Quick IPv6 Question

Hey, guys. I'm reading through Routing TCP/IP and I've always kind of struggled with ipv6. I've started to become a little more comfortable with the addressing scheme until halfway through the ipv6 chapter.

Can someone help clear up how the Global Routing Prefix and Subnet ID work together with the in general prefix(eg. /64, /48/, etc)?

At first I just thought that the in general prefix would work as the network portion. Say that I had 3ffe:1944:100:a::bc:2500:d0b with a /48 prefix, I assumed that the network number would be 3ffe:1944:100/48 and the rest would determine the specific host. But then I got to the next section and it explains that the first 48 bits are to the GRP and the next 16 to the Subnet ID. How does that all work together? What if the prefix is larger than 48 bits? Smaller I'd imagine there would just be some 0's. But I'd REALLY appreciate some help with this. ^.^

EDIT: Okay, maybe this turned out to be more than a quick question.

Find more posts tagged with

Comments

mattau

that's how i look at it. /48 and the rest host. But in theory the next 16 are for subnets because of the EUI 64 rule. if you go past /64 you cant use the EUI 64 way of host addressing but nothing stopping you from manual assigning.

the ISP I am with gives out /56 prefixes which means you are left with 8 bits to allocate to subnets if you want to do EUI 64

elderkai

That helped a liiittle bit, but man I'm just having trouble with ipv6 and I'm really not sure why. D;

elderkai

Are there any good books dedicated to ipv6 for this kind of confusion?

EDIT: Actually, I think I'm starting to get it a little bit. If I understand correctly, you won't be dealing with a prefix less than /48. So you get an address with a /48 mask from the ISP and the 16-bit Subnet field is to help you for your internal usage/preference/common subnetting leaving 64 bits for EUI-64. Please correct that train of thought if it's slightly wrong. :P

mattau

ipv6 addressing is pretty confusing, especially when first learning it. It gets easier when you lab and try stuff out. Specifically the cisco ios tells you what subnet you are in and the host address configured under the interface. That's how I learnt how most of it worked, as majority of books dont explain ipv6 subnetting well at all.

the way I look at it is the longer the network bits the less hosts you have, and shorter network bits you have the more hosts you have.

and if you dont worry about the 16 subnet bits and just focus on network bits and host bits its all similar to ipv4. Like i said I am not an expert but it seems like the 16 subnet bits are there to subnet your organization if you want to use the eui 64 feature. Because for the eui 64 feature to work you need the last 64 bits free so I think they just put the 16 subnet bits in the middle so this process isnt interrupted. But there is nothing stopping you from using a /126 for example. Its just you wont be able to use the EUI-64 and instead you will have a heap more subnets but less hosts on each of those subnets.

but I believe your train of thought is correct

eg say you set fa0/1 to this address like your example

3ffe:1944:100:a::1/64 = the first host on subnet 10

the cisco ios will say

3ffe:1944:100:a::1 subnet is 3ffe:1944:100:a::/64

elderkai

Pretty obvious question, but ill ask anyway. Are the double colons standard in representing a subnet, or are there some 0's in that subnet? Seeing as it's /64 and everything up to the double-colons equal 64 bits, just wanted to make sure. What if there was one or two sections of 0's? Would it still be represented the same just with the prefix, so it knows exactly have many to count?

mattau

you can only have 1 set of :: in an ipv6 address.

a subnet is always represented by :: in the ios

its just like ipv4

like a /24 is represented as 192.168.1.0/24 the 0 at the end is the same as :: in essence

2001:1111:1111:1::/64 is = to 2001:1111:1111:0001:0000:0000:0000:0000 ( all the zeros after the /64 are free for hosts just like all bits after /24 are for hosts.

elderkai

OHH! Oh my gosh hahaha. That just made so much sense to me that I had a rush of satisfaction and I don't know why I didn't see that. I think I try to make ipv6 harder than it is just because it's *face of disgust* new.

Still have problems thinking about it, but it makes a little more sense. Just a little more. That 16-bit Subnet field still gets to me though. Does that mean that you can't use those 16-bits for hosts even if you have a /48 prefix? Oh, man, this is just like pulling hair. Lmao.

mattau

lol yeah its pretty confusing. ask as many questions as you want. I enjoy these forums as I can go back and refresh on stuff and learn at the same time if someone asks a question.

I spent weeks trying to figure out that same question with the 16 bit subnet field and if it can be used for hosts if you have a /48 for example.
I am assuming it can but you must specify it as a /48 because a /48 is 128 - 48 = 80 hosts bits available

for example - underline indicates the network and the rest hosts

fa1/1
ipv6 address 2001:1111:1111::1/48

what happens if you put a 1 in the 16 bit subnet field on a /48 ?

2001:1111:1111:1::1, subnet is 2001:1111:1111::/48

its still in the same subnet so it means if you dont use the extra 16 bit field and specify a /64 you get more hosts.

This might be off topic abit but a major clicking point for me to understanding ipv6 addressing was to see how a /126 works or something higher.

88.jpg

elderkai

Oh, man. This is really starting to click.

So, essentially, a /126 prefix would be the common implementation for a point-to-point link between routers instead of /30, right? And instead of having a class A, B, and C, you just work off of whatever prefix you're assigned! I feel dumb coming to some of these conclusions, but man it's just tough for me to get into the ipv6 mindset flawlessly. Doesn't help that there aren't that great material on it.

You also helped me realize I need to get into thinking binary > hex better. My binary > decimal brain has complications looking at a long ipv6 address, but if I can just make thinking in binary > hex easier things should come a lot faster.

If there's anything else that you've learned in your studies and thought that you could have explained it better or just you think I should know, feel free to share.

mattau

well technically the equivalent to a /30 point to point ipv4 link would be /127 as there is no concept of broadcast in ipv6. You know how on a /30 for example 192.168.1.0/30 the range is 192.168.1.1 - 1.3 with .3 being the broadcast address and .1 and .2 being the 2 "usable" host addresses. with ipv6 you could have a point to point like this

2001:1111:1111::/127 point A
2001:1111:1111::1/127 point B

because a /127 is 2 hosts per subnet ( the cisco ios shows it like this )

2001:1111:1111::, subnet is 2001:1111:1111::/127 - 1st host on the 1st subnet
2001:1111:1111::1, subnet is 2001:1111:1111::/127 - 2nd host on the 1st subnet
2001:1111:1111::2, subnet is 2001:1111:1111::2/127 - now this is the next subnet as we are going up in increments of 2 due to the /127 (1 bit = 2 )
2001:1111:1111::3, subnet is 2001:1111:1111::2/127 - 2nd host on the subnet 2
2001:1111:1111::4, subnet is 2001:1111:1111::4/127 -1st host on subnet 4
2001:1111:1111::5, subnet is 2001:1111:1111::4/127 -2nd host on subnet 4
etc

But yea there is no concept of classes now. But people in books try do comparisons to make it easier to see the similarities with ipv4 and ipv6 and say a /64 is like the basic stock standard /24 address.

The thing that was key for me was the binary to hex logic. It all makes sense when you combine ipv4 logic and just convert it to hex.

like in ipv4 if you wanted 64 hosts per subnet you would know that your subnets go up in increments of 64. In hex you can do this exactly the same except you just need to find out the equivilant of 64 in hex

decimal
0
64
128

hex

0
40
80

I practised this by basically doing labs where Id say to myself ill make subnets like i did in ipv4 with smaller hosts per subnet. Just staying away from the huge ipv6 numbers to begin with. Like just make subnets with 32 hosts in each using a /123. Then after I did that I did the same thing using the 16 bit subnet field for subnets instead and realising its the exact same thing except you can use eui-64 if you want or if you dont want to then understanding that the extra 64 bits could be used for hosts that you number manually.

elderkai

I know that there isn't a broadcast address, but what about the network address? I thought that it was just instead of 2^1-2 it would be 2^1-1, still leaving only one usable?

Also, could you kind of show a small example of a VLSM task being worked out? I know that's probably easy, but I'm just trying to really get a hold of this. Sorry for asking so many questions.

EDIT: In the real world, are we still going to even be creating subnets small enough for just 60 hosts, 30, 110, etc? For some reason I feel like they expect us to be wasteful, but I haven't read that much ipv6 stuff, so..:P

mattau

I just found that to be the case in real labs : ) I dont know the exact specifics sorry. I believe 0 is a usable address now.

The only reason I do the longer subnet masks is to understand it better. Otherwise its easy to get confused with the massive numbers in hex. It is all the same when you have anything below like /120 its just massive numbers involved.

I have watched various videos and CCIE's mention it is good etiquette to not waste address space but to be smart with your allocation. I guess you dont have to be tight but maybe dont use /64 on point to point links for example as that is wasteful.

To be honest I havent found any VLSM examples with ipv6 and I dont think its even used but I would not have a clue about that aspect of ipv6 in the real world. But this is how I would do a VLSM with ipv6. Again using a longer prefix length to help illustrate the similarities with ipv4.

say you want to do this

1 subnet with 64 hosts
1 subnet with 8 hosts
1 subnet with 4 hosts

just work it out with it would be in decimal and change it to Hex

2001:1111:1111::/122 0 to 39 - in hex ( decimal 0 - 63 ) - note if you count the 0 as being an address it actually works out to be 64 usable addresses.

2001:1111:1111::40/125 40 to 47 - in hex ( decimal 64 - 71 )

2001:1111:1111::48/126 48 to 4B - in hex ( decimal 72 - 75 )

0 to 64 subnet
2001:1111:1111::, subnet is 2001:1111:1111::/122
2001:1111:1111::1, subnet is 2001:1111:1111::/122
2001:1111:1111::39, subnet is 2001:1111:1111::/122

64 to 72 subnet
2001:1111:1111::40, subnet is 2001:1111:1111::40/125
2001:1111:1111::47, subnet is 2001:1111:1111::40/125

72 to 76 subnet
2001:1111:1111::48, subnet is 2001:1111:1111::48/126
2001:1111:1111::4B, subnet is 2001:1111:1111::48/126

elderkai

Okay, so that is way more comprehensible than I thought.

Man I REALLY appreciate all the help you've been. That's awesome. Would you mind if I added you as a contact to maybe have a later discussion, seeing as I'm about to start digging into the ROUTE material? I didn't really need to ask TOO many questions outside of the classroom for the CCNA, but doing the CCNP self-study over the summer while waiting to get an internship starting the fall is kind of different. ^.^

mattau

you're welcome. Glad it wasn't too confusing. and if you find out that any of its incorrect let me know so i can learn too.

sure thing. I am always on these forums so feel free to contact me if you need some help or just do a new post and there are heaps of friendly people around here to help.

elderkai

It wasn't confusing at all. The questions and confusion I had were all on me.

I did forget a question, or rather reassurance, though. So for the 16-bit subnet field, if I feel like using a /48 prefix the 16 bits will still be used for hosts, right? And in the case that I do want to use EUI-64, I would have to just switch it to a /64 prefix, but that's the only real thing I should worry myself about it at this point in my studies though right? Also, I did a quick search regarding using the network number to assign, but sadly I couldn't come up with the right combo of keywords for google to give me what I want and kept getting articles about NAT and IPv6 in general. I shall just lab later and find out. ^.^

mattau

yeah that should be correct.

/48 will mean everything after /48 is host bits

if you want to use eui 64 then make it a /64 using the 16 bit subnet field

if you are doing ccnp route go through the various texts and get your hand on the student lab manual and go through the labs in there and you should be fine.

http://www.google.com.au/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CFUQFjAA&url=http%3A%2F%2Fwww.hh.se%2Fdownload%2F18.53cc3a9212c99c4ab2b80001119%2Fen_ROUTE_SLM_v6000.pdf&ei=_EHcT5XXJs2ZiQf3rcyQCg&usg=AFQjCNEtisfUsNNDZ3ntfeuLGG9aEVrCAA&sig2=R1sBu42UGKmIdUaVu07T9A

elderkai

Thanks for the resource.

I plan on taking the ROUTE and SWITCH kinda slow. Not too, slow, but I'm not sure if I'll stress about taking an exam before the summer is over. Mainly because I'm 17 and $200 is a lot without a job, haha, but also because I just want to become a really strong networker.

I've heard a lot of good about Routing TCP/IP so I think I'm going to read through that and really expand my knowledge, along with a bunch of labbing, Chris Bryant's study guides, and I'm going to try Jeremy Ciorra's ROUTE series and see if it's good. His CCNA series lacked a lot, but we'll see. ^.^

EDIT: Also, I think I know why you can use a /127 prefix and use the network number. From where I haven't gotten deep into my ROUTE studies at all, I didn't know about RFC 3021. It covers the ability to use a 31-bit mask for a ptp link which would be the equivalent of using a /127 ipv6. So that confusion was my fault.

MrBrian

Nice discussion between some fellow "colleagues," if you will. I find it cool that none of us know each other, but we're all helping each other through the use of the Internet, and technology in general. I guess that's obvious, but I think when I step back sometimes it amazes me. People had to be face to face to have such conversations in years past. elderkai, sounds like you're gonna be grinding it out this summer, awesome! I'll be spending a lot of time reading/labbing this summer as well, and will be on here frequently I suppose.. feel free to post whatever questions you have to the forum and I'm sure people will be willing to help. You say you're only 17 which is pretty amazing. Keep that level of curiosity and it will take you far! I still have a long way to go in my overall understanding of things too, we're all learning. Who knows what the technology landscape will be like in 20 years, gotta stay updated.

And so this post is somewhat relevant, I too am getting more curious about IPv6. I'm equipped with some basic information on it, as I've passed ROUTE, but that's not enough to give me that confident feeling yet. I just haven't honestly had enough interaction with it in labs yet. I'll surely be seeking out some good IPv6 books this summer to push me along with it. I guess its onwards to the amazon book reviews for me..

Forsaken_GA

MrBrian wrote: »

And so this post is somewhat relevant, I too am getting more curious about IPv6. I'm equipped with some basic information on it, as I've passed ROUTE, but that's not enough to give me that confident feeling yet. I just haven't honestly had enough interaction with it in labs yet. I'll surely be seeking out some good IPv6 books this summer to push me along with it. I guess its onwards to the amazon book reviews for me..

ipv6.he.net

MrBrian

Forsaken_GA wrote: »

ipv6.he.net

Nice, thank you

wave

Some info on point-to-point addressing in IPv6:

IPv6 Addressing Plans - ARIN IPv6 Wiki

IETF expects that you will assign a /64 for point-to-point links

Fewer typos because all subnets are the same size
You can use longer prefixes but what's the point?
/126 will break Mobile IPv6 Home Agent discovery
/112 leaves final 16 bits free for Node IDs
Use /64 unless you have read and understand RFC 3627
Note: on pure point-to-point links (e.g., SONET) anything shorter than /127 is vulnerable to ping-pong packet amplification as described in Maz's APNIC 26 presentation. (On Ethernet, this is at most a neighbor cache DoS)

elderkai

Ohhh. Very good find.