I recently came across a situation in which I needed to assemble some reads from a critter that we gathered from a couple of 454 sequencing runs. Unfortunately, the assembly of these runs using gsAssembler 2.5.3 required a bit more RAM than available on our local workstation running a Fedora image in VMware.
So, I decided to assemble the reads using Amazon EC2. The process was relatively painless once I found an OS image that would boot properly with the instance type I chose (m2.2xlarge. see here for a list).
Here's the steps to follow using a 64-bit Amazon Linux image: