[staunch]

If the hardware goes down, so do the servers (just like EC2), but they can be brought back up on new hardware rather quickly.

You can of course setup redundancy by creating multiple servers on different hardware nodes.