MySQL's innodb has a clustered index. On DB2, SqlServer, Oracle, etc, they are configurable. PostgreSQL is more like MySQL myisam in that the data and index are

MySQL's innodb has a clustered index. On DB2, SqlServer, Oracle, etc, they are configurable. PostgreSQL is more like MySQL myisam in that the data and index are separate. In this case, the random nature only screws up the index (assuming btree), not the data. It can be a bigger issue if you proxy sql queries and shard to 1000 DB servers and you thus rely on part of the query to indicate which shard it is on. You can do this with a type of uuid that embeds the shard, or you have to include additional parts in the query beyond the PK to get the row (it could be a column that the sharing is based on, or a comment). And yes, careful with joins from an a root object's table that joins to other tables that use uuids unless you can guarantee that they are in the same node.