is what the RIRs currently use to guide registry allocations of IPv6 unicast address blocks 🧚 It puts the address efficiency used for IPv4 allocations side by side with what IPv6 does, and the vibe is that for large allocations the target efficiency numbers can be wildly different between the two. Iconic comparison, no cap ✨
Table of Contents
1. Introduction ....................................................2
2. IPv6 Address Structure ..........................................2
3. The Host Density Ratio ..........................................3
4. The Role of an Address Efficiency Metric ........................4
5. Network Structure and Address Efficiency Metric .................6
6. Varying the HD-Ratio ............................................7
6.1. Simulation Results .........................................8
7. Considerations .................................................10
8. Security Considerations ........................................11
9. Acknowledgements ...............................................11
10. References ....................................................12
10.1. Normative References .....................................12
10.2. Informative References ...................................12
Appendix A. Comparison Tables ....................................131. Introduction ✨
Okay so address assignment efficiency metrics are a big deal in the RIRs' (Regional Internet Registries') address allocation function 💖 The whole point of using one common metric is that individual networks can be checked against a threshold value in an objective way, no vibes-based decisions. It's the supporting material for an allocation request, showing the network met or beat the threshold, and it also backs up the case for getting another address block 🎀
Public and private IP networks are super different when it comes to purpose, structure, size, and tech, so trying to use one efficiency metric for all of them is honestly a tough job 😭 The threshold has to strike a balance. It needs to be something any competently designed and operated service can actually hit, but it also can't let networks eat up so many address resources that the protocol's long-term viability gets threatened by scarcity. People have all kinds of takes on this, both from theoretical analysis of assignment efficiency and from the practical targets used in today's Internet 🌸
This document compares the efficiency metric and threshold value used for IPv4 allocations with the framework the IPv6 allocation process uses 🦋
2. IPv6 Address Structure ✨
Before we get into allocation efficiency metrics, it makes sense to recap the address structure for IPv6 global unicast addresses 🧚
The general format for IPv6 global unicast addresses is defined in [RFC4291], and it looks like this (Figure 1):
| 64 - m bits | m bits | 64 bits |
+------------------------+-----------+----------------------------+
| global routing prefix | subnet ID | interface ID |
+------------------------+-----------+----------------------------+
IPv6 Address Structure
Figure 1In the current IPv6 allocation policy for the public Internet, the value of 'm' in that figure (the subnet ID) is usually a 16-bit field 💅 So the end-site global routing prefix ends up being 48 bits long, the per-customer subnet ID is 16 bits, and the interface ID is 64 bits [RFC3177] 💜
Here's the tea: the efficiency metric isn't about using individual 128-bit IPv6 addresses, and it isn't about the 64-bit subnet prefix either 🙈 It's only about how efficiently the end-site global routing prefix gets used. This allocation model assumes each customer is given a minimum of one /48 address block. Since that block allows 2^16 possible subnets, a /48 allocation is assumed to be what gets used in the vast majority of end-customer assignments, which is giving main character energy 👑
The discussion that follows assumes the IPv6 address allocation unit is a 48-bit address prefix, and that assignment efficiency means how efficiently /48 allocation units get assigned. That said, the analysis is more general than /48 prefixes. It's about end-site address allocation practices, so it applies to an end-site global routing prefix of any size ✨
3. The Host Density Ratio 🌈
The "Host Density Ratio" was first described in [RFC1715] and then updated in [RFC3194] 🧚
The "H Ratio", as defined in RFC 1715, looks like this:
log (number of objects)
H = -----------------------
available bits
Figure 2The argument in [RFC1715] uses a bunch of examples to back up the claim that this metric is a useful, generic measure of address assignment efficiency across a range of end-site addressed networks 💖 It also says the optimal point for this utilization efficiency metric is an H Ratio value somewhere between 0.14 and 0.26. Lower H Ratio values mean inefficient address use, and higher H Ratio values tend to come with extra network overhead from forced re-addressing operations 😭
Quick reminder: a low H Ratio means wasted addresses, and a high one means you're stuck doing forced re-addressing. Both are a no-go, bestie 💀
This particular metric has a maximal value of log base 10 of 2, which is 0.30103 🍓
The metric got normalized in RFC 3194, which introduced the "HD-Ratio", defined like this:
log(number of allocated objects)
HD = ------------------------------------------
log(maximum number of allocatable objects)
Figure 3HD-Ratio values are proportional to the H ratio, and they range from 0 to 1 🦋 The analysis in [RFC3194] applied this HD-Ratio metric to the examples from [RFC1715]. Based on those examples, it proposed that HD-Ratio values of 0.85 or higher force a network into some form of renumbering. HD-Ratio values of 0.80 or lower were considered an acceptable network efficiency metric ✨
The HD-Ratio is referenced in the IPv6 address allocation policies used by the Regional Internet Registries, and those policy documents say an HD-Ratio of 0.8 is an acceptable objective for address assignment efficiency in an IPv6 network 💅
By contrast, the commonly used efficiency metric for IPv4 is just the ratio of allocated (or addressed) objects to the maximum number of allocatable objects 🫶 For IPv4, the commonly applied value for this ratio is 0.8 (or 80%) 🔥
A comparison of these two metrics is given in Table 1 of Attachment A 🎀
4. The Role of an Address Efficiency Metric ✨
okay so the whole point of this address efficiency metric 🎀 is to give both the allocation entity AND the applicant some objective numbers about how well a network is actually using its address space. 💅 With those numbers they can figure out whether a new allocation is warranted, and get a hint about how big that allocation should be. It also gives a target utilization level, which tells you the moment a network's address resource can be considered "fully utilized" 🦋
The goal here is for the network service provider to deploy addresses across both the network gear AND the customers without having to renumber every so often 😭 It also needs to keep the internal routing system and the inter-domain routing system from getting fragmented and bloated with a ton of extra route objects. 🌸 The way to pull that off is an addressing plan where every level of structure has its own pool of address blocks, so the network can grow at that level without renumbering everything else. It's giving scalable, no cap 👑
Now, a 100% address utilization efficiency metric is unrealistic in any scenario, fr fr 💀 In a typical network address plan, the address space is considered exhausted not when every single address has been used, but when one element of the structure has used up its pool and borrowing from the other elements' pools would mean a massive renumbering headache. 🙈 It's not possible to give one definitive threshold for what overall efficiency is achievable in every IP network, but experience with IPv4 deployments suggests that at any particular level of a hierarchically structured address plan, an efficiency level of between 60% to 80% is an achievable metric in the general case. ✨
This IPv4 efficiency threshold is way higher than what the examples in the HD-Ratio description in [RFC1715] show. 🍓 Those examples come from sources including the Public Switched Telephone Network (PSTN), and this comparison deserves a closer look. There are a bunch of differences between public IP network deployments and PSTN deployments that could explain the gap. IP addresses are deployed per provider and lined up with network topology. 🧚 PSTN addresses are mostly deployed through a geographical distribution system of "call areas" that share a common number prefix. Inside each call area, there need to be enough number blocks from that prefix to let each operator pull their own block from the area pool. In the IP world, service providers don't pull blocks from a shared geographic number pool. They get address blocks from the Regional Internet Registry on a 'whole of network' basis. 🫶 That difference in address structure lets an IP environment reach a higher overall address utilization efficiency. 🔥
Let's talk about how many levels of internal hierarchy IP networks have. If the interior routing protocol is deployed uniformly, it supports a hierarchical network structure only two levels deep: a fully connected backbone "core" with a bunch of satellite areas attached directly to it. 💜 You can get additional levels of routing hierarchy using various forms of routing confederations, but that's not a super common deployment technique. The most common structure in large IP networks is a three-level one using regions, individual Points of Presence (POPs), and end-customers. 🌈
Also, big IP deployments usually have a relatively flat routing structure instead of a deeply hierarchical one. ⭐ To keep the interior routing protocol's dynamic performance healthy, the number of routes it carries is usually limited to the routes for next-hop destinations of iBGP routes. Customer routes get carried inside the iBGP domain and are aggregated at the point where they're announced over eBGP sessions. 🎀 This means that per-POP or per-region address aggregation based on some fixed address hierarchy isn't a necessary feature of large IP networks, so a strictly hierarchical address structure in every part of the network isn't a must in those routing environments. 🦋
5. Network Structure and Address Efficiency Metric 👑
An address efficiency metric can be written using the number of levels of structure (n) and the efficiency achieved at each level (e). 💖 If the same efficiency threshold is applied at every level of structure, the resulting efficiency threshold is e^n. 🧚 This lets us make some extra observations about the HD-Ratio values. Table 2 of Appendix A (Figure 8) shows how many levels of structure are implied by a given HD-Ratio value of 0.8 for each address allocation block size, assuming a fixed efficiency level at all levels of the structure. 🔥 The takeaway is that for large address blocks, the HD-Ratio assumes either a large number of elements in the hierarchical structure, or a very low level of address efficiency at the lower levels. 😭 In IP network deployments, that second situation isn't commonly the case. 💅
The most common interior routing structure in IP networks is a two-level routing structure. 🌸 Given that constrained routing architecture, network address plans seem to commonly use up to a three-level hierarchical structure, while larger networks may generally use a four-level structure. ✨
Table 3 of Attachment A (Figure 9) shows an example of address efficiency outcomes using a per-level efficiency metric of 0.75 (75%), with a progressively deeper network structure as the address block grows. 🍓 This model (termed here "limited levels") caps the maximal number of levels of internal hierarchy at 6, and adds 1 level to the network hierarchy whenever the network grows by a factor of a little over one order of magnitude. 🌈
It's illustrative to compare these metrics for a larger network deployment. 🫶 Say the network is designed to encompass 8 million end customers, each assigned a 16-bit subnet ID for their end site. Then the following table, Figure 4, shows the associated allocation size determined by the address efficiency metric. 💜
Allocation: 8M Customers
Allocation Relative Ratio
100% Allocation Efficiency /25 1
80% Efficiency (IPv4) /24 2
0.8 HD-Ratio /19 64
75% with Limited Level /23 4
0.94 HD-Ratio /23 4
Figure 4Note that the 0.8 HD-Ratio gives a significantly lower efficiency level than the other metrics. 🙈 The limited-level model seems to point to a more realistic efficiency value for networks of this scale (corresponding to a network with 4 levels of internal hierarchy, each with a target utilization efficiency of 75%). 🦋 This limited-level model corresponds to an HD-Ratio with a threshold value of 0.945. ✨
6. Varying the HD-Ratio 🌸
One way to model the range of outcomes from taking a more limited approach to the number of levels of aggregateable hierarchy is to compare various HD-Ratio values against the fixed-efficiency model and the "Limited Levels" model. 💖 That comparison is shown in Figure 5. 🎀
Prefix Length (bits)
|
|
| Limited HD-Ratio
| Levels 0.98 0.94 0.90 0.86 0.82 0.80
| | | | | | | |
1 0.750 0.986 0.959 0.933 0.908 0.883 0.871
4 0.750 0.946 0.847 0.758 0.678 0.607 0.574
8 0.750 0.895 0.717 0.574 0.460 0.369 0.330
12 0.563 0.847 0.607 0.435 0.312 0.224 0.189
16 0.563 0.801 0.514 0.330 0.212 0.136 0.109
20 0.422 0.758 0.435 0.250 0.144 0.082 0.062
24 0.422 0.717 0.369 0.189 0.097 0.050 0.036
28 0.316 0.678 0.312 0.144 0.066 0.030 0.021
32 0.316 0.642 0.264 0.109 0.045 0.018 0.012
36 0.237 0.607 0.224 0.082 0.030 0.011 0.007
40 0.237 0.574 0.189 0.062 0.021 0.007 0.004
44 0.178 0.543 0.160 0.047 0.014 0.004 0.002
48 0.178 0.514 0.136 0.036 0.009 0.003 0.001
Figure 5As this figure shows, it's possible to pick an HD-Ratio value that models IP level structures in a way that behaves more consistently for very large deployments. 👑 In this case, an HD-Ratio of 0.94 is consistent with a limited-level model of up to 6 levels of hierarchy, with a metric of 75% density at each level. 💅 This correlation is indicated in Table 3 of Attachment A. 🫶
6.1. Simulation Results ✨
okay bestie, let's talk numbers 🧮💅 so when you're trying to figure out what happens if you drop the HD-Ratio to a lower value, the move is to test it against actual address consumption data 🦋 and the people running this study used the IPv4 allocation data that the Regional Internet Registries publish [RIR-Data] 🌸 they assumed every end customer gets an IPv4 /32, and that assignments were made based on an 80% density metric in terms of assumed customer count 💖 then that customer count becomes the basis for an IPv6 address allocation, and the HD-Ratio is what maps a customer count to the size of the allocation 🎀
the results here come from simulating an IPv6 address allocation registry 🔥 using IPv4 allocation data from the RIRs covering January 1, 1999 through August 31, 2004 ⭐ the goal is to see the relative level of IPv6 address consumption when you use an IPv6 request size profile built by applying different HD-Ratio values to the derived customer numbers 🌈
the total address consumption profile for selected HD-Ratio values is shown in Figure 6 🧚 and the simulation says that an HD-Ratio of 0.8 eats a total of 7 times the address space that an HD-Ratio of 0.94 does 😭
HD-Ratio Total Address Consumption
| Prefix Length Count of
| Notation /32 prefixes
0.80 /14.45 191,901
0.81 /14.71 160,254
0.82 /15.04 127,488
0.83 /15.27 108,701
0.84 /15.46 95,288
0.85 /15.73 79,024
0.86 /15.88 71,220
0.87 /16.10 61,447
0.88 /16.29 53,602
0.89 /16.52 45,703
0.90 /16.70 40,302
0.91 /16.77 38,431
0.92 /16.81 37,381
0.93 /16.96 33,689
0.94 /17.26 27,364
0.95 /17.32 26,249
0.96 /17.33 26,068
0.97 /17.33 26,068
0.98 /17.40 24,834
0.99 /17.67 20,595
Figure 6so what this is telling us 👀 is that an IPv6 registry will probably see enough spread in allocation request sizes that picking a threshold HD-Ratio actually changes the total address consumption rate 💅 going from 0.8 all the way to 0.99 is about a factor of one order of magnitude in relative address consumption over an extended period of time 📈 the simulation also shows that most allocations (somewhere between 74% and 95% of all address allocations) fit inside the /32 minimum allocation size 🫶 for the rest, the HD-Ratio really matters for a small slice of transactions: the leftovers range from a /19 to a /31 at HD-Ratio 0.8, and from a /26 to a /31 at HD-Ratio 0.99 🙈
the takeaway is that the HD-Ratio affects about one quarter of all allocations, and the remaining ones are handled by the minimum allocation unit of a /32 prefix 🦋 of those "impacted" allocations that are bigger than the minimum, roughly one tenth are "large" allocations 🍓 those large ones have a huge effect on total consumption, and moving the HD-Ratio between 0.8 and 0.99 for them makes a net difference of about one order of magnitude 💜 this is a heavy-tail distribution, which just means a tiny group of big allocations drives the total consumption rate 😳 changing the HD-Ratio barely touches more than 95% of IPv6 allocations, but it creates big swings within the largest 2% of them, and those swings then hit total consumption rates hard 💀
7. Considerations 🎀
so here's the spicy part 🌶️ an HD-Ratio of 0.8, used as a model of network address utilization efficiency, gives extremely low efficiency outcomes for networks on the order of 10**6 end customers and bigger 😭
the 0.8 value also assumes that as the address allocation block gets bigger, the network it's deployed in adds more levels of hierarchical structure 🧚 and honestly, that ever-deeper hierarchy isn't something we commonly see in public IP network deployments 🙈
the fixed efficiency model, which is what IPv4 address allocation policy uses, assumes that as the allocation block grows, the network structure stays at a fixed number of levels 💅 if you add levels, the efficiency at each level goes up a lot 📈 but there's little evidence that adding levels to a network hierarchy actually raises efficiency at each level 🤷♀️
it's obvious that neither model accurately captures real IP network infrastructure and what address deployment actually needs 💀 the fixed efficiency model puts way too much pressure on the operator to hit very high utilization at every level of the hierarchy 😤 which leads to either customer renumbering or deploying things like Network Address Translation (NAT) just to hit the target efficiency value in a hierarchically structured network 🫠 the HD-Ratio model at 0.8 sets an extremely low efficiency target for bigger networks 🥴 and while that doesn't force network architects into renumbering, there's real concern it's an extremely inefficient use of address resources 😬 if IPv6 is supposed to last for decades and span a public network with many billions of end customers and a huge range of end use devices and components 🌍 then there's legitimate reason to worry that 0.8 is setting too conservative a target for address efficiency, meaning the total address consumption targets might get hit way too early ⏰
the study concludes that we should look at whether a higher HD-Ratio value is viable 🦋 because it could be a more relevant model of internal network structure, internal routing, and internal address aggregation in IPv6 deployments 💖
8. Security Considerations 🔐
considerations of various forms of host density metrics create no new threats to the security of the Internet 🫶
9. Acknowledgements 🙏
the document was reviewed by Kurt Lindqvist, Thomas Narten, Paul Wilson, David Kessens, Bob Hinden, Brian Haberman, and Marcelo Bagnulo 🌸
10. References 📚
10.1. Normative References
[RFC1715] Huitema, C., "The H Ratio for Address Assignment
Efficiency", RFC 1715, November 1994.
[RFC3177] IAB and IESG, "IAB/IESG Recommendations on IPv6 Address
Allocations to Sites", RFC 3177, September 2001.
[RFC3194] Durand, A. and C. Huitema, "The H-Density Ratio for
Address Assignment Efficiency An Update on the H ratio",
RFC 3194, November 2001.
[RFC4291] Hinden, R. and S. Deering, "IP Version 6 Addressing
Architecture", RFC 4291, February 2006.10.2. Informative References
[RIR-Data] RIRs, "RIR Delegation Records", February 2005,
<ftp://ftp.apnic.net/pub/stats/>.Appendix A. Comparison Tables ✨
okay so this appendix is basically the receipts 🧾💅 It shows how many /48 end user allocations you'd need before a block counts as "full" under the threshold, comparing the Fixed Efficiency model (0.8) with the HD-Ratio model (0.8). The HD-Ratio column is where things get spicy: bigger blocks hit their threshold with way fewer allocations, which is exactly why the fixed model looks so different. 😭
Fixed Efficiency Value 0.8
HD-Ratio Value 0.8
Number of /48 allocations to fill the
address block to the threshold level
Prefix Size Fixed Efficiency HD-Ratio
0.8 0.8
/48 1 1 100% 1 100%
/47 2 2 100% 2 87%
/46 4 4 100% 3 76%
/45 8 7 88% 5 66%
/44 16 13 81% 9 57%
/43 32 26 81% 16 50%
/42 64 52 81% 28 44%
/41 128 103 80% 49 38%
/40 256 205 80% 84 33%
/39 512 410 80% 147 29%
/38 1,024 820 80% 256 25%
/37 2,048 1,639 80% 446 22%
/36 4,096 3,277 80% 776 19%
/35 8,192 6,554 80% 1,351 16%
/34 16,384 13,108 80% 2,353 14%
/33 32,768 26,215 80% 4,096 13%
/32 65,536 52,429 80% 7,132 11%
/31 131,072 104,858 80% 12,417 9%
/30 262,144 209,716 80% 21,619 8%
/29 524,288 419,431 80% 37,641 7%
/28 1,048,576 838,861 80% 65,536 6%
/27 2,097,152 1,677,722 80% 114,105 5%
/26 4,194,304 3,355,444 80% 198,668 5%
/25 8,388,608 6,710,887 80% 345,901 4%
/24 16,777,216 13,421,773 80% 602,249 4%
/23 33,554,432 26,843,546 80% 1,048,576 3%
/22 67,108,864 53,687,092 80% 1,825,677 3%
/21 134,217,728 107,374,180 80% 3,178,688 2%
/20 268,435,456 214,748,365 80% 5,534,417 2%
/19 536,870,912 429,496,730 80% 9,635,980 2%
/18 1,073,741,824 858,993,460 80% 16,777,216 2%
/17 2,147,483,648 1,717,986,919 80% 29,210,830 1%
/16 4,294,967,296 3,435,973,837 80% 50,859,008 1%
/15 8,589,934,592 6,871,947,674 80% 88,550,677 1%
/14 17,179,869,184 13,743,895,348 80% 154,175,683 1%
/13 34,359,738,368 27,487,790,695 80% 268,435,456 1%
/12 68,719,476,736 54,975,581,389 80% 467,373,275 1%
/11 137,438,953,472 109,951,162,778 80% 813,744,135 1%
/10 274,877,906,944 219,902,325,556 80% 1,416,810,831 1%
/9 549,755,813,888 439,804,651,111 80% 2,466,810,934 0%
/8 1,099,511,627,776 879,609,302,221 80% 4,294,967,296 0%
/7 2,199,023,255,552 1,759,218,604,442 80% 7,477,972,398 0%
/6 4,398,046,511,104 3,518,437,208,884 80% 13,019,906,166 0%
/5 8,796,093,022,208 7,036,874,417,767 80% 22,668,973,294 0%
Table 1. Comparison of Fixed Efficiency Threshold vs
HD-Ratio Threshold
Figure 7okay the next bit is the "how deep does the hierarchy get" question 🌸 The idea is that the messy stuff you see in big deployments comes from having lots of levels of structure in the network. This table works out how many levels you end up with to hit a 0.8 HD-Ratio, assuming 0.8 utilization efficiency at each level. Main character energy for the hierarchy 👑
Prefix Size 0.8 Structure
HD-Ratio Levels
/48 1 1 1
/47 2 2 1
/46 4 3 2
/45 8 5 2
/44 16 9 3
/43 32 16 4
/42 64 28 4
/41 128 49 5
/40 256 84 5
/39 512 147 6
/38 1,024 256 7
/37 2,048 446 7
/36 4,096 776 8
/35 8,192 1,351 9
/34 16,384 2,353 9
/33 32,768 4,096 10
/32 65,536 7,132 10
/31 131,072 12,417 11
/30 262,144 21,619 12
/29 524,288 37,641 12
/28 1,048,576 65,536 13
/27 2,097,152 114,105 14
/26 4,194,304 198,668 14
/25 8,388,608 345,901 15
/24 16,777,216 602,249 15
/23 33,554,432 1,048,576 16
/22 67,108,864 1,825,677 17
/21 134,217,728 3,178,688 17
/20 268,435,456 5,534,417 18
/19 536,870,912 9,635,980 19
/18 1,073,741,824 16,777,216 19
/17 2,147,483,648 29,210,830 20
/16 4,294,967,296 50,859,008 20
/15 8,589,934,592 88,550,677 21
/14 17,179,869,184 154,175,683 22
/13 34,359,738,368 268,435,456 22
/12 68,719,476,736 467,373,275 23
/11 137,438,953,472 813,744,135 23
/10 274,877,906,944 1,416,810,831 24
/9 549,755,813,888 2,466,810,934 25
/8 1,099,511,627,776 4,294,967,296 25
Table 2: Number of Structure Levels Assumed by HD-Ratio
Figure 8last one 🦋 The original also looks at a different deployment model where the hierarchy gets more levels at a slower rate than the 0.8 HD-Ratio model. This one uses a per-level target efficiency of 0.75 and gets compared against an HD-Ratio value of 0.94.
Per-Level Target Efficiency: 0.75
Prefix Size Stepped Stepped Efficiency HD-Ratio
Levels 0.75 0.94
/48 1 1 1 100% 1 100%
/47 2 1 2 100% 2 100%
/46 4 1 3 75% 4 100%
/45 8 1 6 75% 7 88%
/44 16 1 12 75% 13 81%
/43 32 1 24 75% 25 78%
/42 64 1 48 75% 48 75%
/41 128 1 96 75% 92 72%
/40 256 1 192 75% 177 69%
/39 512 2 384 75% 338 66%
/38 1,024 2 576 56% 649 63%
/37 2,048 2 1,152 56% 1,244 61%
/36 4,096 2 2,304 56% 2,386 58%
/35 8,192 2 4,608 56% 4,577 56%
/34 16,384 2 9,216 56% 8,780 54%
/33 32,768 2 18,432 56% 16,845 51%
/32 65,536 2 36,864 56% 32,317 49%
/31 131,072 3 73,728 56% 62,001 47%
/30 262,144 3 110,592 42% 118,951 45%
/29 524,288 3 221,184 42% 228,210 44%
/28 1,048,576 3 442,368 42% 437,827 42%
/27 2,097,152 3 884,736 42% 839,983 40%
/26 4,194,304 3 1,769,472 42% 1,611,531 38%
/25 8,388,608 3 3,538,944 42% 3,091,767 37%
/24 16,777,216 3 7,077,888 42% 5,931,642 35%
/23 33,554,432 4 14,155,776 42% 11,380,022 34%
/22 67,108,864 4 21,233,664 32% 21,832,894 33%
/21 134,217,728 4 42,467,328 32% 41,887,023 31%
/20 268,435,456 4 84,934,656 32% 80,361,436 30%
/19 536,870,912 4 169,869,312 32% 154,175,684 29%
/18 1,073,741,824 4 339,738,624 32% 295,790,403 28%
/17 2,147,483,648 4 679,477,248 32% 567,482,240 26%
/16 4,294,967,296 4 1,358,954,496 32% 1,088,730,702 25%
/15 8,589,934,592 5 2,717,908,992 32% 2,088,760,595 24%
/14 17,179,869,184 5 4,076,863,488 24% 4,007,346,185 23%
/13 34,359,738,368 5 8,153,726,976 24% 7,688,206,818 22%
/12 68,719,476,736 5 16,307,453,952 24% 14,750,041,884 21%
/11 137,438,953,472 5 32,614,907,904 24% 28,298,371,876 21%
/10 274,877,906,944 5 65,229,815,808 24% 54,291,225,552 20%
/9 549,755,813,888 5 130,459,631,616 24% 104,159,249,331 19%
/8 1,099,511,627,776 5 260,919,263,232 24% 199,832,461,158 18%
Table 3: Limited Levels of Structure
Figure 9Author's Address
Geoff Huston APNIC
EMail: gih@apnic.net
Full Copyright Statement
Copyright (C) The Internet Society (2006).
This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights.
This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Intellectual Property 💅
Okay bestie, this part is the legal stuff, but it's important so let's go 🧚
The IETF is not picking a side on whether any Intellectual Property Rights (IPR) or other rights that might be claimed over the technology in this document are valid or in scope 🫠. It also doesn't say whether a license to those rights is available, and it hasn't done any independent digging to find out what rights exist 🙈. If you want to know how all of that works for RFC documents, check out BCP 78 and BCP 79 📚.
Copies of IPR disclosures sent to the IETF Secretariat are available, and so are any assurances of licenses. You can also find the results of anyone trying to get a general license or permission to use these proprietary rights, which is handy for implementers and users of this spec ✨. All of that lives in the IETF online IPR repository at http://www.ietf.org/ipr 🌸.
The IETF also really wants to hear from you 🫶. If you know about any copyrights, patents, patent applications, or other proprietary rights that might cover technology you'd need to implement this standard, tell them! Send it to ietf-ipr@ietf.org 💜.
Acknowledgement 🎀
Shoutout to the people keeping the lights on: funding for the RFC Editor function comes from the IETF Administrative Support Activity (IASA) 👑. Couldn't do it without them 💖🌈
📦 where this page comes from
- ietf: rfc4692 2006-10
- 💖 girlified by claude-haiku-5-5 on : a language model's rewrite of the original, under its terms