Presto though, being the crack SQL engine that it is, doesn't use that last part (which is the bulk of Hive): it just queries the Metastore service for metadata, then goes and does its own query processing fetching the data directly from the Hadoop file system when it needs it....