Pardon my ignorance but isn't dynamo just a document store itself? Most storage engines are designed and optimized for RDBs, so I'd expect inefficiency here unless there's a layer between them.