We saw in Chapters 5 and 6 how to perform server load balancing. The most common model is to create a VIP and let that be the target or destination to which the user connects. Based on the ability of the content switch, the manipulation of the addresses and inspection of the data is the function of the load balancing process. This process, which replaces the DIP with that of the real server, is by far the most common. It is extremely simple to implement, as all the intelligence is inherent within the switch. Why, then, would you want to use anything else to perform load balancing?
Applications that require the source and destination IP address to be maintained are a prime example of why you would use application redirection.
Application redirection does not do any manipulation of the IP addresses; it merely substitutes the destination MAC address with that of the destination device selected by the load balancing metric and maintains the original source and destination IP address. The difference between this and a traditional router is that a load balancing decision has been made at this stage. This obviously makes our traditional load-balancing model a little harder to implement, as we do not have a VIP we can use as the pivotal point (or destination) of the session. In the normal load-balancing model, you would not use application redirection. However, if you require a session, based on its service (TCP port) used, to be sent to a set of servers or even to a VIP, application redirection is the most efficient mechanism to achieve this. Figure 8-1 shows how we can intercept certain applications and steer them to different devices.
Figure 8-1. Application redirection shows how the content switch inspects the data and redirects sessions, based on service used, to the appliance. Other sessions are passed through normally.