In the previous example, the client port numbers, 1331 and 1555, did not change at the NAT-enabled router. This is not a very likely scenario, because there is a good chance that these port numbers may have already been attached to other active sessions.

PAT attempts to preserve the original source port. However, if the original source port is already used, PAT assigns the first available port number starting from the beginning of the appropriate port group 0–511, 512–1,023, or 1,024–65,535. When there are no more ports available and there is more than one external address in the address pool, PAT moves to the next address to try to allocate the original source port. This process continues until there are no more available ports or external IP addresses.

Click the Play button in the figure to view PAT operation.

In the animation, the hosts have chosen the same port number 1444. This is acceptable for the inside address, because the hosts have unique private IP addresses. However, at the NAT router, the port numbers must be changed; otherwise, packets from two different hosts would exit R2 with the same source address. In this example, PAT has assigned the next available port (1445) to the second host address.