I'm not quite sure I understand what advantages are lost by sharding -- would you mind elaborating? As far as being able to write parallel queries, this is per

I'm not quite sure I understand what advantages are lost by sharding -- would you mind elaborating? As far as being able to write parallel queries, this is perhaps true, but you have to manage the parallelization, synchronization, and all other manner of nastiness. For data that doesn't require frequent updates (like Google), this may be a better route, but for systems where consistent data is important, I would much rather rely on years' worth of RDBMS research. It's not a matter of a lack of creativity; rather, I would prefer to spend my time solving problems as they are related to my application, not duplicating the bits of an RDBMS that aren't present in SimpleDB. I find point (c) to be a bit surprising -- SQL is based around the concepts of set theory, and works in a fairly straightforward fashion. While there are differences between implementations, you certainly don't have to 're-learn everything' when you switch vendors, and the basic tasks of database management are not terribly complicated. And yes, I regularly bounce between about three different SQL implementations (Postgres, MySQL, and SQLlite). Spent some time with Sybase as well, and I am not a DBA. There are differences, definitely, but it never took me more than a day or two to get up to speed, and I am not a genius by any stretch of the imagination.