While it is important to devise a workable algorithm to load balance traffic, it is also important to make this information transparently available to users.
Technologies such as The Linux Virtual Server[6] that implement Layer 4 Switching5 while very effective for load balancing traffic on a Local Area Network (LAN), do not extend well to load balancing over a Wide Area Network (WAN) such as the Internet. These technologies rely on all in-bound packets and, often, all return packets passing through a single point. While this is acceptable on a LAN where all packets must pass through a limited number of switches and routers, the fundamental problem with using this in the context of a WAN is that having all traffic pass through one site, only to be sent to another, has the potential to significantly increase latency. This also has the potential to reduce reliability as packets are traversing more hops across potentially uncontrolled networks.
A step forward would be to provide a mechanism for connections to be redirected to another site, such that once the redirection has been made clients communicate directly to the site they have been redirected to. It also makes sense to allow any participating site to make this redirection. Thus, no site would be a single point of failure for establishing or maintaining connections for the network presence as a whole. Two ways of achieving this are by using DNS and HTTP redirects.