Scalable Operation of Address Translators with Per-Interface Bindings
draft-arkko-dual-stack-extra-lite-00

Abstract

This document explains how to employ address translation in networks that serve a large number of individual customers without requiring a correspondingly large amount of private IPv4 address space.

Status of this Memo

This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts.

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as “work in progress.”

The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt.

The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html.

This Internet-Draft will expire on August 13, 2010.

Copyright Notice

Copyright (c) 2010 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. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the BSD License.

1.  Introduction

This document explains how to employ address translation without consuming a large amount of private address space. This is important in networks that serve a large number of individual customers. Networks that serve more than 2^24 (16 million) users cannot assign a unique private IPv4 address to each user, because the largest reserved private address block reserved is 10/8 [RFC1918] (Rekhter, Y., Moskowitz, R., Karrenberg, D., Groot, G., and E. Lear, “Address Allocation for Private Internets,” February 1996.). Many networks are already hitting these limits today, for instance, in the consumer Internet service market. Even some individual devices may approach these limits, for instance, cellular network gateways or mobile IP home agents.

If ample IPv4 address space was available, this would be a non-issue, because the current practice of assigning public IPv4 addresses to each user would remain viable, and the complications associated with using the more limited private address space could be avoided. However, as the IPv4 address pool is becoming depleted, this practice is becoming increasingly difficult to sustain.

It has been suggested that more of the unassigned IPv4 space should be converted for private use, in order to allow the provisioning of larger networks with private IPv4 address space. At the time of writing, the IANA "free pool" contained only 24 unallocated unicast IPv4 /8 prefixes. Although reserving a few of those for private use would create some breathing room for such deployments, it would not result in a solution with long-term viability, would result in significant operational and management overheads, and would further reduce the number of available IPv4 addresses.

Segmenting a network into areas of overlapping private address space is another possible technique, but it severely complicates the design and operation of a network.

Finally, the transition to IPv6 will eventually eliminate these addressing limitations. However, during the migration period when IPv4 and IPv6 have to co-exist, there will be the need to reach IPv4 destinations, which involves the use of address or protocol translation.

The rest of this document is organized as follows. Section 2 (Solution Outline) gives an outline of the solution, Section 3 (Terms) introduces some terms, Section 4 (Per-Interface Bindings) specifies the required behavior for managing NAT bindings and Section 5 (IPv6 Considerations) discusses the use of this technique with IPv6.

2.  Solution Outline

The need for address or protocol translation during the migration period to IPv6 creates the opportunity to deploy these mechanisms in a way that allows the support of a large user base without the need for a correspondingly large IPv4 address block.

A Network Address Translator (NAT) is typically configured to connect a network domain that uses private IPv4 addresses to the public Internet. The NAT device - which is configured with a public IPv4 address - creates and maintains a mapping for each communication session from a device inside the domain it serves to devices in the public Internet. It does that by translating the packet flow of each session such that the externally visible traffic uses only public addresses.

In most NAT deployments, the network domain connected by the NAT to the public Internet is a broadcast network sharing the same media, where each individual device must have a unique IPv4 address. In such deployments it is natural to also implement the NAT functionality such that it uses this unique IPv4 address when looking up which mapping should be used to translate a given communication session.

It is important to note, however, that this is not an inherent requirement. When other methods of identifying the correct mapping are available, and the NAT is not connecting a shared-media broadcast network to the Internet, there is no need to assign each device in the domain a unique IPv4 address.

This is the case, for example, when the NAT connects devices to the Internet that connect to it with individual point-to-point links. In this case, it becomes possible to use the same private addresses many times, making it possible to support any number of devices behind a NAT using very few IPv4 addresses.

There are tunneling-based techniques to reach the same benefits, by establishing new tunnels over any IP network [I‑D.ietf‑softwire‑dual‑stack‑lite] (Durand, A., Droms, R., Haberman, B., Woodyatt, J., Lee, Y., and R. Bush, “Dual-Stack Lite Broadband Deployments Following IPv4 Exhaustion,” March 2010.). However, where the point-to- point links already exists, creating an additional layer of tunneling is unnecessary. The approach described in this document can be implemented and deployed within a single device and has no effect to hosts behind it. In addition, as no additional layers of tunneling are introduced, there is no effect to the Maximum Transfer Unit (MTU) settings.

Note, however, that existing tunnels are a common special case of point-to-point links. For instance, cellular network gateways terminate a large number of tunnels that are already needed for mobility management reasons. Implementing the approach described in this document is particularly attractive in such environments, given that no additional tunneling mechanisms, negotiation, or host changes are required. In addition, since there is no additional tunneling, packets continue to take the same path as they would normally take. Other commonly appearing network technology that may be of interest include Point-to-Point Protocol (PPP) [RFC1661] (Simpson, W., “The Point-to-Point Protocol (PPP),” July 1994.) links, PPP over Ethernet (PPPoE) [RFC2516] (Mamakos, L., Lidl, K., Evarts, J., Carrel, D., Simone, D., and R. Wheeler, “A Method for Transmitting PPP Over Ethernet (PPPoE),” February 1999.) encapsulation, Asynchronous Transfer Mode (ATM) Permanent Virtual Circuits (PVCs), and per-subscriber virtual LAN (VLAN) allocation in consumer broadband networks.

The approach described here also results overlapping private address space, like the segmentation of the network to different areas. However, this overlap is applied only at the network edges, and does not impact routing or reachability of servers in a negative way.

3.  Terms

In this document, the key words "MAY", "MUST, "MUST NOT", "OPTIONAL", "RECOMMENDED", "SHOULD", and "SHOULD NOT", are to be interpreted as described in [RFC2119] (Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” March 1997.).

"NAT" in this document includes both "Basic NAT" and "Network Address/Port Translator (NAPT)" as defined by [RFC2663] (Srisuresh, P. and M. Holdrege, “IP Network Address Translator (NAT) Terminology and Considerations,” August 1999.). The term "NAT Session" is adapted from [RFC5382] (Guha, S., Biswas, K., Ford, B., Sivakumar, S., and P. Srisuresh, “NAT Behavioral Requirements for TCP,” October 2008.) and is defined as follows.

NAT Session - A NAT session is an association between a transport layer session as seen in the internal realm and a session as seen in the external realm, by virtue of NAT translation. The NAT session will provide the translation glue between the two session representations.

This document uses the term mapping as defined in [RFC4787] (Audet, F. and C. Jennings, “Network Address Translation (NAT) Behavioral Requirements for Unicast UDP,” January 2007.) to refer to state at the NAT necessary for network address and port translation of sessions.

4.  Per-Interface Bindings

To support a mode of operation that uses a fixed number of IPv4 addresses to serve an arbitrary number of devices, a NAT MUST manage its mappings on a per-interface basis, by associating a particular NAT session not only with the five tuples used for the transport connection on both sides of the NAT, but also with the internal interface on which the user device is connected to the NAT. This approach allows each internal interface to use the same private IPv4 address range.

For deployments where exactly one user device is connected with a separate tunnel interface and all tunnels use the same IPv4 address for the user devices, it is redundant to store this address in the mapping in addition to the internal interface identifier. When the internal interface identifier is shorter than a 32-bit IPv4 address, this may decrease the storage requirements of a mapping entry by a small measure, which may aid NAT scalability. For other deployments, it is likely necessary to store both the user device IPv4 address and the internal interface identifier, which slightly increases the size of the mapping entry.

This mode of operation is only suitable in deployments where user devices connect to the NAT over point-to-point links. If supported, this mode of operation SHOULD be configurable, and it should be disabled by default.

5.  IPv6 Considerations

Private address space conservation is important even during the migration to IPv6, because it will be necessary to communicate with the IPv4 Internet for a long time. This document specifies two recommended deployment models for IPv6. In the first deployment model the mechanisms specified in this document are useful. In the second deployment model no additional mechanisms are needed, because IPv6 addresses are already sufficient to distinguish mappings from each other.

The first deployment model employs dual stack [RFC4213] (Nordmark, E. and R. Gilligan, “Basic Transition Mechanisms for IPv6 Hosts and Routers,” October 2005.). The IPv6 side of dual stack operates based on global addresses and direct end-to-end communication. However, on the IPv4 side private addressing and NATs are a necessity. The use of per-interface NAT mappings is RECOMMENDED for the IPv4 side under these circumstances. Per-interface mappings help the NAT scale, while dual stack operation helps reduce the pressure on the NAT device by moving key types of traffic to IPv6, eliminating the need for NAT processing.

The second deployment model involves the use of address and protocol translation, such as the one defined in [I‑D.ietf‑behave‑v6v4‑xlate‑stateful] (Bagnulo, M., Matthews, P., and I. Beijnum, “Stateful NAT64: Network Address and Protocol Translation from IPv6 Clients to IPv4 Servers,” March 2010.). In this deployment model there is no IPv4 in the internal network at all. This model is applicable only in situations where all relevant devices and applications are IPv6-capable. In this situation, per-interface mappings could be employed as specified above, but they are generally unnecessary as the IPv6 address space is large enough to provide a sufficient number of mappings.

6.  Security Considerations

This practices outlined in this document do not affect the security properties of address translation.

7.  IANA Considerations

This document has no IANA implications.

8.  References

8.1. Normative References

8.2. Informative References

Appendix A.  Acknowledgments

The ideas in this draft were first presented in [I‑D.ietf‑softwire‑dual‑stack‑lite] (Durand, A., Droms, R., Haberman, B., Woodyatt, J., Lee, Y., and R. Bush, “Dual-Stack Lite Broadband Deployments Following IPv4 Exhaustion,” March 2010.). This document also in debt to [I‑D.arkko‑townsley‑coexistence] (Arkko, J. and M. Townsley, “IPv4 Run-Out and IPv4-IPv6 Co-Existence Scenarios,” July 2009.) and [I‑D.miles‑behave‑l2nat] (Miles, D. and M. Townsley, “Layer2-Aware NAT,” March 2009.). However, all of these documents focused on additional components, such as tunneling protocols or the allocation of special IP address ranges. We wanted to publish a specification that just focuses on the core functionality of a per-interface NAT mappings.

The authors would also like to thank Alain Durand, Randy Bush, Fredrik Garneij, Dan Wing, Christian Vogt, Marcelo Braun, Joel Halpern, Wassim Haddad and others for interesting discussions in this problem space.

Authors' Addresses