Don’s write-up of how they use Amazon’s compute cluster service EC2 is an interesting read if you’re into those types of things. :) Using EC2, they’re able to change the server resources they’re using on the fly to match with the current demand on the system, without having to over-buy extra hardware for peak times.