Here's an interesting treatment of the scaling issue which cropped up on HN a few days ago:
Thank you. That was an interesting and insightful observation [http://news.ycombinator.com/item?id=859617]. His argument for reliability made scaling seem almost moot. You rarely see someone toss around the idea of 'reliability' when (they are) discussing whatever they think 'the cloud' means. Nevertheless, it was a refreshing reminder that newer (better) technology is not always about optimization, its about mitigating risk by building a sense of predictability into your solution (application).
Please correct if I'm misunderstanding anything or everything. Also would someone care to elaborate more (educate me) on scalability? Is this same sort of scalability Amazon's EC2 servers are optimized for?