Network Address Translation (NAT) is a method by which Internet Protocol (IP) addresses are ranslated from one group to another, in a manner transparent to the end users. It translates the source and destination addresses and ports in the Internet Protocol datagram. There are several benefits for using NAT. NAT can be installed without changes to hosts or routers, it allows reuse of globally routable addresses, it facilitates easy migration or addition of new networks and it provides a method to keep private network addresses hidden from the outside world.
NAT, however, is a processor- and memory-intensive activity for any device that implements it. This is because NAT involves reading from and writing to the header and payload information of every IP packet to do the address translation, a performance-intensive activity. It causes an increase in Central Processing Unit (CPU) and memory utilization and may impair throughput and increase the latency experienced by a packet. Thus, understanding the performance impact of NAT on a network device (in particular, a router) becomes an important factor when implementing NAT in any live network.
This thesis aims to understand and quantify the impact of Network Address Translation on a network router by doing a series of performance tests after specifying the performance parameters to measure and, then, clearly defining the performance testing methodology that is used to study each of the performance parameters. After a discussion of previous research, the measurement system and subsequent measurement results are described.
