From their whitepapers you get the sense that much of the point of Amazon Dynamo, Google Protocol Buffers and BigTable, YAHOO PNUTS and UDB, Facebook Thrift obj

From their whitepapers you get the sense that much of the point of Amazon Dynamo, Google Protocol Buffers and BigTable, YAHOO PNUTS and UDB, Facebook Thrift objects, etc is to release big systems from the tyranny of the SQL schema. When you can run multiple versions of code against the same dataset, upgrading the application layer one bit at a time becomes much easier.