Network Working Group                                    G. Van de Velde
Request for Comments: 5375                                  C. Popoviciu
Category: Informational                                    Cisco Systems
                                                               T. Chown
                                              University of Southampton
                                                             O. Bonness
                                                                C. Hahn
                                     T-Systems Enterprise Services GmbH
                                                          December 2008


            IPv6 Unicast Address Assignment Considerations

Status of This Memo

  This memo provides information for the Internet community.  It does
  not specify an Internet standard of any kind.  Distribution of this
  memo is unlimited.

Copyright Notice

  Copyright (c) 2008 IETF Trust and the persons identified as the
  document authors.  All rights reserved.

  This document is subject to BCP 78 and the IETF Trust's Legal
  Provisions Relating to IETF Documents (http://trustee.ietf.org/
  license-info) in effect on the date of publication of this document.
  Please review these documents carefully, as they describe your rights
  and restrictions with respect to this document.

Abstract

  One fundamental aspect of any IP communications infrastructure is its
  addressing plan.  With its new address architecture and allocation
  policies, the introduction of IPv6 into a network means that network
  designers and operators need to reconsider their existing approaches
  to network addressing.  Lack of guidelines on handling this aspect of
  network design could slow down the deployment and integration of
  IPv6.  This document aims to provide the information and
  recommendations relevant to planning the addressing aspects of IPv6
  deployments.  The document also provides IPv6 addressing case studies
  for both an enterprise and an ISP network.









Van de Velde, et al.         Informational                      [Page 1]

RFC 5375             IPv6 Addressing Considerations        December 2008


Table of Contents

  1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  3
  2.  Network-Level Addressing Design Considerations . . . . . . . .  4
    2.1.  Globally Unique Addresses  . . . . . . . . . . . . . . . .  4
    2.2.  Unique Local IPv6 Addresses  . . . . . . . . . . . . . . .  5
    2.3.  6bone Address Space  . . . . . . . . . . . . . . . . . . .  6
    2.4.  Network-Level Design Considerations  . . . . . . . . . . .  6
      2.4.1.  Sizing the Network Allocation  . . . . . . . . . . . .  8
      2.4.2.  Address Space Conservation . . . . . . . . . . . . . .  8
  3.  Subnet Prefix Considerations . . . . . . . . . . . . . . . . .  8
    3.1.  Considerations for /64 Prefixes  . . . . . . . . . . . . . 10
  4.  Allocation of the IID of an IPv6 Address . . . . . . . . . . . 10
    4.1.  Automatic EUI-64 Format Option . . . . . . . . . . . . . . 10
    4.2.  Using Privacy Extensions . . . . . . . . . . . . . . . . . 10
    4.3.  Manual/Dynamic Assignment Option . . . . . . . . . . . . . 11
  5.  Security Considerations  . . . . . . . . . . . . . . . . . . . 11
  6.  Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 11
  7.  Informative References . . . . . . . . . . . . . . . . . . . . 12
  Appendix A.  Case Studies  . . . . . . . . . . . . . . . . . . . . 16
    A.1.  Enterprise Considerations  . . . . . . . . . . . . . . . . 16
      A.1.1.  Obtaining General IPv6 Network Prefixes  . . . . . . . 16
      A.1.2.  Forming an Address (Subnet) Allocation Plan  . . . . . 17
      A.1.3.  Other Considerations . . . . . . . . . . . . . . . . . 18
      A.1.4.  Node Configuration Considerations  . . . . . . . . . . 18
    A.2.  Service Provider Considerations  . . . . . . . . . . . . . 19
      A.2.1.  Investigation of Objective Requirements for an
              IPv6 Addressing Schema of a Service Provider . . . . . 19
      A.2.2.  Exemplary IPv6 Address Allocation Plan for a
              Service Provider . . . . . . . . . . . . . . . . . . . 23
      A.2.3.  Additional Remarks . . . . . . . . . . . . . . . . . . 28
  Appendix B.  Considerations for Subnet Prefixes Different than
               /64 . . . . . . . . . . . . . . . . . . . . . . . . . 30
    B.1.  Considerations for Subnet Prefixes Shorter than /64  . . . 30
    B.2.  Considerations for Subnet Prefixes Longer than /64 . . . . 31
      B.2.1.  /126 Addresses . . . . . . . . . . . . . . . . . . . . 31
      B.2.2.  /127 Addresses . . . . . . . . . . . . . . . . . . . . 31
      B.2.3.  /128 Addresses . . . . . . . . . . . . . . . . . . . . 31
      B.2.4.  EUI-64 'u' and 'g' Bits  . . . . . . . . . . . . . . . 31
      B.2.5.  Anycast Addresses  . . . . . . . . . . . . . . . . . . 32
      B.2.6.  Addresses Used by Embedded-RP (RFC 3956) . . . . . . . 33
      B.2.7.  ISATAP Addresses . . . . . . . . . . . . . . . . . . . 34









Van de Velde, et al.         Informational                      [Page 2]

RFC 5375             IPv6 Addressing Considerations        December 2008


1.  Introduction

  The Internet Protocol Version 6 (IPv6) Addressing Architecture
  [RFC4291] defines three main types of addresses: unicast, anycast,
  and multicast.  This document focuses on unicast addresses, for which
  there are currently two principal allocated types: Globally Unique
  Addresses ('globals') [RFC3587] and Unique Local IPv6 Addresses
  (ULAs) [RFC4193].  In addition, until recently there has been the
  'experimental' 6bone address space [RFC3701], though its use has been
  deprecated since June 2006 [RFC3701].

  The document covers aspects that should be considered during IPv6
  deployment for the design and planning of an addressing scheme for an
  IPv6 network.  The network's IPv6 addressing plan may be for an IPv6-
  only network, or for a dual-stack infrastructure where some or all
  devices have addresses in both protocols.  These considerations will
  help an IPv6 network designer to efficiently and prudently assign the
  IPv6 address space that has been allocated to their organization.

  The address assignment considerations are analyzed separately for the
  two major components of the IPv6 unicast addresses -- namely,
  'Network-Level Addressing' (the allocation of subnets) and the
  'interface-id' (the identification of the interface within a subnet).
  Thus, the document includes a discussion of aspects of address
  assignment to nodes and interfaces in an IPv6 network.  Finally, the
  document provides two examples of deployed addressing plans in a
  service provider (ISP) and an enterprise network.

  Parts of this document highlight the differences that an experienced
  IPv4 network designer should consider when planning an IPv6
  deployment, for example:

  o  IPv6 devices will more likely be multi-addressed in comparison
     with their IPv4 counterparts.

  o  The practically unlimited size of an IPv6 subnet (2^64 bits)
     reduces the requirement to size subnets to device counts for the
     purposes of (IPv4) address conservation.

  o  The vastly increased subnet size has implications on the threat of
     address-based host scanning and other scanning techniques, as
     discussed in [RFC5157].

  We do not discuss here how a site or ISP should proceed with
  acquiring its globally routable IPv6 address prefix.  In each case,
  the prefix received is either provider assigned (PA) or provider
  independent (PI).




Van de Velde, et al.         Informational                      [Page 3]

RFC 5375             IPv6 Addressing Considerations        December 2008


  We do not discuss PI policy here.  The observations and
  recommendations of this text are largely independent of the PA or PI
  nature of the address block being used.  At this time, we assume that
  when an IPv6 network changes provider, typically it will need to
  undergo a renumbering process, as described in [RFC4192].  A separate
  document [THINKABOUT] makes recommendations to ease the IPv6
  renumbering process.

  This document does not discuss implementation aspects related to the
  transition from the now obsoleted site-local addresses to ULAs.  Some
  implementations know about site-local addresses even though they are
  deprecated, and do not know about ULAs even though they represent
  current specification.  As a result, transitioning between these
  types of addresses may cause difficulties.

2.  Network-Level Addressing Design Considerations

  This section discusses the kind of IPv6 addresses used at the network
  level for the IPv6 infrastructure.  The kind of addresses that can be
  considered are Globally Unique Addresses and ULAs.  We also comment
  here on the deprecated 6bone address space.

2.1.  Globally Unique Addresses

  The most commonly used unicast addresses will be Globally Unique
  Addresses ('globals').  No significant considerations are necessary
  if the organization has an address space assignment and a single
  prefix is deployed through a single upstream provider.

  However, a multihomed site may deploy addresses from two or more
  service-provider-assigned IPv6 address ranges.  Here, the network
  administrator must have awareness on where and how these ranges are
  used on the multihomed infrastructure environment.  The nature of the
  usage of multiple prefixes may depend on the reason for multihoming
  (e.g., resilience failover, load balancing, policy-based routing, or
  multihoming during an IPv6 renumbering event).  IPv6 introduces
  improved support for multi-addressed hosts through the IPv6 default
  address selection methods described in RFC 3484 [RFC3484].  A
  multihomed host may thus have two or more addresses, one per prefix
  (provider), and select source and destination addresses to use as
  described in that RFC.  However, multihoming also has some
  operational and administrative burdens besides choosing multiple
  addresses per interface [RFC4218] [RFC4219].








Van de Velde, et al.         Informational                      [Page 4]

RFC 5375             IPv6 Addressing Considerations        December 2008


2.2.  Unique Local IPv6 Addresses

  ULAs have replaced the originally conceived site-local addresses in
  the IPv6 addressing architecture, for reasons described in [RFC3879].
  ULAs improve on site-locals by offering a high probability of the
  global uniqueness of the prefix used, which can be beneficial when
  there is (deliberate or accidental) leakage or when networks are
  merged.  ULAs are akin to the private address space [RFC1918]
  assigned for IPv4 networks, except that in IPv6 networks we may
  expect to see ULAs used alongside global addresses, with ULAs used
  internally and globals used externally.  Thus, use of ULAs does not
  imply use of NAT for IPv6.

  The ULA address range allows network administrators to deploy IPv6
  addresses on their network without asking for a globally unique
  registered IPv6 address range.  A ULA prefix is 48 bits, i.e., a /48,
  the same as the currently recommended allocation for a site from the
  globally routable IPv6 address space [RFC3177].

  A site that wishes to use ULAs can have (a) multiple /48 prefixes
  (e.g., a /44) (b) one /48, or (c) a less-than-/48 prefix (e.g., a /56
  or /64).  In all of the above cases, the ULAs can be randomly chosen
  according to the principles specified in [RFC4193].  However, in case
  (a) the use of randomly chosen ULAs will provide suboptimal
  aggregation capabilities.

  ULAs provide the means to deploy a fixed addressing scheme that is
  not affected by a change in service provider and the corresponding PA
  global addresses.  Internal operation of the network is thus
  unaffected during renumbering events.  Nevertheless, this type of
  address must be used with caution.

  A site using ULAs may or may not also deploy global addresses.  In an
  isolated network, ULAs may be deployed on their own.  In a connected
  network that also deploys global addresses, both may be deployed,
  such that hosts become multi-addressed (one global and one ULA), and
  the IPv6 default address selection algorithm will pick the
  appropriate source and destination addresses to use, e.g., ULAs will
  be selected where both the source and destination hosts have ULAs.
  Because a ULA and a global site prefix are both /48 length, an
  administrator can choose to use the same subnetting (and host
  addressing) plan for both prefixes.

  As an example of the problems ULAs may cause, when using IPv6
  multicast within the network, the IPv6 default address selection
  algorithm prefers the ULA as the source address for the IPv6
  multicast streams.  This is NOT a valid option when sending an IPv6
  multicast stream to the IPv6 Internet for two reasons.  For one,



Van de Velde, et al.         Informational                      [Page 5]

RFC 5375             IPv6 Addressing Considerations        December 2008


  these addresses are not globally routable, so Reverse Path Forwarding
  checks for such traffic will fail outside the internal network.  The
  other reason is that the traffic will likely not cross the network
  boundary due to multicast domain control and perimeter security
  policies.

  In principle, ULAs allow easier network mergers than RFC 1918
  addresses do for IPv4 because ULA prefixes have a high probability of
  uniqueness, if the prefix is chosen as described in the RFC.

2.3.  6bone Address Space

  The 6bone address space was used before the Regional Internet
  Registries (RIRs) started to distribute 'production' IPv6 prefixes.
  The 6bone prefixes have a common first 16 bits in the IPv6 Prefix of
  3FFE::/16.  This address range has been deprecated as of 6 June 2006
  [RFC3701] and must not be used on any new IPv6 network deployments.
  Sites using 6bone address space should renumber to production address
  space using procedures as defined in [RFC4192].

2.4.  Network-Level Design Considerations

  IPv6 provides network administrators with a significantly larger
  address space, enabling them to be very creative in how they can
  define logical and practical addressing plans.  The subnetting of
  assigned prefixes can be done based on various logical schemes that
  involve factors such as:

  o  Using existing systems

     *  translate the existing subnet numbers into IPv6 subnet IDs

     *  translate the VLAN IDs into IPv6 subnet IDs

  o  Redesign

     *  allocate according to your need

  o  Aggregation

     *  Geographical Boundaries - by assigning a common prefix to all
        subnets within a geographical area.

     *  Organizational Boundaries - by assigning a common prefix to an
        entire organization or group within a corporate infrastructure.






Van de Velde, et al.         Informational                      [Page 6]

RFC 5375             IPv6 Addressing Considerations        December 2008


     *  Service Type - by reserving certain prefixes for predefined
        services such as: VoIP, content distribution, wireless
        services, Internet access, security areas, etc.  This type of
        addressing may create dependencies on IP addresses that can
        make renumbering harder if the nodes or interfaces supporting
        those services on the network are sparse within the topology.

  Such logical addressing plans have the potential to simplify network
  operations and service offerings, and to simplify network management
  and troubleshooting.  A very large network would not need to consider
  using private address space for its infrastructure devices, thereby
  simplifying network management.

  The network designer must however keep in mind several factors when
  developing these new addressing schemes for networks with and without
  global connectivity:

  o  Prefix aggregation - The larger IPv6 addresses can lead to larger
     routing tables unless network designers are actively pursuing
     aggregation.  While prefix aggregation will be enforced by the
     service provider, it is beneficial for the individual
     organizations to observe the same principles in their network
     design process.

  o  Network growth - The allocation mechanism for flexible growth of a
     network prefix, documented in RFC 3531 [RFC3531] can be used to
     allow the network infrastructure to grow and be numbered in a way
     that is likely to preserve aggregation (the plan leaves 'holes'
     for growth).

  o  ULA usage in large networks - Networks that have a large number of
     'sites' that each deploy a ULA prefix that will by default be a
     'random' /48 under fc00::/7 will have no aggregation of those
     prefixes.  Thus, the end result may be cumbersome because the
     network will have large amounts of non-aggregated ULA prefixes.
     However, there is no rule to disallow large networks from using a
     single ULA prefix for all 'sites', as a ULA still provides 16 bits
     for subnetting to be used internally.

  o  Compact numbering of small sites - It is possible that as registry
     policies evolve, a small site may experience an increase in prefix
     length when renumbering, e.g., from /48 to /56.  For this reason,
     the best practice is to number subnets compactly rather than
     sparsely, and to use low-order bits as much as possible when
     numbering subnets.  In other words, even if a /48 is allocated,
     act as though only a /56 is available.  Clearly, this advice does
     not apply to large sites and enterprises that have an intrinsic
     need for a /48 prefix.



Van de Velde, et al.         Informational                      [Page 7]

RFC 5375             IPv6 Addressing Considerations        December 2008


  o  Consider assigning more than one /64 to a site - A small site may
     want to enable routing amongst interfaces connected to a gateway
     device.  For example, a residential gateway that receives a /48
     and is situated in a home with multiple LANs of different media
     types (sensor network, wired, Wi-Fi, etc.), or has a need for
     traffic segmentation (home, work, kids, etc.), could benefit
     greatly from multiple subnets and routing in IPv6.  Ideally,
     residential networks would be given an address range of a /48 or
     /56 [RIPE_Nov07] such that multiple /64 subnets could be used
     within the residence.

2.4.1.  Sizing the Network Allocation

  We do not discuss here how a network designer sizes their application
  for address space.  By default, a site will receive a /48 prefix
  [RFC3177]; however, different RIR service regions policies may
  suggest alternative default assignments or let the ISPs decide on
  what they believe is more appropriate for their specific case (see
  Section 6.5.4, "Assignments from LIRs/ISPs", of [ARIN]).  The default
  provider allocation via the RIRs is currently a /32 [RIPE_Nov07].
  These allocations are indicators for a first allocation for a
  network.  Different sizes may be obtained based on the anticipated
  address usage [RIPE_Nov07].  At the time of writing, there are
  examples of allocations as large as /19 having been made from RIRs to
  providers.

2.4.2.  Address Space Conservation

  Despite the large IPv6 address space, which enables easier
  subnetting, it still is important to ensure an efficient use of this
  resource.  Some addressing schemes, while facilitating aggregation
  and management, could lead to significant numbers of addresses being
  unused.  Address conservation requirements are less stringent in
  IPv6, but they should still be observed.

  The proposed Host-Density (HD) value [RFC3194] for IPv6 is 0.94
  compared to the current value of 0.96 for IPv4.  Note that with IPv6,
  HD is calculated for sites (e.g., on a basis of /56), instead of for
  addresses as with IPv4.

3.  Subnet Prefix Considerations

  An important part of an IPv4 addressing plan is deciding the length
  of each subnet prefix.  Unlike in IPv4, the IPv6 addressing
  architecture [RFC4291] specifies that all subnets using Globally
  Unique Addresses and ULAs always have the same prefix length of 64
  bits.  (This also applies to the deprecated 6bone and site-local
  addresses.)



Van de Velde, et al.         Informational                      [Page 8]

RFC 5375             IPv6 Addressing Considerations        December 2008


  The only exception to this rule are special addresses starting with
  the binary value 000, such as IPv4-compatible IPv6 addresses.  These
  exceptions are largely beyond the scope of this document.

  Using a subnet prefix length other than a /64 will break many
  features of IPv6, including Neighbor Discovery (ND), Secure Neighbor
  Discovery (SEND) [RFC3971], privacy extensions [RFC4941], parts of
  Mobile IPv6 [RFC4866], Protocol Independent Multicast - Sparse Mode
  (PIM-SM) with Embedded-RP [RFC3956], and Site Multihoming by IPv6
  Intermediation (SHIM6) [SHIM6], among others.  A number of other
  features currently in development, or being proposed, also rely on
  /64 subnet prefixes.

  Nevertheless, many IPv6 implementations do not prevent the
  administrator from configuring a subnet prefix length shorter or
  longer than 64 bits.  Using subnet prefixes shorter than /64 would
  rarely be useful; see Appendix B.1 for discussion.

  However, some network administrators have used prefixes longer than
  /64 for links connecting routers, usually just two routers on a
  point-to-point link.  On links where all the addresses are assigned
  by manual configuration, and all nodes on the link are routers (not
  end hosts) that are known by the network, administrators do not need
  any of the IPv6 features that rely on /64 subnet prefixes, this can
  work.  Using subnet prefixes longer than /64 is not recommended for
  general use, and using them for links containing end hosts would be
  an especially bad idea, as it is difficult to predict what IPv6
  features the hosts will use in the future.

  Appendix B.2 describes some practical considerations that need to be
  taken into account when using prefixes longer than /64 in limited
  cases.  In particular, a number of IPv6 features use interface
  identifiers that have a special form (such as a certain fixed value
  in some bit positions).  When using prefixes longer than /64, it is
  prudent to avoid certain subnet prefix values so that nodes who
  assume that the prefix is /64 will not incorrectly identify the
  addresses in that subnet as having a special form.  Appendix B.2
  describes the subnet prefix values that are currently believed to be
  potentially problematic; however, the list is not exhaustive and can
  be expected to grow in the future.

  Using /64 subnets is strongly recommended, also for links connecting
  only routers.  A deployment compliant with the current IPv6
  specifications cannot use other prefix lengths.  However, the V6OPS
  WG believes that despite the drawbacks (and a potentially expensive
  network redesign, if IPv6 features relying on /64 subnets are needed
  in the future), some networks administrators will use prefixes longer
  than /64.



Van de Velde, et al.         Informational                      [Page 9]

RFC 5375             IPv6 Addressing Considerations        December 2008


3.1.  Considerations for /64 Prefixes

  Based on RFC 3177 [RFC3177], 64 bits is the prescribed subnet prefix
  length to allocate to interfaces and nodes.

  When using a /64 subnet length, the address assignment for these
  addresses can be made either by manual configuration, by a Dynamic
  Host Configuration Protocol [RFC3315], by stateless autoconfiguration
  [RFC4862], or by a combination thereof [RFC3736].

  Note that RFC 3177 strongly prescribes 64-bit subnets for general
  usage, and that stateless autoconfiguration on most link layers
  (including Ethernet) is only defined for 64-bit subnets.  While in
  theory it might be possible that some future autoconfiguration
  mechanisms would allow longer than 64-bit prefix lengths to be used,
  the use of such prefixes is not recommended at this time.

4.  Allocation of the IID of an IPv6 Address

  In order to have a complete IPv6 address, an interface must be
  associated with a prefix and an Interface Identifier (IID).  Section
  3 of this document analyzed the prefix selection considerations.
  This section discusses the elements that should be considered when
  assigning the IID portion of the IPv6 address.

  There are various ways to allocate an IPv6 address to a device or
  interface.  The option with the least amount of caveats for the
  network administrator is that of EUI-64 [RFC4862] based addresses.
  For the manual or dynamic options, the overlap with well-known IPv6
  addresses should be avoided.

4.1.  Automatic EUI-64 Format Option

  When using this method, the network administrator has to allocate a
  valid 64-bit subnet prefix.  Once that allocation has been made, the
  EUI-64 [RFC4862] allocation procedure can assign the remaining 64 IID
  bits in a stateless manner.  All the considerations for selecting a
  valid IID have been incorporated into the EUI-64 methodology.

4.2.  Using Privacy Extensions

  The main purpose of IIDs generated based on RFC 4941 [RFC4941] is to
  provide privacy to the entity using an IPv6 address.  While there are
  no particular constraints in the usage of IPv6 addresses with IIDs as
  defined in [RFC4941], there are some implications to be aware of when
  using privacy addresses as documented in Section 4 of RFC 4941
  [RFC4941]




Van de Velde, et al.         Informational                     [Page 10]

RFC 5375             IPv6 Addressing Considerations        December 2008


4.3.  Manual/Dynamic Assignment Option

  This section discusses those IID allocations that are not implemented
  through stateless address configuration (Section 4.1).  They are
  applicable regardless of the prefix length used on the link.  It is
  out of scope for this section to discuss the various assignment
  methods (e.g., manual configuration, DHCPv6, etc).

  In this situation, the actual allocation is done by human
  intervention, and consideration needs to be given to the complete
  IPv6 address so that it does not result in overlaps with any of the
  well-known IPv6 addresses:

  o  Subnet Router Anycast Address (Appendix B.2.5.1)

  o  Reserved Subnet Anycast Address (Appendix B.2.5.2)

  o  Addresses used by Embedded-RP (Appendix B.2.6)

  o  Intra-Site Automatic Tunnel Addressing Protocol (ISATAP) Addresses
     (Appendix B.2.7)

  When using an address assigned by human intervention, it is
  recommended to choose IPv6 addresses that are not obvious to guess
  and/or to avoid any IPv6 addresses that embed IPv4 addresses used in
  the current infrastructure.  Following these two recommendations will
  make it more difficult for malicious third parties to guess targets
  for attack, and thus reduce security threats to a certain extent.

5.  Security Considerations

  This document doesn't add any new security considerations that aren't
  already outlined in the security considerations of the references.

  It must be noted that using subnet prefixes other than /64 breaks
  security mechanisms such as Cryptographically Generated Addresses
  (CGAs) and Hash-Based Addresses (HBAs), and thus makes it impossible
  to use protocols that depend on them.

6.  Acknowledgements

  Constructive feedback and contributions have been received during
  IESG review cycle and from Marla Azinger, Stig Venaas, Pekka Savola,
  John Spence, Patrick Grossetete, Carlos Garcia Braschi, Brian
  Carpenter, Mark Smith, Janos Mohacsi, Jim Bound, Fred Templin, Ginny
  Listman, Salman Assadullah, Krishnan Thirukonda, and the IESG.





Van de Velde, et al.         Informational                     [Page 11]

RFC 5375             IPv6 Addressing Considerations        December 2008


7.  Informative References

  [RFC1918]       Rekhter, Y., Moskowitz, R., Karrenberg, D., Groot,
                  G., and E. Lear, "Address Allocation for Private
                  Internets", BCP 5, RFC 1918, February 1996.

  [RFC2526]       Johnson, D. and S. Deering, "Reserved IPv6 Subnet
                  Anycast Addresses", RFC 2526, March 1999.

  [RFC3021]       Retana, A., White, R., Fuller, V., and D. McPherson,
                  "Using 31-Bit Prefixes on IPv4 Point-to-Point Links",
                  RFC 3021, December 2000.

  [RFC3053]       Durand, A., Fasano, P., Guardini, I., and D. Lento,
                  "IPv6 Tunnel Broker", RFC 3053, January 2001.

  [RFC3056]       Carpenter, B. and K. Moore, "Connection of IPv6
                  Domains via IPv4 Clouds", RFC 3056, February 2001.

  [RFC3177]       IAB and IESG, "IAB/IESG Recommendations on IPv6
                  Address Allocations to Sites", RFC 3177,
                  September 2001.

  [RFC3180]       Meyer, D. and P. Lothberg, "GLOP Addressing in
                  233/8", BCP 53, RFC 3180, 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.

  [RFC3315]       Droms, R., Bound, J., Volz, B., Lemon, T., Perkins,
                  C., and M. Carney, "Dynamic Host Configuration
                  Protocol for IPv6 (DHCPv6)", RFC 3315, July 2003.

  [RFC3484]       Draves, R., "Default Address Selection for Internet
                  Protocol version 6 (IPv6)", RFC 3484, February 2003.

  [RFC3531]       Blanchet, M., "A Flexible Method for Managing the
                  Assignment of Bits of an IPv6 Address Block",
                  RFC 3531, April 2003.

  [RFC3587]       Hinden, R., Deering, S., and E. Nordmark, "IPv6
                  Global Unicast Address Format", RFC 3587,
                  August 2003.

  [RFC3627]       Savola, P., "Use of /127 Prefix Length Between
                  Routers Considered Harmful", RFC 3627,
                  September 2003.



Van de Velde, et al.         Informational                     [Page 12]

RFC 5375             IPv6 Addressing Considerations        December 2008


  [RFC3633]       Troan, O. and R. Droms, "IPv6 Prefix Options for
                  Dynamic Host Configuration Protocol (DHCP) version
                  6", RFC 3633, December 2003.

  [RFC3701]       Fink, R. and R. Hinden, "6bone (IPv6 Testing Address
                  Allocation) Phaseout", RFC 3701, March 2004.

  [RFC3736]       Droms, R., "Stateless Dynamic Host Configuration
                  Protocol (DHCP) Service for IPv6", RFC 3736,
                  April 2004.

  [RFC3879]       Huitema, C. and B. Carpenter, "Deprecating Site Local
                  Addresses", RFC 3879, September 2004.

  [RFC3956]       Savola, P. and B. Haberman, "Embedding the Rendezvous
                  Point (RP) Address in an IPv6 Multicast Address",
                  RFC 3956, November 2004.

  [RFC3971]       Arkko, J., Kempf, J., Zill, B., and P. Nikander,
                  "SEcure Neighbor Discovery (SEND)", RFC 3971,
                  March 2005.

  [RFC4192]       Baker, F., Lear, E., and R. Droms, "Procedures for
                  Renumbering an IPv6 Network without a Flag Day",
                  RFC 4192, September 2005.

  [RFC4193]       Hinden, R. and B. Haberman, "Unique Local IPv6
                  Unicast Addresses", RFC 4193, October 2005.

  [RFC4218]       Nordmark, E. and T. Li, "Threats Relating to IPv6
                  Multihoming Solutions", RFC 4218, October 2005.

  [RFC4219]       Lear, E., "Things Multihoming in IPv6 (MULTI6)
                  Developers Should Think About", RFC 4219,
                  October 2005.

  [RFC4271]       Rekhter, Y., Li, T., and S. Hares, "A Border Gateway
                  Protocol 4 (BGP-4)", RFC 4271, January 2006.

  [RFC4291]       Hinden, R. and S. Deering, "IP Version 6 Addressing
                  Architecture", RFC 4291, February 2006.

  [RFC4477]       Chown, T., Venaas, S., and C. Strauf, "Dynamic Host
                  Configuration Protocol (DHCP): IPv4 and IPv6 Dual-
                  Stack Issues", RFC 4477, May 2006.






Van de Velde, et al.         Informational                     [Page 13]

RFC 5375             IPv6 Addressing Considerations        December 2008


  [RFC4798]       De Clercq, J., Ooms, D., Prevost, S., and F. Le
                  Faucheur, "Connecting IPv6 Islands over IPv4 MPLS
                  Using IPv6 Provider Edge Routers (6PE)", RFC 4798,
                  February 2007.

  [RFC4862]       Thomson, S., Narten, T., and T. Jinmei, "IPv6
                  Stateless Address Autoconfiguration", RFC 4862,
                  September 2007.

  [RFC4866]       Arkko, J., Vogt, C., and W. Haddad, "Enhanced Route
                  Optimization for Mobile IPv6", RFC 4866, May 2007.

  [RFC4941]       Narten, T., Draves, R., and S. Krishnan, "Privacy
                  Extensions for Stateless Address Autoconfiguration in
                  IPv6", RFC 4941, September 2007.

  [RFC5214]       Templin, F., Gleeson, T., and D. Thaler, "Intra-Site
                  Automatic Tunnel Addressing Protocol (ISATAP)",
                  RFC 5214, March 2008.

  [RFC5157]       Chown, T., "IPv6 Implications for Network Scanning",
                  RFC 5157, March 2008.

  [SHIM6]         IETF, "Site Multihoming by IPv6 Intermediation
                  (shim6) Charter", <http://www.ietf.org/html.charters/
                  shim6-charter.html>.

  [ARIN]          ARIN, "ARIN Number Resource Policy Manual",
                  Version 2008.4, September 2008,
                  <http://www.arin.net/policy/nrpm.html>.

  [RIPE_Nov07]    APNIC, ARIN, RIPE NCC, "IPv6 Address Allocation and
                  Assignment Policy", ripe-421, November 2007,
                  <http://www.ripe.net/ripe/docs/ipv6policy.html>.

  [RIPE_Jul07]    APNIC, ARIN, RIPE NCC, "IPv6 Address Allocation and
                  Assignment Policy", ripe-412, July 2007,
                  <http://www.ripe.net/ripe/docs/ripe-412.html>.

  [APNIC_IPv6]    APNIC, "IPv6 Address Allocation and Assignment
                  Policy", APNIC-089, August 2008, <http://
                  www.apnic.net/policy/ipv6-address-policy.html>.

  [LACNIC_IPv6]   LACNIC, "Internet Resource Management Policies in
                  Latin America and the Caribbean: IPv6 Address
                  Allocation and Assignment Policy",
                  <http://lacnic.net/en/politicas/ipv6.html>.




Van de Velde, et al.         Informational                     [Page 14]

RFC 5375             IPv6 Addressing Considerations        December 2008


  [AFRINIC_IPv6]  AfriNIC, "AfriNIC IPv6 Address Allocation and
                  Assignment Policy", March 2004,
                  <http://www.afrinic.net/docs/policies/
                  afpol-v6200407-000.htm>.

  [THINKABOUT]    Chown, T., Thompson, M., Ford, A., and S. Venaas,
                  "Things to think about when Renumbering an IPv6
                  network", Work in Progress, March 2007.











































Van de Velde, et al.         Informational                     [Page 15]

RFC 5375             IPv6 Addressing Considerations        December 2008


Appendix A.  Case Studies

  This appendix contains two case studies for IPv6 addressing schemas
  that have been based on the statements and considerations of this
  document.  These case studies illustrate how this document has been
  used in two specific network scenarios.  The case studies may serve
  as basic considerations for an administrator who designs the IPv6
  addressing schema for an enterprise or ISP network, but are not
  intended to serve as a general design proposal for every kind of IPv6
  network.  All subnet sizes used in this appendix are for practical
  visualization and do not dictate RIR policy.

A.1.  Enterprise Considerations

  In this section, one considers a case study of a campus network that
  is deploying IPv6 in parallel with existing IPv4 protocols in a dual-
  stack environment.  The specific example is the University of
  Southampton (UK), focusing on a large department within that network.
  The deployment currently spans around 1,000 hosts and over 1,500
  users.

A.1.1.  Obtaining General IPv6 Network Prefixes

  In the case of a campus network, the site will typically take its
  connectivity from its National Research and Education Network (NREN).
  Southampton connects to JANET, the UK academic network, via its local
  regional network LeNSE (Learning Network South East).  JANET
  currently has a /32 allocation from RIPE NCC.  The current
  recommended practice is for sites to receive a /48 allocation; on
  this basis, Southampton has received such a prefix for its own use.
  The regional network also uses its own allocation from the NREN
  provider.

  No ULA addressing is used on site.  The campus is not multihomed
  (JANET is the sole provider), nor does it expect to change service
  provider, and thus does not plan to use ULAs for the (perceived)
  benefit of easing network renumbering.  Indeed, the campus has
  renumbered following the aforementioned renumbering procedure
  [RFC4192] on two occasions, and this has proven adequate (with
  provisos documented in [THINKABOUT]).  The campus does not see any
  need to deploy ULAs for in-band or out-of-band network management;
  there are enough IPv6 prefixes available in the site allocation for
  the infrastructure.  In some cases, use of private IP address space
  in IPv4 creates problems, so University of Southampton believes that
  the availability of ample global IPv6 address space for
  infrastructure may be a benefit for many sites.





Van de Velde, et al.         Informational                     [Page 16]

RFC 5375             IPv6 Addressing Considerations        December 2008


  No 6bone addressing is used on site any more.  Since the 6bone
  phaseout of June 2006 [RFC3701], most transit ISPs have begun
  filtering attempted use of such prefixes.

  Southampton does participate in global and organizational scope IPv6
  multicast networks.  Multicast address allocations are not discussed
  here as they are not in scope for the document.  It is noted that
  IPv6 has advantages for multicast group address allocation.  In IPv4,
  a site needs to use techniques like GLOP [RFC3180] to pick a globally
  unique multicast group to use.  This is problematic if the site does
  not use the Border Gateway Protocol (BGP) [RFC4271] and does not have
  an Autonomous System Number (ASN).  In IPv6,0 unicast-prefix-based
  IPv6 multicast addresses empower a site to pick a globally unique
  group address based on its own unicast site or link prefix.
  Embedded-RP is also in use, is seen as a potential advantage for IPv6
  and multicast, and has been tested successfully across providers
  between sites (including paths to/from the US and UK).

A.1.2.  Forming an Address (Subnet) Allocation Plan

  The campus has a /16 prefix for IPv4 use; in principle, 256 subnets
  of 256 addresses.  In reality, the subnetting is muddier, because of
  concerns of IPv4 address conservation; subnets are sized to the hosts
  within them, e.g., a /26 IPv4 prefix is used if a subnet has 35 hosts
  in it.  While this is efficient, it increases management burden when
  physical deployments change, and IPv4 subnets require resizing (up or
  down), even when DHCP is in use.

  The /48 IPv6 prefix is considerably larger than the IPv4 allocation
  already in place at the site.  It is loosely equivalent to a 'Class
  A' IPv4 prefix in that it has 2^16 (over 65,000) subnets, but has an
  effectively unlimited subnet address size (2^64) compared to 256 in
  the IPv4 equivalent.  The increased subnet size means that /64 IPv6
  prefixes can be used on all subnets, without any requirement to
  resize them at a later date.  The increased subnet volume allows
  subnets to be allocated more generously to schools and departments in
  the campus.  While address conservation is still important, it is no
  longer an impediment to network management.  Rather, address (subnet)
  allocation is more about embracing the available address space and
  planning for future expansion.

  In a dual-stack network, it was chosen to deploy the IP subnets
  congruently for IPv4 and IPv6.  This is because the systems are still
  in the same administrative domains and the same geography.  It is not
  expected to have IPv6-only subnets in production use for a while yet,
  outside the test beds and some early Mobile IPv6 trials.  With
  congruent addressing, the firewall policies are also aligned for IPv4
  and IPv6 traffic at the site border.



Van de Velde, et al.         Informational                     [Page 17]

RFC 5375             IPv6 Addressing Considerations        December 2008


  The subnet allocation plan required a division of the address space
  per school or department.  Here, a /56 was allocated to the school
  level of the university; there are around 30 schools currently.  A
  /56 of IPv6 address space equates to 256 /64 subnet allocations.
  Further /56 allocations were made for central IT infrastructure, the
  network infrastructure, and the server side systems.

A.1.3.  Other Considerations

  The network uses a Demilitarized Zone (DMZ) topology for some level
  of protection of 'public' systems.  Again, this topology is congruent
  with the IPv4 network.

  There are no specific transition methods deployed internally to the
  campus; everything is using the conventional dual-stack approach.
  There is no use of ISATAP [RFC5214] for example.

  For the Mobile IPv6 early trials, there is one allocated prefix for
  Home Agent (HA) use.  However, there has been no detailed
  consideration yet regarding how Mobile IPv6 usage may grow, and
  whether more subnets (or even every subnet) will require HA support.

  The university operates a tunnel broker [RFC3053] service on behalf
  of the United Kingdom Education and Research Network Association
  (UKERNA) for JANET sites.  This uses separate address space from
  JANET, not the university site allocation.

A.1.4.  Node Configuration Considerations

  Currently, stateless autoconfiguration is used on most subnets for
  IPv6 hosts.  There is no DHCPv6 service deployed yet, beyond tests of
  early code releases.  It is planned to deploy DHCPv6 for address
  assignment when robust client and server code is available (at the
  time of writing, the potential for this looks good, e.g., via the
  Internet Systems Consortium (ISC) implementation).  University of
  Southampton is also investigating a common integrated DHCP/DNS
  management platform, even if the servers themselves are not co-
  located, including integrated DHCPv4 and DHCPv6 server configuration,
  as discussed in [RFC4477].  Currently, clients with statelessly
  autoconfigured addresses are added to the DNS manually, though
  dynamic DNS is an option.  The network administrators would prefer
  the use of DHCP because they believe it gives them more management
  control.








Van de Velde, et al.         Informational                     [Page 18]

RFC 5375             IPv6 Addressing Considerations        December 2008


  Regarding the implications of the larger IPv6 subnet address space on
  scanning attacks [RFC5157], it is noted that all the hosts are dual-
  stack, and thus are potentially exposed over both protocols anyway.
  All addresses are published in DNS, and the site does not operate a
  two-faced DNS.

  Currently, there is internal usage of RFC 4941 privacy addresses
  [RFC4941] (certain platforms ship with it on by default), but network
  administrators may desire to disable this (perhaps via DHCP) to ease
  management complexity.  However, it is desired to determine the
  feasibility of this on all systems, e.g., for guests on wireless LAN
  or other user-maintained systems.  Network management and monitoring
  should be simpler without RFC 4941 in operation, in terms of
  identifying which physical hosts are using which addresses.  Note
  that RFC 4941 is only an issue for outbound connections, and that
  there is potential to assign privacy addresses via DHCPv6.

  Manually configured server addresses are used to avoid address
  changes based upon change of network adaptor.  With IPv6 you can pick
  ::53 for a DNS server, or you can pick 'random' addresses for
  obfuscation, though that's not an issue for publicly advertised
  addresses (dns, mx, web, etc.).

A.2.  Service Provider Considerations

  In this section an IPv6 addressing schema is sketched that could
  serve as an example for an Internet Service Provider.

  Appendix A.2.1 starts with some thoughts regarding objective
  requirements of such an addressing schema and derives a few general
  rules of thumb that have to be kept in mind when designing an ISP
  IPv6 addressing plan.

  Appendix A.2.2 illustrates the findings of Appendix A.2.1 with an
  exemplary IPv6 addressing schema for an MPLS-based ISP offering
  Internet services as well as network access services to several
  millions of customers.

A.2.1.  Investigation of Objective Requirements for an IPv6 Addressing
       Schema of a Service Provider

  The first step of the IPv6 addressing plan design for a service
  provider should identify all technical, operational, political, and
  business requirements that have to be satisfied by the services
  supported by this addressing schema.






Van de Velde, et al.         Informational                     [Page 19]

RFC 5375             IPv6 Addressing Considerations        December 2008


  According to the different technical constraints and business models
  as well as the different weights of these requirements (from the
  point of view of the corresponding service provider), it is very
  likely that different addressing schemas will be developed and
  deployed by different ISPs.  Nevertheless, the addressing schema of
  Appendix A.2.2 is one possible example.

  For this document, it is assumed that our exemplary ISP has to
  fulfill several roles for its customers such as:

  o  Local Internet Registry

  o  Network Access Provider

  o  Internet Service Provider

A.2.1.1.  Recommendations for an IPv6 Addressing Schema from the LIR
         Perspective of the Service Provider

  In its role as Local Internet Registry (LIR), the service provider
  has to care about the policy constraints of the RIRs and the
  standards of the IETF regarding IPv6 addressing.  In this context,
  the following basic recommendations have to be considered and should
  be satisfied by the IPv6 address allocation plan of a service
  provider:

  o  As recommended in RFC 3177 [RFC3177] and in several RIR policies,
     "Common" customers sites (normally private customers) should
     receive a /48 prefix from the aggregate of the service provider.
     (Note: The addressing plan must be flexible enough and take into
     account the possible change of the minimum allocation size for end
     users currently under definition by the RIRs.)

  o  "Big customers" (like big enterprises, governmental agencies,
     etc.) may receive shorter prefixes according to their needs, when
     their needs can be documented and justified to the RIR.

  o  The IPv6 address allocation schema has to be able to meet the HD-
     ratio that is proposed for IPv6.  This requirement corresponds to
     the demand for an efficient usage of the IPv6 address aggregate by
     the service provider.  (Note: The currently valid IPv6 HD-ratio of
     0.94 means an effective usage rate of about 22% of a /20 prefix of
     the service provider, on the basis of /56 assignments.)

  o  All assignments to customers have to be documented and stored into
     a database that can also be queried by the RIR.





Van de Velde, et al.         Informational                     [Page 20]

RFC 5375             IPv6 Addressing Considerations        December 2008


  o  The LIR has to make available the means for supporting the reverse
     DNS mapping of the customer prefixes.

  o  IPv6 Address Allocation and Assignment Policies can be found at
     RIRs and are similar in many aspects.  See [RIPE_Nov07],
     [RIPE_Jul07], [APNIC_IPv6], [LACNIC_IPv6], [AFRINIC_IPv6], and
     Section 6 of [ARIN].

A.2.1.2.  IPv6 Addressing Schema Recommendations from the ISP
         Perspective of the Service Provider

  From the ISP perspective, the following basic requirements can be
  identified:

  o  The IPv6 address allocation schema must be able to realize a
     maximal aggregation of all IPv6 address delegations to customers
     into the address aggregate of the service provider.  Only this
     provider aggregate will be routed and injected into the global
     routing table (DFZ, "Default-Free Zone").  This strong aggregation
     keeps the routing tables of the DFZ small and eases filtering and
     access control very much.

  o  The IPv6 addressing schema of the SP should contain optimal
     flexibility since the infrastructure of the SP will change over
     time with new customers, transport technologies, and business
     cases.  The requirement of optimal flexibility is contrary to the
     recommendation of strong IPv6 address aggregation and efficient
     address usage, but each SP has to decide which of these
     requirements to prioritize.

  o  While keeping the multilevel network hierarchy of an ISP in mind,
     note that due to addressing efficiency reasons, not all hierarchy
     levels can and should be mapped into the IPv6 addressing schema of
     an ISP.  Sometimes it is much better to implement a more "flat"
     addressing for the ISP network than to lose big chunks of the IPv6
     address aggregate in addressing each level of network hierarchy.
     (Note: In special cases, it is even recommended for really "small"
     ISPs to design and implement a totally flat IPv6 addressing schema
     without any level of hierarchy.)

  o  A decoupling of provider network addressing and customer
     addressing is recommended.  (Note: A strong aggregation (e.g., on
     POP, Aggregation Router (AG), or Label Edge Router (LER) level)
     limits the numbers of customer routes that are visible within the
     ISP network, but also brings down the efficiency of the IPv6
     addressing schema.  That's why each ISP has to decide how many
     internal aggregation levels it wants to deploy.)




Van de Velde, et al.         Informational                     [Page 21]

RFC 5375             IPv6 Addressing Considerations        December 2008


A.2.1.3.  IPv6 Addressing Schema Recommendations from the Network Access
         Provider Perspective of the Service Provider

  As already done for the LIR and the ISP roles of the SP it is also
  necessary to identify requirements that come from its Network Access
  Provider role.  Some of the basic requirements are:

  o  The IPv6 addressing schema of the SP, it must be chosen in a way
     that it can handle new requirements that are triggered from
     customer side.  For instance, this can be the customer's growing
     needs for IPv6 addresses as well as customer-driven modifications
     within the access network topology (e.g., when the customer moves
     from one point of network attachment (POP) to another).  (See
     Appendix A.2.3.4, "Changing the Point of Network Attachment".)

  o  For each IPv6 address assignment to customers, a "buffer zone"
     should be reserved that allows the customer to grow in its
     addressing range without renumbering or assignment of additional
     prefixes.

  o  The IPv6 addressing schema of the SP must deal with multiple
     attachments of a single customer to the SP network infrastructure
     (i.e., multihomed network access with the same SP).

  These few requirements are only part of the requirements a service
  provider has to investigate and keep in mind during the definition
  phase of its addressing architecture.  Each SP will most likely add
  more constraints to this list.

A.2.1.4.  A Few Rules of Thumb for Designing an ISP IPv6 Addressing
         Architecture

  As a result of the above enumeration of requirements regarding an ISP
  IPv6 addressing plan, the following design "rules of thumb" have been
  derived:

  o  No "One size fits all".  Each ISP must develop its own IPv6
     address allocation schema depending on its concrete business
     needs.  It is not practical to design one addressing plan that
     fits for all kinds of ISPs (small / big, routed / MPLS-based,
     access / transit, LIR / No LIR, etc.).

  o  The levels of IPv6 address aggregation within the ISP addressing
     schema should strongly correspond to the implemented network
     structure, and their number should be minimized because of
     efficiency reasons.  It is assumed that the SP's own





Van de Velde, et al.         Informational                     [Page 22]

RFC 5375             IPv6 Addressing Considerations        December 2008


     infrastructure will be addressed in a fairly flat way, whereas
     part of the customer addressing architecture should contain
     several levels of aggregation.

  o  Keep the number of IPv6 customer routes inside your network as
     small as possible.  A totally flat customer IPv6 addressing
     architecture without any intermediate aggregation level will lead
     to lots of customer routes inside the SP network.  A fair trade-
     off between address aggregation levels (and hence the size of the
     internal routing table of the SP) and address conservation of the
     addressing architecture has to be found.

  o  The ISP IPv6 addressing schema should provide maximal flexibility.
     This has to be realized for supporting different sizes of customer
     IPv6 address aggregates ("big" customers vs. "small" customers) as
     well as to allow future growth rates (e.g., of customer
     aggregates) and possible topological or infrastructural changes.

  o  A limited number of aggregation levels and sizes of customer
     aggregates will ease the management of the addressing schema.
     This has to be weighed against the previous "rule of thumb" --
     flexibility.

A.2.2.  Exemplary IPv6 Address Allocation Plan for a Service Provider

  In this example, the service provider is assumed to operate an MPLS-
  based backbone and to implement IPv6 Provider Edge Routers (6PE)
  [RFC4798] to provide IPv6 backbone transport between the different
  locations (POPs) of a fully dual-stacked network access and
  aggregation area.

  In addition, it is assumed that the service provider:

  o  has received a /20 from its RIR

  o  operates its own LIR

  o  has to address its own IPv6 infrastructure

  o  delegates prefixes from this aggregate to its customers

  This addressing schema should illustrate how the /20 IPv6 prefix of
  the SP can be used to address the SP's own infrastructure and to
  delegate IPv6 prefixes to its customers, following the above-
  mentioned requirements and rules of thumb as far as possible.






Van de Velde, et al.         Informational                     [Page 23]

RFC 5375             IPv6 Addressing Considerations        December 2008


  The figure below summarizes the device types in an SP network and the
  typical network design of a MPLS-based service provider.  The network
  hierarchy of the SP has to be taken into account for the design of an
  IPv6 addressing schema; it defines the basic shape of the addressing
  schema and the various levels of aggregation.

  +------------------------------------------------------------------+
  |               LSRs of the MPLS Backbone of the SP                |
  +------------------------------------------------------------------+
     |        |             |              |                 |
     |        |             |              |                 |
  +-----+  +-----+     +--------+     +--------+         +--------+
  | LER |  | LER |     | LER-BB |     | LER-BB |         | LER-BB |
  +-----+  +-----+     +--------+     +--------+         +--------+
   |   |    |   |        |    |      /     |              |     |
   |   |    |   |        |    |     /      |              |     |
   |   |    |   |  +------+  +------+   +------+          |     |
   |   |    |   |  |BB-RAR|  |BB-RAR|   |  AG  |          |     |
   |   |    |   |  +------+  +------+   +------+          |     |
   |   |    |   |    |  |      |  |      |    |           |     |
   |   |    |   |    |  |      |  |      |    |           |     |
   |   |    |   |    |  |      |  | +-----+  +-----+  +-----+  +-----+
   |   |    |   |    |  |      |  | | RAR |  | RAR |  | RAR |  | RAR |
   |   |    |   |    |  |      |  | +-----+  +-----+  +-----+  +-----+
   |   |    |   |    |  |      |  |  |   |    |   |    |   |    |   |
   |   |    |   |    |  |      |  |  |   |    |   |    |   |    |   |
  +-------------------------------------------------------------------+
  |                       Customer networks                           |
  +-------------------------------------------------------------------+

  LSR     Label Switch Router
  LER     Label Edge Router
  LER-BB  Broadband Label Edge Router
  RAR     Remote Access Router
  BB-RAR  Broadband Remote Access Router
  AG      Aggregation Router

                   Exemplary Service Provider Network

  The following should be taken into consideration when making the
  basic design decisions for the exemplary service provider IPv6
  addressing plan regarding customer prefixes.

  o  The prefixes assigned to all customers behind the same LER (or
     LER-BB) are aggregated under one LER prefix.  This ensures that
     the number of labels that have to be used for 6PE is limited and
     hence provides strong MPLS label conservation.




Van de Velde, et al.         Informational                     [Page 24]

RFC 5375             IPv6 Addressing Considerations        December 2008


  o  The /20 prefix of the SP is separated into 3 different pools that
     are used to allocate IPv6 prefixes to the customers of the SP:

     1.  A pool (e.g., /24) for satisfying the addressing needs of
         really "big" customers (as defined in Appendix A.2.2.1.1) that
         need IPv6 prefixes larger than /48 (e.g., /32).  These
         customers are assumed to be connected to several POPs of the
         access network, so that this customer prefix will be visible
         in each of these POPs.

     2.  A pool (e.g., /24) for the LERs with direct customer
         connections (e.g., dedicated line access) and without an
         additional aggregation area between the customer and the LER.
         (These LERs are mostly connected to a limited number of
         customers because of the limited number of interfaces/ports.)

     3.  A larger pool (e.g., 14*/24) for LERs (or LER-BBs) that serve
         a high number of customers that are normally connected via
         some kind of aggregation network (e.g., DSL customers behind a
         BB-RAR or dial-in customers behind a RAR).

  o  The IPv6 address delegation within each pool (the end customer
     delegation or the aggregates that are dedicated to the LER itself)
     should be chosen with an additional buffer zone of 100-300% for
     future growth.  That is, 1 or 2 additional prefix bits should be
     reserved according to the expected future growth rate of the
     corresponding customer or the corresponding network device
     aggregate.

A.2.2.1.  Defining an IPv6 Address Allocation Plan for Customers of the
         Service Provider

A.2.2.1.1.  "Big" Customers

  The SP's "big" customers receive their prefix from the /24 IPv6
  address aggregate that has been reserved for their "big" customers.
  A customer is considered a "big" customer if it has a very complex
  network infrastructure and/or huge IPv6 address needs (e.g., because
  of very large customer numbers) and/or several uplinks to different
  POPs of the SP network.

  The assigned IPv6 address prefixes can have a prefix length in the
  range 32-48 and for each assignment a 100 or 300% future growing zone
  is marked as "reserved" for this customer.  For instance, this means
  that with a delegation of a /34 to a customer the corresponding /32
  prefix (which contains this /34) is reserved for the customer's
  future usage.




Van de Velde, et al.         Informational                     [Page 25]

RFC 5375             IPv6 Addressing Considerations        December 2008


  The prefixes for the "big" customers can be chosen from the
  corresponding "big customer" pool by either using an equidistant
  algorithm or using mechanisms similar to the Sparse Allocation
  Algorithm (SAA) [RIPE_Nov07].

A.2.2.1.2.  "Common" Customers

  All customers that are not "big" customers are considered as "common"
  customers.  They represent the majority of customers, hence they
  receive a /48 out of the IPv6 customer address pool of the LER where
  they are directly connected or aggregated.

  Again a 100-300% future growing IPv6 address range is reserved for
  each customer, so that a "common" customer receives a /48 allocation
  but has a /47 or /46 reserved.

  (Note: If it is obvious that the likelihood of needing a /47 or /46
  in the future is very small for a "common" customer, then no growing
  buffer should be reserved for it, and only a /48 will be assigned
  without any growing buffer.)

  In the network access scenarios where the customer is directly
  connected to the LER, the customer prefix is directly taken out of
  the customer IPv6 address aggregate (e.g., /38) of the corresponding
  LER.

  For other cases (e.g., the customer is attached to a RAR that is
  itself aggregated to an AG or to a LER-BB), at least 2 different
  approaches are possible.

  1)  Mapping of Aggregation Network Hierarchy into Customer IPv6
      Addressing Schema.  The aggregation network hierarchy could be
      mapped into the design of the customer prefix pools of each
      network level in order to achieve a maximal aggregation at the
      LER level as well as at the intermediate levels.  (Example:
      Customer - /48, RAR - /38, AG - /32, LER-BB - /30).  At each
      network level, an adequate growing zone should be reserved.
      (Note: Of course, this approach requires some "fine tuning" of
      the addressing schema based on a very good knowledge of the
      Service Provider network topology including actual growing ranges
      and rates.)

      When the IPv6 customer address pool of a LER (or another device
      of the aggregation network -- AG or RAR) is exhausted, the
      related LER (or AG or RAR) prefix is shortened by 1 or 2 bits
      (e.g., from /38 to /37 or /36) so that the originally reserved
      growing zone can be used for further IPv6 address allocations to




Van de Velde, et al.         Informational                     [Page 26]

RFC 5375             IPv6 Addressing Considerations        December 2008


      customers.  In the case where this growing zone is exhausted as
      well, a new prefix range from the corresponding pool of the next-
      higher hierarchy level can be requested.

  2)  "Flat" Customer IPv6 Addressing Schema.  The other option is to
      allocate all the customer prefixes directly out of the customer
      IPv6 address pool of the LER where the customers are attached and
      aggregated and to ignore the intermediate aggregation network
      infrastructure.  Of course, this approach leads to a higher
      amount of customer routes at the LER and aggregation network
      level, but it takes a great amount of complexity out of the
      addressing schema.  Nevertheless, the aggregation of the customer
      prefixes to one prefix at the LER level is realized as required
      above.

  Note: The handling of changes (e.g., technically triggered changes)
  within the ISP access network is discussed briefly in
  Appendix A.2.3.5.

  If the actual observed growing rates show that the reserved growing
  zones are not needed, then they can be freed and used for assignments
  for prefix pools to other devices at the same level of the network
  hierarchy.

A.2.2.2.  Defining an IPv6 Address Allocation Plan for the Service
         Provider Network Infrastructure

  For the IPv6 addressing of the SP's own network infrastructure, a /32
  (or /40) from the "big" customers address pool can be chosen.

  This SP infrastructure prefix is used to code the network
  infrastructure of the SP by assigning a /48 to every POP/location and
  using (for instance) a /56 for coding the corresponding router within
  this POP.  Each SP internal link behind a router interface could be
  coded using a /64 prefix.  (Note: While it is suggested to choose a
  /48 for addressing the POP/location of the SP network, it is left to
  each SP to decide what prefix length to assign to the routers and
  links within the POP.)

  The IIDs of the router interfaces may be generated by using EUI-64 or
  through plain manual configuration, e.g., for coding additional
  network or operational information into the IID.

  Again, it is assumed that 100-300% growing zones are needed for each
  level of network hierarchy, and additional prefix bits may be
  assigned to POPs and/or routers if needed.





Van de Velde, et al.         Informational                     [Page 27]

RFC 5375             IPv6 Addressing Considerations        December 2008


  Loopback interfaces of routers may be chosen from the first /64 of
  the /56 router prefix (in the example above).

  (Note: The /32 (or /40) prefix that has been chosen for addressing
  the SP's own IPv6 network infrastructure leaves enough space to code
  additional functionalities like security levels or private and test
  infrastructure, although such approaches haven't been considered in
  more detail for the above-described SP until now.)

  Point-to-point links to customers (e.g., PPP links, dedicated lines,
  etc.) may be addressed using /126 prefixes out of the first /64 of
  the access routers that could be reserved for this reason.

A.2.3.  Additional Remarks

A.2.3.1.  ULA

  There are no compelling reasons for service providers to use ULAs.
  See Section 2.2.

  ULAs could be used inside the SP network in order to have an
  additional "site-local scoped" IPv6 address for the SP's own
  infrastructure, for instance, for network management reasons and in
  order to have an addressing schema that can't be reached from outside
  the SP network.

  When ULAs are used, it is possible to map the proposed internal IPv6
  addressing of the SP's own network infrastructure (as described in
  Appendix A.2.2.2) directly to the ULA addressing schema by
  substituting the /48 POP prefix with a /48 ULA site prefix.

A.2.3.2.  Multicast

  IPv6 multicast-related addressing issues are out of the scope of this
  document.

A.2.3.3.  POP Multihoming

  POP multihoming (or better, LER multihoming) of customers with the
  same SP can be realized within the proposed IPv6 addressing schema of
  the SP by assigning multiple LER-dependent prefixes to this customer
  (i.e., considering each customer location as a single customer) or by
  choosing a customer prefix out of the pool of "big" customers.  The
  second solution has the disadvantage that in every LER where the
  customer is attached, this prefix will appear inside the IGP routing
  table, thus requiring an explicit MPLS label.





Van de Velde, et al.         Informational                     [Page 28]

RFC 5375             IPv6 Addressing Considerations        December 2008


  Note: The negative effects (described above) of POP/LER multihoming
  on the addressing architecture in the SP access network are not
  resolved by implementing the Site Multihoming by IPv6 Intermediation
  (SHIM6) approach.  SHIM6 only targets a mechanism for dealing with
  multiple prefixes in end systems.  The SP is expected to have
  unaggregated customer prefixes in its internal routing tables.

A.2.3.4.  Changing the Point of Network Attachment

  In the possible case that a customer has to change its point of
  network attachment to another POP/LER within the ISP access network,
  two different approaches can be applied, assuming that the customer
  uses PA addresses out of the SP aggregate:

  1)  The customer has to renumber its network with an adequate
      customer prefix out of the aggregate of the corresponding LER/RAR
      of its new network attachment.  To minimize the administrative
      burden for the customer, the prefix should be of the same size as
      the former.  This conserves the IPv6 address aggregation within
      the SP network (and the MPLS label space) but adds additional
      burden to the customer.  Hence, this approach will most likely
      only be chosen in the case of "small customers" with temporary
      addressing needs and/or prefix delegation with address
      autoconfiguration.

  2)  The customer does not need to renumber its network and keeps its
      address aggregate.

      This approach leads to additional more-specific routing entries
      within the IGP routing table of the LER and will hence consume
      additional MPLS labels, but it is totally transparent to the
      customer.  Because this results in additional administrative
      effort and will stress the router resources (label space, memory)
      of the ISP, this solution will only be offered to the most
      valuable customers of an ISP (e.g., "big customers" or
      "enterprise customers").

      Nevertheless, the ISP again has to find a fair trade-off between
      customer renumbering and sub-optimal address aggregation (i.e.,
      the generation of additional more-specific routing entries within
      the IGP and the waste of MPLS label space).










Van de Velde, et al.         Informational                     [Page 29]

RFC 5375             IPv6 Addressing Considerations        December 2008


A.2.3.5.  Restructuring of SP (Access) Network and Renumbering

  A technically triggered restructuring of the SP (access) network (for
  instance, because of split of equipment or installation of new
  equipment) should not lead to a customer network renumbering.  This
  challenge should be handled in advance by an intelligent network
  design and IPv6 address planning.

  In the worst case, the customer network renumbering could be avoided
  through the implementation of more-specific customer routes.  (Note:
  Since this kind of network restructuring will mostly happen within
  the access network (at the level) below the LER, the LER aggregation
  level will not be harmed and the more-specific routes will not
  consume additional MPLS label space.)

A.2.3.6.  Extensions Needed for the Later IPv6 Migration Phases

  The proposed IPv6 addressing schema for an SP needs some slight
  enhancements / modifications for the later phases of IPv6
  integration, for instance, when the whole MPLS backbone
  infrastructure (LDP, IGP, etc.) is realized over IPv6 transport, and
  an IPv6 addressing of the LSRs is needed.  Other changes may be
  necessary as well but should not be explained at this point.

Appendix B.  Considerations for Subnet Prefixes Different than /64

B.1.  Considerations for Subnet Prefixes Shorter than /64

  An allocation of a prefix shorter then 64 bits to a node or interface
  is considered bad practice.  One exception to this statement is when
  using 6to4 technology where a /16 prefix is utilized for the pseudo-
  interface [RFC3056].  The shortest subnet prefix that could
  theoretically be assigned to an interface or node is limited by the
  size of the network prefix allocated to the organization.

  A possible reason for choosing the subnet prefix for an interface
  shorter than /64 is that it would allow more nodes to be attached to
  that interface compared to a prescribed length of 64 bits.  The
  prescribed /64 does include 2 functional bits, the 'g' bit and the
  inverted 'u' (universal/local) bit and these can not be chosen at
  will.  However, a larger address space then a /64 is unnecessary for
  most networks, considering that 2^62 provides plenty of node
  addresses.

  The subnet prefix assignments can be made by manual configuration, by
  a stateful Host Configuration Protocol [RFC3315], by a stateful
  prefix delegation mechanism [RFC3633], or implied by stateless
  autoconfiguration from prefix Router Advertisements (RAs).



Van de Velde, et al.         Informational                     [Page 30]

RFC 5375             IPv6 Addressing Considerations        December 2008


B.2.  Considerations for Subnet Prefixes Longer than /64

  The following subsections describe subnet prefix values that should
  be avoided in deployments because nodes who assume that the subnet
  prefix is /64 could treat them incorrectly.

B.2.1.  /126 Addresses

  126-bit subnet prefixes are typically used for point-to-point links
  similar to a the IPv4 address-conservative /30 allocation for point-
  to-point links.  The usage of this subnet address length does not
  lead to any considerations beyond those discussed earlier in this
  section, particularly those related to the 'u' and 'g' bits (see
  B.2.4.

B.2.2.  /127 Addresses

  The usage of the /127 addresses, the equivalent of IPv4's RFC 3021
  [RFC3021], is not valid and should be strongly discouraged as
  documented in RFC 3627 [RFC3627].

B.2.3.  /128 Addresses

  The 128-bit address prefix may be used in those situations where we
  know that one, and only one, address is sufficient.  Example usage
  would be the off-link loopback address of a network device.

  When choosing a 128 bit prefix, it is recommended to take the 'u' and
  'g' bits into consideration and to make sure that there is no overlap
  with any of the following well-known addresses:

  o  Subnet Router Anycast Address

  o  Reserved Subnet Anycast Address

  o  Addresses used by Embedded-RP

  o  ISATAP Addresses

B.2.4.  EUI-64 'u' and 'g' Bits

  When using subnet prefix lengths other than /64, the interface
  identifier cannot be in Modified EUI-64 format as required by
  [RFC4291].  However, nodes not aware that a prefix length other than
  /64 is used might still think it's an EUI-64; therefore, it's prudent
  to take into account the following points when setting the bits.





Van de Velde, et al.         Informational                     [Page 31]

RFC 5375             IPv6 Addressing Considerations        December 2008


  Address space conservation is the main motivation for using a subnet
  prefix length longer than 64 bits; however, this kind of address
  conservation is of little benefit compared with the additional
  considerations one must make when creating and maintaining an IPv6
  addressing plan.

  The address assignment can be made either by manual configuration or
  by a stateful Host Configuration Protocol [RFC3315].

  When assigning a subnet prefix of more then 70 bits, according to RFC
  4291 [RFC4291], 'u' and 'g' bits (the 71st and 72nd bit,
  respectively) need to be taken into consideration and should be set
  correctly.

  The 71st bit of a IPv6 address is the inverted 'u' (universal/local)
  bit and is used to determine whether the address is universally or
  locally administered.  If 1, the IEEE, through the designation of a
  unique company ID, has administered the address.  If 0, the address
  is locally administered.  The network administrator has overridden
  the manufactured address and specified a different address.

  The 'g' (the individual/group) bit is the 72nd bit and is used to
  determine whether the address is an individual address (unicast) or a
  group address (multicast).  If '0', the address is a unicast address.
  If '1', the address is a multicast address.

  In current IPv6 protocol stacks, the relevance of the 'u' and 'g'
  bits is marginal and typically will not give an error when configured
  wrongly; however, future implementations may turn out differently if
  they process the 'u' and 'g' bits in IEEE-like behavior.

  When using subnet lengths longer then 64 bits, it is important to
  avoid selecting addresses that may have a predefined use and could
  confuse IPv6 protocol stacks.  The alternate usage may not be a
  simple unicast address in all cases.  The following points should be
  considered when selecting a subnet length longer then 64 bits.

B.2.5.  Anycast Addresses

B.2.5.1.  Subnet Router Anycast Address

  RFC 4291 [RFC4291] provides a definition for the required Subnet
  Router Anycast Address as follows:

   |                   n bits                   |   128-n bits   |
   +--------------------------------------------+----------------+
   |               subnet prefix                | 00000000000000 |
   +--------------------------------------------+----------------+



Van de Velde, et al.         Informational                     [Page 32]

RFC 5375             IPv6 Addressing Considerations        December 2008


  It is recommended to avoid allocating this IPv6 address to a device
  that expects to have a normal unicast address.

B.2.5.2.  Reserved IPv6 Subnet Anycast Addresses

  RFC 2526 [RFC2526] stated that within each subnet, the highest 128
  interface identifier values are reserved for assignment as subnet
  anycast addresses.

  The construction of a reserved subnet anycast address depends on the
  type of IPv6 addresses used within the subnet, as indicated by the
  format prefix in the addresses.

  The first type of Subnet Anycast addresses have been defined as
  follows for the Modified EUI-64 format:

   |           64 bits            |      57 bits     |   7 bits   |
   +------------------------------+------------------+------------+
   |        subnet prefix         | 1111110111...111 | anycast ID |
   +------------------------------+------------------+------------+

  The anycast address structure implies that it is important to avoid
  creating a subnet prefix where the bits 65 to 121 are defined as
  "1111110111...111" (57 bits in total) in order to prevent confusion.

  For other IPv6 address types (that is, with format prefixes other
  than those listed above), the interface identifier is not in 64-bit
  extended unique identifier (EUI-64) format and may not be 64 bits in
  length.  The reserved subnet anycast addresses for such address types
  are constructed as follows:

   |           n bits             |    121-n bits    |   7 bits   |
   +------------------------------+------------------+------------+
   |        subnet prefix         | 1111111...111111 | anycast ID |
   +------------------------------+------------------+------------+
                                  |   interface identifier field  |

  It is recommended to avoid allocating this IPv6 address to a device
  that expects to have a normal unicast address.

B.2.6.  Addresses Used by Embedded-RP (RFC 3956)

  Embedded-RP [RFC3956] reflects the concept of integrating the
  Rendezvous Point (RP) IPv6 address into the IPv6 multicast group
  address.  Due to this embedding and the fact that the length of the
  IPv6 address AND the IPv6 multicast address are 128 bits, it is not
  possible to have the complete IPv6 address of the multicast RP
  embedded as such.



Van de Velde, et al.         Informational                     [Page 33]

RFC 5375             IPv6 Addressing Considerations        December 2008


  This results in a restriction of 15 possible RP-addresses per prefix
  that can be used with embedded-RP.  The space assigned for the
  embedded-RP is based on the 4 low-order bits, while the remainder of
  the Rendezvous Interface ID (RIID) is set to all '0'.  The format of
  the IPv6 multicast group address used by embedded-RP is as follows:

              (IPv6-prefix (64 bits))(60 bits all '0')(RIID)

                  where: (RIID) = 4 bits.

  This format implies that when selecting subnet prefixes longer than
  64, and when the bits beyond the 64th bit are non-zero, the subnet
  cannot use embedded-RP.

  In addition, it is discouraged to assign a matching embedded-RP IPv6
  address to a device that is not a real Multicast Rendezvous Point,
  even though it would not generate major problems.

B.2.7.  ISATAP Addresses

  ISATAP [RFC5214] is an experimental automatic tunneling protocol used
  to provide IPv6 connectivity over an IPv4 campus or enterprise
  environment.  In order to leverage the underlying IPv4
  infrastructure, the IPv6 addresses are constructed in a special
  format.

  An IPv6 ISATAP address has the IPv4 address embedded, based on a
  predefined structure policy that identifies them as an ISATAP
  address.  The format is as follows:

               [IPv6 Prefix (64 bits)][0000:5EFE][IPv4 address]

  When using a subnet prefix length longer then 64 bits, it is good
  engineering practice to ensure that the portion of the IPv6 prefix
  from bit 65 to the end of the host-ID does not match with the well-
  known ISATAP [0000:5EFE] address when assigning an IPv6 address to a
  non-ISATAP interface.

  Note that the definition of ISATAP does not support multicast.












Van de Velde, et al.         Informational                     [Page 34]

RFC 5375             IPv6 Addressing Considerations        December 2008


Authors' Addresses

  Gunter Van de Velde
  Cisco Systems
  De Kleetlaan 6a
  Diegem  1831
  Belgium
  Phone: +32 2704 5473
  EMail: [email protected]


  Ciprian Popoviciu
  Cisco Systems
  7025-6 Kit Creek Road
  Research Triangle Park, North Carolina
  USA
  EMail: [email protected]


  Tim Chown
  University of Southampton
  Highfield
  Southampton  SO17 1BJ
  United Kingdom
  Phone: +44 23 8059 3257
  EMail: [email protected]


  T-Systems Enterprise Services GmbH
  Goslarer Ufer 35
  Berlin  10589
  Germany
  Phone: +49 30 3497 3124
  EMail: [email protected]


  Christian Hahn
  T-Systems Enterprise Services GmbH
  Goslarer Ufer 35
  Berlin  10589
  Germany
  Phone: +49 30 3497 3164
  EMail: [email protected]








Van de Velde, et al.         Informational                     [Page 35]


You are viewing proxied material from gopher.fnord.one. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.