The process of NAT overload is the same whether a pool of addresses is used or a single address is used. Continuing with the previous PAT example, using a single public IPv4 address, PC1 wants to communicate with the web server, Svr1. At the same time another client, PC2, wants to establish a similar session with the web server Svr2. Both PC1 and PC2 are configured with private IPv4 addresses, with R2 enabled for PAT.
PC to Server Process
1. Figure 1 shows both PC1 and PC2 sending packets to Svr1 and Svr2, respectively. PC1 has the source IPv4 address 192.168.10.10 and is using TCP source port 1444. PC2 has the source IPv4 address 192.168.10.11 and is coincidentally assigned the same source port of 1444.
2. The packet from PC1 reaches R2 first. Using PAT, R2 modifies the source IPv4 address to 209.165.200.225 (inside global address). There are no other devices in the NAT table using port 1444, so PAT maintains the same port number. The packet is then forwarded towards Svr1 at 209.165.201.1.
3. Next, the packet from PC2 arrives at R2. PAT is configured to use a single inside global IPv4 address for all translations, 209.165.200.225. Similar to the translation process for PC1, PAT changes PC2’s source IPv4 address to the inside global address 209.165.200.225. However, PC2 has the same source port number as a current PAT entry, the translation for PC1. PAT increments the source port number until it is a unique value in its table. In this instance, the source port entry in the NAT table and the packet for PC2 receives 1445.
Although PC1 and PC2 are using the same translated address, the inside global address of 209.165.200.225, and the same source port number of 1444; the modified port number for PC2 (1445) makes each entry in the NAT table unique. This will become evident with the packets sent from the servers back to the clients.
Server to PC Process
4. As shown in Figure 2, in a typical client-server exchange, Svr1 and Svr2 respond to the requests received from PC1 and PC2, respectively. The servers use the source port from the received packet as the destination port, and the source address as the destination address for the return traffic. The servers seem as if they are communicating with the same host at 209.165.200.225; however, this is not the case.
5. As the packets arrive, R2 locates the unique entry in its NAT table using the destination address and the destination port of each packet. In the case of the packet from Svr1, the destination IPv4 address of 209.165.200.225 has multiple entries but only one with the destination port 1444. Using the entry in its table, R2 changes the destination IPv4 address of the packet to 192.168.10.10, with no change required for the destination port. The packet is then forwarded toward PC1.
6. When the packet from Svr2 arrives R2 performs a similar translation. The destination IPv4 address of 209.165.200.225 is located, again with multiple entries. However, using the destination port of 1445, R2 is able to uniquely identify the translation entry. The destination IPv4 address is changed to 192.168.10.11. In this case, the destination port must also be modified back to its original value of 1444, which is stored in the NAT table. The packet is then forwarded toward PC2.