3

IPv6 Prefix delegation is designed to allow an ISP to give an organisation a block of addresses to issue throughout their network. However, enterprise networks extend beyond a single tier of devices, and I cannot find a method of assigning this address block to the rest of the network beyond tertiary devices.

What method is meant to be used to allocate a /48 or /56 network in a network any larger than a home network? Is the intention that organisations still allocate addresses statically? I had thought that this was frowned upon, or is this thinking for home users only? I had read that home customers where advised for a /56 prefix while business customers were advised for a /48, and that seems to suggest that PD is also intended for business customers.

To explain my question better, the diagram below outlines the different layers, and where addresses can be allocated to. So far, I can allocate addresses to the primary interface connected to secondary devices, but not anything beyond. I have used command conventions outlined in this document.

What is the current best practice for distributing IPv6 addresses that have been PD assigned from an ISP?

       +-----------------------+
       |                       |
       |          ISP          |
       |                       |
       +-----------+-----------+
                   |
                   |
                   |ipv6 dhcp client pd mypool
       +-----------+-----------+
       |                       |
       |      First router     |
       |                       |
       +-----------+-----------+
                   |ipv6 address mypool ::1:0:0:0:1/64
                   |
                   |
                   |ipv6 address autoconfig default
      +------------+-----------+
      |                        |
      |      Second router     |
      |                        |
      +-----+--------------+---+
            |?             |?
            |              |
            |?             |?
+-----------+---------+ +--+-----------------+
|                     | |                    |
|     Tertiary 1      | |    Tertiary 2      |
|                     | |                    |
+---------------------+ +--------------------+
2
  • We need to know the specific models of your routers. For example, Cisco has methods to do IPv6 Prefix Delegation on all the routers in a site. – Ron Maupin Jan 12 at 16:30
  • Mostly Cisco with a few Junipers. 4451, 3650, 4500x, etc – fileinster Jan 12 at 16:33
3

Prefix delegation through a site can be performed via DHCP relay chaining; each router in the chain bootstraps the next. The PD should be originated internally to allow you to scope the subnets accordingly. My only concern is how to build resilience in to this. Also, my provider has only given a /56 which makes this method not very scalable with that amount of address space, hence why the address authorities advise for a /48 for business connections.

Further device tiers can be configured the same as the tertiary devices; the important consideration is to identify the direction to the DHCP server, but I'm not sure how to make this a resilient/redundant configuration in the event of multiple paths. The same configuration allows for 8 ipv6 interfaces per router and 16 routers; not scalable by any means. There is some unused address space that can be allocated statically:

**First Router**
ipv6 local pool mypool x:x:x:80::/57 61
ipv6 dhcp pool dhcpv6pool
 prefix-delegation pool mypool lifetime 1800 600
!
ipv6 general-prefix INTERNET_PREFIX x::/61
interface in
 ipv6 address <manual>/64
interface out
 ipv6 address INTERNET_PREFIX ::1:0:0:0:1/64
 ipv6 dhcp server dhcpv6pool

**Second Router**
interface in
 ipv6 address autoconfig default
 ipv6 dhcp client pd mypool
!
interface out
 ipv6 address mypool::1/64
 ipv6 dhcp relay destination x::x

**Tier 3 Router**
interface in
 ipv6 address autoconfig default
 ipv6 dhcp client pd mypool
!
interface out
 ipv6 address mypool::1/64

A summary diagram to explain:

   +-----------------------+
   |                       |
   |          ISP          |
   |                       |
   +-----------+-----------+
               |
               |ipv6 address <manual>
   +-----------------------+
   |                       |
   |      First router     |
   |                       |
   +-----------+-----------+
               |ipv6 address mypool ::1:0:0:0:1/64
               |ipv6 dhcp server dhcpv6
               |
               |ipv6 address autoconfig default
               |ipv6 dhcp client pd mypool
   +-----------+-----------+
   |                       |
   |     Second router     |
   |                       |
   +-----------+-----------+
               |ipv6 address mypool ::1:0:0:0:1/64
               |ipv6 dhcp relay destination x::x
               |
               |ipv6 address autoconfig default
               |ipv6 dhcp client pd mypool
  +------------+-----------+
  |                        |
  |      Tier 3 router     |
  |                        |
  +------------+-----------+
               |ipv6 address mypool ::1/64
5
  • In reality, a complex network will not use IPv6 PD, and it will use the subnet values to encode some information specific to the networks. If an ISP does not give each site a /48 is it really not understanding IPv6, yet. IPv6 is designed to actually waste addresses and network addresses. There is no reason to be so frugal with business networks. Your business could actually get a provider-independent, smaller prefix from your RIR, and you would not be tied to any one ISP, just run BGP with any ISP you have or get. – Ron Maupin Jan 14 at 1:02
  • 1
  • Thanks, Ron. This was more of an academic endeavour than a real world IPv6 rollout on a site specifically to test address assignment. I now have an entire site using DHCPv6 PD to assign addresses with full Internet access. I'll also be more wary with providers when ordering my next synchronous 1GbE Internet line next time to ensure they allocate a more appropriate prefix. – fileinster Jan 14 at 10:04
  • 1
    Also, you should consider that in the IPv6 Global Conference that I attended, one of the big statistics that kept coming up is that, on average, a company implementing IPv6 will readdress three times before settling on its IPv6 addressing. That is because you can use the subnet nibbles to encode information, such as Region, Site, Function, etc., and most companies start out ignoring those possibilities, just continuing the IPv4 way of doing things, but that is a very bad idea because IPv6 does things differently. – Ron Maupin Jan 14 at 15:49
  • Yes, I am aware that it is officially advised now. It's a natural evolution from good addressing strategies developed in IPv4 that required good housekeeping and active management and planning. It just becomes easier to plan a this strategy when you have more address space at your disposal. – fileinster Jan 14 at 22:42
-1

The answer to your question is routing protocols and network architecture. We don't automatically assign addresses across multiple network tiers.

IPv6 PD is a great system for dynamically assigning addresses to end-user networks such as homes and branch offices; but it was never intended to scale up further, the way your diagram describes.

1
  • 2
    Actually, it does scale out just fine... if your routers know how to do it. (PD of a prefix learned by DHCP is not a common feature.) – Ricky Jan 13 at 4:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.