Amazon SimpleDB has just been released as a Beta, its like S3 but for databases, allowing structured queries.

What this means is that you don't need to worry about database clustering or possibly backup (although I have never gotten a decent answer to the question of whether you still need to backup S3 data).

This potentially provides a beautiful solution for startups - EC2 as application servers, SimpleDB for structured data and S3 for binaries.

It is not clear whether SimpleDB can be used, somehow for efficient full text search.

If only S3 had a front end like Squid to enable automatic cache on demand for binaries, and there were a better front end to instantiate, configure and manage EC2 instances, then Amazon would be the default choice for most startups.