Load Balancing (LB) is a very important feature for many organizations, because the number of users that the organization needs to service may be greater than what a single UAG server can handle. Once a single server has maxed out its CPU or memory, the only thing you can do is to distribute the load between two or more servers. Load balancing is done via a mechanism that intercepts a user's request, and redirects it to one of the servers configured as a load-balanced cluster. There are plenty of load-balancer devices on the market, such as F5's Big-IP or a Celestix XLB device, and Windows itself also has a built-in mechanism to do the same thing.
Most load balancing services can also check the servers they are balancing for signs of life, and in case a server stops responding, re-distribute the load between the other members; so that users have as little interruption as possible (it's still interrupted, though—more about that later).
When servers are load balanced, they are...