A 6to4 gateway with co-located NAT

Abstract

IPv6 was proposed as the next-generation IP primarily to deal with the problem of IPv4 address depletion caused by the rapid growth of the Internet. 6to4 tunneling is one of the currently used transition mechanisms for enabling IPv6 devices and networks to connect to today's Internet, which is primarily IPv4-based. Since most internal networks use private IPv4 addresses, it becomes necessary to provide both 6to4 and NAT functionality at the network boundary in order to handle IPv4 and IPv6 traffic. We evaluate the performance of a 6to4 Linux gateway with a co-located NAT. To enable 6to4 and NAT overhead to be determined, we also compare performance of the Linux gateway and a compatible 6to4 gateway with a co-located NAT that runs on a bare PC with no operating system or kernel installed. We describe the design and implementation of the bare PC 6to4 gateway with a co-located NAT, and also compare the performance of 6to4 in a test LAN with a co-located and a decoupled NAT (where 6to4 and NAT run on different devices). We conducted experiments with HTTP and VoIP traffic, and also measured RTT and CPU utilization. The results show that performance using 6to4 with a co-located NAT is better than with a decoupled NAT regardless of whether a Linux or a bare 6to4 gateway is used. In general, performance improvements with a co-located versus a decoupled NAT range from 34%-57% for the bare PC gateway and 7%-45% for the Linux gateway. Furthermore, performance improvements for a bare PC versus a Linux gateway range from 23%-86% with co-located and decoupled NATs. A 6to4 gateway with a co-located NAT can be used to improve network performance during the IPv6-IPv4 transition.