You've been downvoted without an explanation as to why that's not a useful suggestion.
The problem is that if you take a "just restart it if it crashes" approach, the service will be very unreliable, and you're not allowing for analysing what went wrong. In fact, taking this approach, Amazon's EC2 images are a really bad option, because you lose any chance of gathering diagnostic info as soon as you switch them off - all the data vanishes at that point.
Secondly, if you're going for a long-running VM, Amazon's machines are very expensive for what they are; I pay roughly a third of what Amazon charge for mine. If you're using them as base load rather than on-demand peak smoothing, you're doing it wrong.
The problem is that if you take a "just restart it if it crashes" approach, the service will be very unreliable, and you're not allowing for analysing what went wrong. In fact, taking this approach, Amazon's EC2 images are a really bad option, because you lose any chance of gathering diagnostic info as soon as you switch them off - all the data vanishes at that point.
Secondly, if you're going for a long-running VM, Amazon's machines are very expensive for what they are; I pay roughly a third of what Amazon charge for mine. If you're using them as base load rather than on-demand peak smoothing, you're doing it wrong.