Jump to content


IPV6


  • Please log in to reply
37 replies to this topic

#31 cistechie

cistechie

    Newbie

  • Members
  • Pip
  • 1 posts

Posted 18 February 2010 - 01:46 AM

No problem at all opening this document....
  • 0

#32 greywolf82

greywolf82

    Newbie

  • Members
  • Pip
  • 5 posts

Posted 17 April 2010 - 08:01 AM

I just finished studying for my BSCI (passed it, YAY) and struggled with one aspect of IPv6: subnetting. Specifically, subnetting on point-to-point serial links, loopback interfaces, etc.

Early on in the IPv4 world, we had lots of addresses and we wasted them without realizing the cost. After a while, we realized that we could no longer afford to do this and subnetting was created. We started being very smart about how we allocated subnets, using /30s on point-to-point links and /32s on loopbacks, etc. We did this because addresses were scarce and we needed to minimize their waste. Those of us who learned networking after this were taught to conserve addresses early on and had it drilled into us.

Now, in the IPv6 world, there appears to be a wanton disregard for this lesson. The Cisco Press book I used (here) says that everything gets a /64 mask. I was stunned. Even if my ISP gives me a /48 and I can subnet that to 65536 /64s, I am wasting millions of addresses on my serial, loopback, tunnel, and dialer interfaces.

After I calmed down, I remembered that the folks who designed this are a lot smarter than I am and couldn't have done this without good reason. After some searching, I found that the Global/Local bit (71st bit in a 128 bit address) indicates if an address is globally unique. If we subnet beyond the 64th bit, we run into problems where this bit doesn't represent what the designers wanted it to.

Again, I was stunned - and here is where I need help. I've configured IPv6 labs in GNS3 and done it both ways: one lab with 64 bit subnets for everything, and one using /126s or /128s where appropriate. Both work. But everywhere I search, I run into the same thing: don't subnet beyond /64 or you'll run into problems because of the G/L bit.

Does anyone have or know of a simple, straightforward tutorial or write up that can clear this up for me? Has anyone here used IPv6 in a production network and have used long subnet masks with or without problems? Just because I have lots of addresses doesn't mean that I should waste them. Again, I know that people way smarter than I designed this and I have the utmost respect for them. I'm just unable to get my head around this design choice after we've been forced to conserve for the last couple of decades. Either I'm missing something very big or I need to shown the way to the new IPv6 kool-aid bowl.

Thanks for your help.

Keith
  • 0

#33 notgoing2fail

notgoing2fail

    Cisco Routing/Switching Expert

  • Members
  • PipPipPipPip
  • 958 posts
  • Gender:Male
  • Location:New York
  • Interests:Life is short. Play with Cisco.

Posted 17 April 2010 - 08:11 AM

I'm just unable to get my head around this design choice after we've been forced to conserve for the last couple of decades. Either I'm missing something very big or I need to shown the way to the new IPv6 kool-aid bowl.



I know what you mean. I too am still coming to terms with IPv6. You see, I think with IPv6, you have to change your philosophy a bit on IP addressing. You have to really step back and look at the big picture which is, YES YOU CAN AFFORD TO WASTE IP ADDRESSES.

I've heard it here and there, the numbers may not be exact but you'll get the picture, there is basically 1.3 million unique IP addresses available per square inch on Earth.

One of the features of IPv6, automatically assigning the host with its own IP addresses basically takes advantage of the fact that there are so many IP addresses that it's actually taking the risk that it won't ever create an IP address that's the same as another host on the same network.

As far as the /64 goes, I think according to the RFC, you have to stick with /64, if you go any further, it won't comply with IPv6 and the router/hosts are suppose to ignore or drop those packets. Someone else can confirm in more detail exactly about that...

We are just so fine-tuned to use NATS/PATS, subnetting, ip subnet zero etc etc that it's hard to step away from that and drink the IPv6 kool-aid. But remember, these people who created IPv6 made the agreement that they don't EVER want society to go through this ever ever again. So apparently 128 bit provides more than anyone will ever need....

I know, hard to comprehend, but it's something that we just have to accept.....

I beleive that SUBNETTING, will be something of historical value one day down the road. Maybe not for us, but for the next generation kids who will say, "haha you actually had to do subnetting? We only read about that in class...."

I was too young for this, but back in the 70's they had calculators that you actually had to plug into the wall....so ....Haha for them! :P
  • 0

#34 DarkFiber

DarkFiber

    Cisco Expert

  • Members
  • PipPipPipPip
  • 927 posts
  • Gender:Male
  • Location:Cairo-Egypt
  • Interests:Networks

Posted 17 April 2010 - 07:45 PM

I know what you mean. I too am still coming to terms with IPv6. You see, I think with IPv6, you have to change your philosophy a bit on IP addressing. You have to really step back and look at the big picture which is, YES YOU CAN AFFORD TO WASTE IP ADDRESSES.

I've heard it here and there, the numbers may not be exact but you'll get the picture, there is basically 1.3 million unique IP addresses available per square inch on Earth.

One of the features of IPv6, automatically assigning the host with its own IP addresses basically takes advantage of the fact that there are so many IP addresses that it's actually taking the risk that it won't ever create an IP address that's the same as another host on the same network.

As far as the /64 goes, I think according to the RFC, you have to stick with /64, if you go any further, it won't comply with IPv6 and the router/hosts are suppose to ignore or drop those packets. Someone else can confirm in more detail exactly about that...

We are just so fine-tuned to use NATS/PATS, subnetting, ip subnet zero etc etc that it's hard to step away from that and drink the IPv6 kool-aid. But remember, these people who created IPv6 made the agreement that they don't EVER want society to go through this ever ever again. So apparently 128 bit provides more than anyone will ever need....

I know, hard to comprehend, but it's something that we just have to accept.....

I beleive that SUBNETTING, will be something of historical value one day down the road. Maybe not for us, but for the next generation kids who will say, "haha you actually had to do subnetting? We only read about that in class...."

I was too young for this, but back in the 70's they had calculators that you actually had to plug into the wall....so ....Haha for them! :P



Totally agree, couldn't say more.....remember when the network was using the ipv4 with ospf mulyipoint...hahahah,
Same as we say IPX


I'm in the middle of my honey moon, 5 days with no internet...couldn't resist more ;)
  • 0

#35 mongolio

mongolio

    Newbie

  • Members
  • Pip
  • 20 posts
  • Gender:Male
  • Location:Kyrgyzstan / Bishkek

Posted 27 August 2010 - 10:26 PM

I have done with reading IPv6 chapters in Odom ROUTE 642-902 and read Academy course chapters from BCSI aswell. I have question about subnetting concepts in IPv6 - as it mentioned in every related articles - IPv6 hasnt concept of Classfull addressing, but i have never see prefix-length other than /64 in any configuration example in any guide. Instead i found this note in BCSI guide:

An IPv6 address has two parts:
- A subnet prefix representing the network to which the interface is connected. The subnet prefix is a fixed 64-bit length for all current definitions.
- A local identifier, sometimes called a token, which uniquely identifies the host on the local network. The local identifier is always 64 bits and is dynamically created based on Layer 2 media and encapsulation. In the simple case of an Ethernet medium, the local identifier is usually derived from the EUI-48 MAC address.


Let omit part about methods of creation local identifier, but look at

The local identifier is always 64 bits

What does it mean? We have should not use VLSM with IPv6? Ok, maybe here is some mistake take place - let see in other sources. Wendell Odom have no any notes about subnetting in IPv6. But in Cisco Self-Study: Implementing Cisco IPv6 Networks (IPV6) i found "entire" paragraph being devoted to IPv6 subnetting

the number of bits for node addressing within a site prefix (48-bit) in IPv6 is so large that it is not necessary to make an addressing plan for a site using different network mask values. Therefore, network mask calculation for each subnet and the use of Variable-Length Subnet Masks (VLSMs) are not required. In IPv6, the subnetting allocation is much simpler than in IPv4.

On that moment i already have done checking how routers deal with prefix-length other than /64, and logic seems to be preserved from IPv4. So in practice we can borrow bits from host part, and from network part aswell. All that gief me partially answer on my questions, but i want to ask proffesional here for confirmation:
1)Its assumed that using of 64 bit host part is more convenient method of adressing and such methods of address assigment as stateless autoconfiguration, EUI-64, ISATAP tunnel adress assigment (modifyed EUI-64) was developed only from that perspective - not because /64 is only allowed prefix length for global unicast adresses.
2) If answer on my first question is "YES" than i have another one - So we can use prefix lengths other than 64 and can borrow bits from subnet prefix if needed (its hard to imagine when it can be necessary, but when i will need 2^64 adresses hard to imagine aswell ^^) and from host part - all same as for IPv4. But what with site prefixes? Is that mandatory that independent of number host adresses i needed i will receive /48 prefix for a total number of hosts 2^80 = 1208925819614629174706176?

In IPv6, the subnetting allocation is much simpler than in IPv4.

:unsure:
  • 0

#36 rpseenu

rpseenu

    Newbie

  • Members
  • Pip
  • 11 posts

Posted 29 August 2010 - 09:12 PM

Hi,

i want to know how to convert Mac to IP V6 (Link local address).. pls explin waiting from yours reply...

I have done with reading IPv6 chapters in Odom ROUTE 642-902 and read Academy course chapters from BCSI aswell. I have question about subnetting concepts in IPv6 - as it mentioned in every related articles - IPv6 hasnt concept of Classfull addressing, but i have never see prefix-length other than /64 in any configuration example in any guide. Instead i found this note in BCSI guide:

Let omit part about methods of creation local identifier, but look at What does it mean? We have should not use VLSM with IPv6? Ok, maybe here is some mistake take place - let see in other sources. Wendell Odom have no any notes about subnetting in IPv6. But in Cisco Self-Study: Implementing Cisco IPv6 Networks (IPV6) i found "entire" paragraph being devoted to IPv6 subnetting

On that moment i already have done checking how routers deal with prefix-length other than /64, and logic seems to be preserved from IPv4. So in practice we can borrow bits from host part, and from network part aswell. All that gief me partially answer on my questions, but i want to ask proffesional here for confirmation:
1)Its assumed that using of 64 bit host part is more convenient method of adressing and such methods of address assigment as stateless autoconfiguration, EUI-64, ISATAP tunnel adress assigment (modifyed EUI-64) was developed only from that perspective - not because /64 is only allowed prefix length for global unicast adresses.
2) If answer on my first question is "YES" than i have another one - So we can use prefix lengths other than 64 and can borrow bits from subnet prefix if needed (its hard to imagine when it can be necessary, but when i will need 2^64 adresses hard to imagine aswell ^^) and from host part - all same as for IPv4. But what with site prefixes? Is that mandatory that independent of number host adresses i needed i will receive /48 prefix for a total number of hosts 2^80 = 1208925819614629174706176?

:unsure:


  • 0

#37 James Lawrence

James Lawrence

    Newbie

  • Members
  • Pip
  • 22 posts
  • Gender:Male

Posted 01 October 2010 - 01:16 AM

A site-local address can be placed in the region, but not in the public world. But again, I'm confused with the above examples that came out of my mind. Please explain.
  • 0

#38 ZigAA

ZigAA

    Advanced Member

  • Members
  • PipPipPip
  • 179 posts
  • Gender:Male
  • Interests:lots...

Posted 09 November 2010 - 08:09 PM

Hello,

I'm working on IPv6 for ROUTE exam and I came to Ipv6 tunnels and here is a "problem":
interface Tunnel0
 no ip address
 ipv6 address 2002::1/48
 tunnel source Loopback0
 tunnel destination 3.3.3.3
source & destination are IPv4 IPs and they encapsulate IPv6 connectivity. It's all OK. I don't understand "tunnel mode ipv6ip" command:
interface Tunnel0
 no ip address
 ipv6 address 2002::1/48
 tunnel source Loopback0
 tunnel destination 3.3.3.3
 tunnel mode ipv6ip


Posted Image

This packet above is Layer3 type 47 - GRE and what is lower (layer4) is payload of GRE protocol

Posted Image

This packet is Layer3 type 41 - IPv6 (???). I don't understand this type. It's like Layer3 is IPv4 and Layer4 is Ipv6 ???
The reason of using tunnel mode ipv6ip command is to have smaller overhead, but as I see, the overhead is smaller for 4bytes! Is that all?
  • 0





0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users