TCP load distribution is a dynamic form of destination IP address translation that can be configured
for certain outside network traffic to be mapped to a valid inside network for IP traffic
destined for more than one node. After a mapping scheme is created, destination IP addresses
matching an access list are replaced with an address from a rotary pool on a round-robin basis.
When a new connection is established from the outside network to the inside network, all
non-TCP traffic will be passed without being translated, unless another translation type is
applied to the interfaces. Figure 31.4 illustrates TCP load distribution, which is explained in further
detail next.
Let’s look at the process NAT uses to map one virtual host to several real hosts:
1. In Figure 31.4, the PC using global IP address 206.2.2.25 opens a TCP connection to a virtual
host at 200.1.1.25.
2. The NAT border router receives this new connection request and creates a new translation,
which allocates the next real host of 10.1.2.25 for the inside local IP address and adds this
information to the NAT table.
3. The NAT border router replaces the destination IP address with the selected real host IP
address and then forwards the packet.
4. The real host at IP address 10.1.2.25 receives the packet and responds.
5. The NAT border router receives the packet and performs another NAT table lookup by
using the inside local IP address and port number and the outside IP address and port number
as the key. The NAT border router then translates the source address to the virtual
host’s address and forwards the packet.
6. The next connection request to that inside global IP address causes the NAT border router
to allocate 10.1.2.26 for the inside local address.
TCP load distribution steps
200.1.1.0
NAT border
router
NAT Table
Inside IP
10.1.2.25:80
Inside global IP
200.1.1.25:80
Outside global IP
206.2.2.25:3058
200.1.1.25
PC
206.2.2.25
10.1.2.25
4 1
2
3
Internet