NAT on Cisco routers allows network engineers to provide TCP Load
Distribution among hosts. A representation of TCP Load Distribution is given in
Figure
. Steps to
configure TCP load distribution with NAT are in Figure
.
Although TCP load distribution can improve the performance of some types of
network transactions (such as accessing a corporate intranet web service), it
can also be a source of network complexity, resulting in intermittent fault
behaviors when something goes wrong.
Example 1 – host
fault
Assume that the network is configured as depicted in Figure
. The router is
distributing TCP connections evenly among the hosts, as it should. Power to
Host 1.1.1.2 fails and the device is no longer on the network. Because the
router does not know this, it continues to forward TCP connection requests to
the failed host. The result of this is that one in three connection attempts
fail, presenting a seemingly intermittent fault.
The network engineer
realizes that every third connection attempt is failing and immediately
suspects that one of the hosts may have failed. Power is restored to Host
1.1.1.2 and network performance returns to normal.
Example 2 – host
misconfiguration
Using the same network configuration from Figure
, assume a new
network engineer is instructed to build a new server. Because the new network
engineer does not understand how the TCP load distribution system works, the
new server is configured with the IP address assigned to the virtual host on
the NAT router. When the new server is powered up, it detects an IP address
conflict and cannot establish a connection to the network. The new engineer
examines the interface configuration on the router to locate and confirm the
duplicate IP address, but cannot find it in the interface configuration.
After discussing the problem with a more experienced engineer, the new
engineer reconfigures the new server with correct IP settings, reboots, and
connects to the network.