To Be or NAT to Be?
by George Lawton
The imminent exhaustion of IPv4 addresses is driving the transition to IPv6. But carriers face the challenge of supporting the new addressing scheme, while maintaining backward compatibility with existing IPv4 traffic. Originally, the IETF advocated a transition approach, called dual stack, of having both IP stacks running simultaneously — mainly because IPv6 is not backwards compatible with IPv4. Given this, some legacy applications might always require some way to the bridge the two protocols.
This is where network address translation (NAT) came in. The quick fix was a NAT and protocol translation (NAT-PT) approach. But multiple performance and security issues plague NAT-PT, and the approach can break applications.
Recently, Comcast developed another approach called Dual-Stack Lite (DS-Lite), which promises to solve these problems. Core-router vendors Cisco and Juniper, along with tunnel-broker vendor GoGoNET6 announced support for DS-Lite, and carriers are waiting on the sidelines to see if it works as well as expected. Others believe that a better implementation of NAT, called NAT64 could provide a smoother transition.
Planning the Path to IPv6
IPv6 proponents started planning the transition to IPv6 in 1993, but infighting among various camps and the adoption of NAT technology forestalled the process. Although NAT helped slow the need for the transition, the growth of the Internet has finally exhausted the supply of new IPv4 addresses.
"I believe that IPv6 is here now," said Bob Fink, who helped develop the new protocol while a computer scientist at Lawrence Berkeley National Laboratory. "I'm hearing from more than enough people of a newer generation that doesn’t care about all of the old reasons they didn't want to go to IPv6 20 years ago. They know that if they want a single network that grows, you have to implement IPv6."
The IETF has been working on a technical solution to the IPv4/IPv6 transition for over a decade. In 2000, it issued RFC 2766 which proposed NAT-PT as one transition approach. But researchers uncovered numerous problems with it, such as breaking the end-to-end use of IPSec, said Jeremy Duncan, IPv6 architect for Command Information.
NAT-PT also uses up random TCP and UDP ports, which could cause applications that rely on the ports to malfunction. It also causes packet fragmentation and breaks the function of DNS record translation, which reduces network efficiency and functionality. After the first round of experimentation, the IETF deprecated NAT-PT to historical status, elaborating its technical problems in RFC 4966 in 2007.
Building a Better NAT
Interest resurged in finding other ways to support ITv4-IPv6 translation and make it friendlier on the network, said Duncan. Most of the approaches involve two NAT levels: one on the customer premise and another between the customer premises and the carrier. These schemes are usually described in terms of where the transition takes place. For example, NAT 444 uses multiple banks of IPv4 addresses. This extends the use of IPv4 protocols but doesn’t solve the security, latency, and performance issues, said Duncan. It therefore hasn't seen much traction.
Carrier Grade NAT (CGN) is a more robust approach that solves many of these problems and provides multiple configurations such as NAT 64, DNS 64, and NAT 464, which connect IPv4 equipment on the premise with either the IPv6 network or the IPv4 network. But CGN's extra layers of translation — around three — can introduce extra computation and latency when IPv6 equipment on the customer’s premise needs to be translated to IPv4 and back.
"A lot of carriers looked at this issue and thought it was an incredibly complex solution," said Duncan.
Tunnel or Translate?
To solve these problems, Comcast's DS-Lite encapsulates the IPv4 packets into IPv6 and leaves native IPv6 packets untouched. Given that IPv6 is never translated, DS-Lite requires less overhead than approaches that translate between the two protocols. With tunneling IPv4, Duncan explained, the router just has to encapsulate the IPv4 header, which is simpler than having to map the local private IPv4 address to a port number.
CGN support predates DS-Lite in core routers, but most carriers are holding off on implementing either technology until the Comcast trials have demonstrated real-world robustness, said Duncan. If it works as expected, he sees DS-Lite becoming the dominant approach in the service provider transition to IPv6.
Other industry experts believe that NAT 64 provides a cleaner migration path for most network providers because it uses only one box on the network for translation instead of the two required for tunneling.
Doug Junkins, CTO of NTT America explained, "DS-Lite works well for cases when the network operator controls the router in the home and can make sure it's implemented properly there. But NAT 64 doesn't have the requirement for both ends of the connection to be similar."
George Lawton is a freelance journalist in Guerneville, CA. Contact him via his website http://glawton.com.
