I work for a company that develops a web server that runs in a customer datacenter. The customer wants the server to have hot failover and I’m currently investigating HA approaches such as Pacemaker and LVM (might as well have the backup doing something so I’m using load balancing as well).
It occurs to me that some of these features are also provided by cloud computing (e.g. private cloud) technologies. Specifically auto-scaling which is provided by EC2 and cloud foundry addresses load balancing and failover support to some degree.
As a new application would I be better off spending my time trying to setup the application as a private cloud or by utilizing the tried and true HA architectures?
That depends on whether or not scaling is in the foreseeable future.
The jump from hot failover to cloud is not trivial; the big difference is that you have more then one machine doing the same thing, which introduces a host of new problems to deal with.
OTOH, if scaling is definitely going to happen, then you should probably jump straight to it.