dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
3104

aefstoggaflm
Open Source Fan
Premium Member
join:2002-03-04
Bethlehem, PA
Linksys E4200
ARRIS SB6141

aefstoggaflm

Premium Member

[ipv6] Current IPv6 Addressing Waste Full?

Considering that IPv6 is being handled out like candy, does anyone else think that the current IPv6 Addressing is waste full and we will have Deja-Vu (IPv6 NAT and a newer standard IPv7) all over again?

Thank you

PS. Cross Posted over at news.grc.com and at least at DSLR (dslreports.com)

cp
Premium Member
join:2004-05-14
Wheaton, IL

1 recommendation

cp

Premium Member

Re: [ipv6] Current IPv6 Addressing Waste Full?

Yup. A friend and I were talking about this the other day. The EUI-64 format is far too wasteful. We'll find ourselves out of addresses again not too far down the road.
cramer
Premium Member
join:2007-04-10
Raleigh, NC
Westell 6100
Cisco PIX 501

1 recommendation

cramer to aefstoggaflm

Premium Member

to aefstoggaflm
Indeed. But everyone plays the "128bit address, yo!" card. I'm of the opinion we're dooming ourselves to the exact same mistakes from the early days of IPv4. And even 'tho we're only using a ::/3 (1/8th of the space), that doesn't make the mistake remotely correctable in the future. Luckily, I'll be senile or in the grave long before our children/grandchildren are bitten by it.

EUI-64 (aka, let's use our mac address to make our GLOBAL ip address) is the single worst idea to come out of the IPng working group -- and that committee was made entirely of fail. There is absolutely no forseeable reason to need 2^64 lan segments -- and internet of internets for a single f'ing LAN.

I don't have the energy to enumerate all the crap that came out of that working group. IPv6 is the poster child for why you should never design anything my committee. (all they created was a huge ball of glued together personal agendas and political views)
quesix
join:2005-12-19
Cary, IL
ARRIS SB6141
Cisco 2851
Asus RT-AC66

4 edits

1 recommendation

quesix to aefstoggaflm

Member

to aefstoggaflm
worst case i figure we got 50 years....2000::/3 for earth, 4000::/4 for mars 1000::/4 for venus/mercury/solar and 5000::/4 for outer planets... yeah i think we got enough, long as we don't add in FTL and other solar systems.

40 bits max population of full earth inside and out (one trillion)
3 bits redundant links (8 links)
3 bits ISP wastefullness in assignments from stuff like redundant DHCP-PD pools
8 bits subnetting (/56) (128 nanobot swarms will only use half)
64 bits LAN (48 bits mac, 16 bits addresses per device)
===========
118 bits out of 128 bits

that leaves 10 bits or 1,024x for our robot overlords to outnumber us 1000:1
so problem comes with interstellar

The number of atoms in the entire observable universe is estimated to be within the range of 10^78 to 10^82

So yeah it's only wastefulness that will lead to using up all 2^128, which WILL happen, even if it's 5000 years from now as human colonies spread out over 50 light years, not wanting to assign same addresses because of chance someone figures out FTL communications. Inter-Universe communication will require a whole new protocol anyway.

also keep in mind i believe plan is to have more and more devices on a LAN segment, when you got IPv6 only and millions or even billions of hosts (nanobots) on a segment, 64 bits won't seem so much of a waste. SLAAC can be replaced with something that more efficiently uses that 64 bits in far future. It's matter of future proofing not only the number of networks but hosts as well. I'm sure it will only be millions in near future (less than 50 years).

»xkcd.com/865/

leibold
MVM
join:2002-07-09
Sunnyvale, CA
Netgear CG3000DCR
ZyXEL P-663HN-51

3 recommendations

leibold to aefstoggaflm

MVM

to aefstoggaflm
I'd like to see any sane person arguing that the way the IPv6 address space is being utilized is anything other then incredibly wasteful.

For most of the residential/consumer space the lower (LAN) half of the 128-bit address is wasted.

However given that even a 64-bit IP address still means 4 billion times the addresses of the old IPv4 address space I'm not worried about running out of IPv6 addresses during my lifetime.

By the time we run out of 2/3 somebody will hopefully fix SLAAC (or replace it) to allow for longer prefixes and smaller LANs.

NetDog
Premium Member
join:2002-03-04
Hollywood, FL

1 recommendation

NetDog to aefstoggaflm

Premium Member

to aefstoggaflm
said by aefstoggaflm:

Considering that IPv6 is being handled out like candy, does anyone else think that the current IPv6 Addressing is waste full and we will have Deja-Vu (IPv6 NAT and a newer standard IPv7) all over again?

I see it going the classful\classless route like IPv4, I don't see NAT ever coming up as a option.. But you never know, like who would ever use more then 640k for base memory...

justbits
DSL is dead. Long live DSL!
Premium Member
join:2003-01-08
Chicago, IL

3 recommendations

justbits to aefstoggaflm

Premium Member

to aefstoggaflm
There was a discussion about this on the NANOG mailing list recently. Here's an insightful post... »seclists.org/nanog/2013/Dec/115 Read the other posts in the same thread.

The battle between routing bits and LAN bits has been there a long time.

Responsible assignment is going to be a problem if an "ISP" gets a /12, but only ever expects to have a few million customers.

What disgusts me is that ISPs are monetizing IPv6 static addressing.
quesix
join:2005-12-19
Cary, IL
ARRIS SB6141
Cisco 2851
Asus RT-AC66

1 edit

1 recommendation

quesix to aefstoggaflm

Member

to aefstoggaflm
comcat has 2601::/28 (arin assigned) and 2001:558::/31 (legacy), no ::/12's. Small ISPs are only going to get a ::/32 which will add up to 16 million ::/56 DHCP-PD assignments, so only 85% waste is not completely bonkers estimate.

1 million Small ISPs ::/32s with 99.9% waste (10 bits) with 1 billion customers
1000 Large ISPs ::/28s with 80% waste (2.9 bits) with 10 billion customers
average: 3 bits waste or 85%

remember the 60-63bits waste at LAN level (and 8 at subnet level) is already factored in, this is just ISP level waste

NetDog
Premium Member
join:2002-03-04
Hollywood, FL

NetDog to justbits

Premium Member

to justbits
said by justbits:

There was a discussion about this on the NANOG mailing list recently. Here's an insightful post... »seclists.org/nanog/2013/Dec/115 Read the other posts in the same thread.

Yep I have been watching this ..
HELLFIRE
MVM
join:2009-11-25

HELLFIRE to aefstoggaflm

MVM

to aefstoggaflm
said by aefstoggaflm:

and we will have Deja-Vu (IPv6 NAT and a newer standard IPv7) all over again?

...in my lifetime? Not likely.

...in my descendant's decendant's lifetime? Pretty sure... but hey, "it's not my problem" (as human nature attitudes are apt to take in such "long term" concerns...)

Regards
cramer
Premium Member
join:2007-04-10
Raleigh, NC

cramer to justbits

Premium Member

to justbits
(It comes up all the time on NANOG. Pretty much every time IPv6 is mentioned, in fact. )

ropeguru
Premium Member
join:2001-01-25
Mechanicsville, VA

ropeguru

Premium Member

Yep. I really like the current discussion that is going on with the "turning on comcast v6" subject.

»seclists.org/nanog/2013/Dec/243

NetDog
Premium Member
join:2002-03-04
Hollywood, FL

NetDog

Premium Member

said by ropeguru:

Yep. I really like the current discussion that is going on with the "turning on comcast v6" subject.

»seclists.org/nanog/2013/Dec/243

I didnt see that looking at it now..
34764170 (banned)
join:2007-09-06
Etobicoke, ON

34764170 (banned) to HELLFIRE

Member

to HELLFIRE
said by HELLFIRE:

said by aefstoggaflm:

and we will have Deja-Vu (IPv6 NAT and a newer standard IPv7) all over again?

...in my lifetime? Not likely.

...in my descendant's decendant's lifetime? Pretty sure... but hey, "it's not my problem" (as human nature attitudes are apt to take in such "long term" concerns...)

Regards

It wouldn't be in your descendant's descendant's descendant's lifetime. The world would have much bigger issues to deal with before that would be an issue like say... where all these people are going to live on this planet.

aefstoggaflm
Open Source Fan
Premium Member
join:2002-03-04
Bethlehem, PA
Linksys E4200
ARRIS SB6141

aefstoggaflm

Premium Member

said by 34764170:

said by HELLFIRE:

said by aefstoggaflm:

and we will have Deja-Vu (IPv6 NAT and a newer standard IPv7) all over again?

...in my lifetime? Not likely.

...in my descendant's decendant's lifetime? Pretty sure... but hey, "it's not my problem" (as human nature attitudes are apt to take in such "long term" concerns...)

Regards

It wouldn't be in your descendant's descendant's descendant's lifetime. The world would have much bigger issues to deal with before that would be an issue like say... where all these people are going to live on this planet.

Umm..

#1 Last I heard/read there is 7 billion people.

#2 Also last, I heard there is more computers/devices than people.

#3 And remember each computer must have it's own IP Address.

»www.silabs.com/products/ ··· ngs.aspx
34764170 (banned)
join:2007-09-06
Etobicoke, ON

34764170 (banned)

Member

said by aefstoggaflm:

Umm..

#1 Last I heard/read there is 7 billion people.

#2 Also last, I heard there is more computers/devices than people.

#3 And remember each computer must have it's own IP Address.

»www.silabs.com/products/ ··· ngs.aspx

You could assign hundreds of thousands of IPv6 addresses to each device/thing on earth and still only use a small percentage of the address space. Even if the number of devices were to go up by 10 times or more that still wouldn't be an issue.
Shady Bimmer
Premium Member
join:2001-12-03

3 recommendations

Shady Bimmer to aefstoggaflm

Premium Member

to aefstoggaflm
said by aefstoggaflm:

Umm..

#1 Last I heard/read there is 7 billion people.

#2 Also last, I heard there is more computers/devices than people.

#3 And remember each computer must have it's own IP Address.

»www.silabs.com/products/ ··· ngs.aspx

With a 64-bit network prefix, there are over 18 quintillion networks, each with 18 quintillion "devices" (IPv6 address is 128 bits, split in the middle)

1 quintillion is 1 billion billion.

While current thinking is that there is plenty of address space, any scheme with a fixed-length address is at risk of exhaustion regardless of how large that scheme may be.

The thinking is that each "client" would receive its own network space. The question will be when the 18 quintillion individual client/customer limit is reached globally.

Regardless of the actual sizes (before anyone seeks to "correct" the above numbers), an IPv6 address is a fixed-length. While the address space is currently greater than what could conceivably be used today the future is really a big unknown.
cramer
Premium Member
join:2007-04-10
Raleigh, NC
Westell 6100
Cisco PIX 501

2 recommendations

cramer

Premium Member

said by Shady Bimmer:

(IPv6 address is 128 bits, split in the middle)

WRONG. This is the entire problem with this SLAAC bullshit: it teaches people to think the wrong thing from the start. IPv6 IS A 128 BIT ADDRESS - PERIOD - END OF DISCUSSION. It's classless; there is no network part and host part. The only backwards, brain damaged part that demands a LAN segment be ::/64 is the asinine Stateless Address Auto-Configuration, that has no real reason to require that division. PLUS, no one is required to use SLAAC. (and, in fact, many enterprises won't use it.)

[Sure, put forward from first chair around the conference table in 1990, it was a great idea. And in the early '90's -- an era of 100Mhz, 16MB computing -- it was an important concept... the ability to determine an address quickly, with zero complexity, in the least code possible. However, that optimization evaporated by the time it got to the fifth chair where a complete g** d**** IPsec engine was bolted into the specs. Of course, by the time anyone bothered to read any of the RFC's (circa 2000), the landscape was very different -- people were walking around with fast processors with significant memory and storage. The optimization afforded by SLAAC has been moot pretty much forever, as the "embedded devices" of the era were never going to be capable of running IPv6 -- a Z80 or VIC20 CPU with an IPv6 statck, just, no; that's even more so with the advent of privacy extensions, which adds random address generation and duplicate peer detection.]
Shady Bimmer
Premium Member
join:2001-12-03

1 recommendation

Shady Bimmer

Premium Member

said by cramer:

said by Shady Bimmer:

(IPv6 address is 128 bits, split in the middle)

WRONG. This is the entire problem with this SLAAC bullshit: it teaches people to think the wrong thing from the start.

I don't disagree one bit. This is why I specifically made the comment about anyone trying to "correct" my statements - specifically to (hopefully) prevent this.

I know the details about an IPv6 address very well. I also know that there are probably about five other thoughts that would also contradict my statements (and yours similarly). Different providers have their own interpretations and have their own assignments. What is "right" and what is "wrong" really makes no difference in this thread. In the end it is all completely irrelevant to this discussion.

The need to rant about this was not called for, especially after I already noted that the details really are not pertinent in this discussion. It is a fixed length address and regardless of the actual length is there is still a fixed limit that can not be changed. That limit may be very large by current expectations but it is still a fixed limit.
AVonGauss
Premium Member
join:2007-11-01
Boynton Beach, FL

AVonGauss

Premium Member

said by Shady Bimmer:

It is a fixed length address and regardless of the actual length is there is still a fixed limit that can not be changed.

Are you referring to the 128-bit address or the concept of 64+64 in this statement? If the later, while he wasn't friendly about it, Cramer is spot on when he stated IPv6 is classless - albeit with some current policy conventions that could be changed at any time.
cramer
Premium Member
join:2007-04-10
Raleigh, NC
Westell 6100
Cisco PIX 501

cramer

Premium Member

I've never been "friendly" about it. Even as far back as '95(?) -- it was 80+48 then.

(also, the notion of "RA"s... that crap was abandoned in the IPv4 world before anyone learned to spell "internet" -- see also: ICMP router advertisement. aka, the reason step two of installing SunOS was "touch /etc/notrouter" -- everyone was taught to do it, but not why.)
Shady Bimmer
Premium Member
join:2001-12-03

1 recommendation

Shady Bimmer to AVonGauss

Premium Member

to AVonGauss
said by AVonGauss:

said by Shady Bimmer:

It is a fixed length address and regardless of the actual length is there is still a fixed limit that can not be changed.

Are you referring to the 128-bit address or the concept of 64+64 in this statement?

This thread is about "wasteful" allocation of IPv6 blocks. It has nothing to do with whether or not IPv6 addressing is classless. I will also note that there were several other references to the exact same split earlier in this thread that were completely ignored (likely because it really is not relevant to the discussion).

I certainly don't disagree with cramer See Profile's sentiments, but that is neither here nor there. Many ISPs are allocating blocks that avoid any potential conflicts or issues with SLAAC. Regardless of one's own opinions this is what is being done and that is my only reference.

And again, this is why I made a specific statement to hopefully avoid attempts to "correct" my statement taking this thread off topic.

Back to clarify my original statement, perhaps I should have said "if" - would that have made everyone all cheery and happy? If I had used another split instead of 64 + 64, would it have made everyone happy?

Perhaps to add more to the point: An IPv6 address is 128 bits long. Period. End of story. It is a fixed length and as a result has a fixed number of addresses. Period. End of story. Being a fixed length, there is opportunity that the address space will be exhausted, regardless of whether this can be conceived today or not. Even if ISPs gave out only single addresses (instead of blocks), there is still a fixed number of addresses - regardless of how "big" that number may be.

Going to the next step, with a 64-bit prefix there are still 18 quintillion address block allocations available. Even with a 48-bit prefix there are over 280 trillion allocations available.

In the long (or very long term) it is very possible there will be address space exhaustion, regardless of the size of blocks allocated. In the shorter term, I would not disagree that it is wasteful, but does it really matter?

As a follow-on to the original question by the OP, if we concede that current allocations are "wasteful", what would be considered a "not wasteful" allocation?
quesix
join:2005-12-19
Cary, IL
ARRIS SB6141
Cisco 2851
Asus RT-AC66

1 recommendation

quesix to aefstoggaflm

Member

to aefstoggaflm
yep it's wasteful but more than current technology can handle, like border routers running up against BGP route limits already, even with only 500k total Ipv4/IPv6 routes per table and only two tables, or 500 host limit per network segment, but who knows what the future will bring, and whether the 64/64 split was a good idea. Who knows which will grow faster, number of ISPs and hence routes, or nano bots in everyone's home serving every whim of human kind, or gosh a self replicating von Neumann probe with FTL communications.

»www.google.com/#q=bgp+mi ··· te+limit

»www.google.com/#q=hosts+ ··· +segment

»en.wikipedia.org/wiki/Se ··· acecraft

mackey
Premium Member
join:2007-08-20

3 recommendations

mackey

Premium Member

Reminds me of »xkcd.com/865/

/M
ViperJohn
join:2000-03-23
Santa Ana, CA

4 recommendations

ViperJohn to aefstoggaflm

Member

to aefstoggaflm
said by aefstoggaflm:

Considering that IPv6 is being handled out like candy, does anyone else think that the
current IPv6 Addressing is waste full and we will have Deja-Vu (IPv6 NAT and a newer standard IPv7) all over again?

While it seems like it is wasteful In reality it isn't. Early on I felt the same way until I actually did the math. The size of
the numbers we are talking about with 128 bit IPV6 addressing is truely mind bending. They are damn near impossible
for most people to relate to and get their heads around. So lets run some numbers here.

IANA has only released a small portion of the IPv6 address space for PUBLIC addressing. Only addresses in the range
from 2000:0000:0000:0000:0000:0000:0000:0000 to 3FFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF are in use for public
IP addressing.

That is 1FFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF hex available addresses which is decimal
42,535,295,865,117,307,932,921,825,928,971,026,431 individual public IPV6 addresses. (Yeah like get your head around
that lol).

By design Stateless Address Autoconfiguration (SLAAC) subnets require a /64 address block, as defined in RFC 4291
so a /64 is the standard address allocation to the end user networks (or rather typically our routers for all users behind
it). That works out to a FFFF FFFF FFFF FFFF hex address block or 18,446,744,073,709,551,615 (not quite 18.5 million
billion) individual public addresses available for each end user network. Now this may seem wasteful but it is pretty
much a moot point as we will see.

Now if we devide 42,535,295,865,117,307,932,921,825,928,971,026,431 by 18,446,744,073,709,551,615 we get
2,305,843,009,213,693,952 (that over 2.3 MILLION BILLION) /64 address blocks for allocation to end user networks
with just the small IANA IPV6 public address space release.

Now the current world population at a bit over 7 billion so in all likely hood by the time we run out of IPV6 addresses
the earth will just be a burnt out cinder orbiting a long ago dead sun.

Good read here - »rednectar.net/2012/05/24 ··· -really/

Viper
IamGimli (banned)
join:2004-02-28
Canada

1 recommendation

IamGimli (banned) to leibold

Member

to leibold
said by leibold:

I'd like to see any sane person arguing that the way the IPv6 address space is being utilized is anything other then incredibly wasteful.

For most of the residential/consumer space the lower (LAN) half of the 128-bit address is wasted.

Unused != wasted. Just because you don't use it today doesn't mean you won't use it in the future and allocating it ahead of time means faster implementation of the technologies which will require it in the future. It's actually very forward-looking. Waste only exists if the thing not being used is actually required for another purpose. That is not the case.

Just think of it this way: when they came up with 10-digit phone numbers for North America most households had less than 2 telephone units. They figured that'd be more than enough numbers for everyone for all of our life times yet we are running out of phone numbers today because the number of phones has exploded. The result is that phone numbers as we know them will become obsolete within our lifetime and will be replaced with... IP addresses. What if when they came up with 10-digit phone numbers they had actually allowed for up to 20-digit phone numbers instead, with each individual getting a 10-digit allocation to further define as they see fit. Now instead of having different numbers for pretty much every device you have they could all share the same 10-digit allocation, and you'd have a lot more power to control your own devices (such as, for example, make rules so that all of your devices ring at the same time when someone dials your 10-digit number, or create subnets of devices for more detailed dialing, etc.) and we wouldn't be running out of phone numbers.

It's that same approach they've taken with IPv6, and it's the right one. Some day each individual will receive it's /64 assignment from birth and will "own" it throughout their lifetime, no matter what addressable devices they associate with it, who their ISP is, where they are, etc. Technology isn't at that level right now but it will get there eventually, and we won't require another addressing standard to adopt it as it will already exist. On top of that we can still use that addressing standard for "legacy" applications and technologies with no fear of "running out". What more could we ask for?
quesix
join:2005-12-19
Cary, IL
ARRIS SB6141
Cisco 2851
Asus RT-AC66

4 edits

quesix to aefstoggaflm

Member

to aefstoggaflm
individual ::/64 assignments are not going to happen any time soon. It's going to be dynamically assigned ::/60s and ::/56s from your ISPs depending on service level, that your future router and PC will need to work with to auto assign to devices, and route backup connections by using multiple ::/64 sub-assignments.

so:
ISP1 Comcast 2601:x:xxx:xxx0::/60 DHCP-PD assigned
ISP2 local WISP 2001:678:xxxx:xxxx::/64 static assigned
ISP3 AT&T 2602:30x:xxxx:xx00::/60 6rd tunneled

combined so device on primary LAN gets
2601:x:xxx:xxx0::/64 address 2601:x:xxx:xxx0:xxxx:xxff:fexx:xxxx/128
2001:678:xxxx:xxxx::/64 address 2001:678:xxxx:xxxx:xxxx:xxff:fexx:xxxx/128
2602:30x:xxxx:xx00::/64 address 2602:300:xxxx:xx00:xxxx:xxff:fexx:xxxx/128
with 2 more addresses from each block for privacy extensions/outbound connections

with say a 5min timeout on each netblock in case of failure, and priority of comcast/WISP/6rd or similar in that order... all this handled by OS automatically and by SLAAC/Router Advertisements. Much updating is needed of routers and OSes to make this work right long term. (given that many routers have trouble with basic DHCP-PD)

Only high end Businesses with multiple fiber connections might use summary only BGP and a ::/32 assignment. Servers will need to be virtual at places like Amazon with DNS pointing at amazon's static block otherwise. Many small businesses won't have servers on site in future, and if they do it would use IPv4 in short term with VPN, and IPv6 with dynamic IP type DNS thru DNS hosting provider based on current ::/56 or ::/60 assignments (paid services.. free ones seem to be all shutdown now).

p.s. This is all for routing efficiency. With only ::/48 or larger in global routing tables (preferably ::/32 except where only one or two ::/48 of the ::/32 is being used and routed over only a few upstream connections) even at this level we can expect global tables to increase from today's 15k upto 200 million in far future (up from Ipv4's ~500k which is increasing at 100k per year, works out to about 400 years for 200 million IPv6 routes) [200 million is assuming ::/28s per ISP like AT&T and Comcast use now. with ::/32s assigned to most that can increase to billions]

correction AT&T 6rd is only a ::/28 so ::/60 prefix per IPv4 address not ::/56 (28+32 of IPv4 = 60) 2602:030H:HHHH:HHH0::/60 H = hex of IPv4 public address of your 6rd capable router. first 7 digits *4 = 28 bits, next 8 *4 = 32 bits IPv4.
lestat99
join:2000-08-04
Piscataway, NJ

5 recommendations

lestat99 to leibold

Member

to leibold
First you have to get your head around just how many IPv6 addresses there are. We are talking about 79 trillion trillion (yes two trillions) *times* the number of IPv4 addresses. In other terms we could assign 523 Quadrillion IPv6 address for every human brain cell on the planet.

Second you have to understand that subnetting consistently at a /64 has significant efficiencies for internet routing tables not to mention the ease of administration when not having to deal with VLSM. One could also argue that IPv4 VLSM is wasteful as well.

Third, subnetting on a /64 allows for the plug and play features of IPv6. You can't do SLAAC without the /64 boundary.

No to say that we will never run out of address (maybe we will start assigning addresses to every cell in our body) but frankly I think the benefits of a /64 far outweigh that risk.
lestat99

lestat99 to quesix

Member

to quesix
Sorry, didn't see that you were talking about a ::/64

timcuth
Braves Fan
Premium Member
join:2000-09-18
Pelham, AL

timcuth to aefstoggaflm

Premium Member

to aefstoggaflm
What lestat99 See Profile said.

Tim