There was a coupon site that I moved to AWS because all the customers were in EU, but our servers were in Los Angeles. I implemented it in two t1.micro servers,

There was a coupon site that I moved to AWS because all the customers were in EU, but our servers were in Los Angeles. I implemented it in two t1.micro servers, db+nginx. Every night at 10PM in LA the first week I'd get a slew of notifications from monitoring that the CPUs were pegged out. I think I put squid or something on the nginx instance so that the most common stuff came out of the proxy instead of having to be php-unblinkered. However another thing that could have been done is to use a slightly larger instance for the DB to get more ram for cache, and then use a single front end until 9PM, spin another up, and spin one down at like 1AM. This is obviously a trivial example, but lots of sites and apps get everyone refreshing at once maybe once or twice a day, and that's when scaling helps, at least from a "costs $X per hour, so don't use it unless we need it". Loads were under 20% the other 23 hours during the day, averaged.