When using Spider it would be useful to take advantage of data from several tables being colocated on the same set of Spider data nodes. This enforces that the column used for colocating data is also part of the primary key in all the tables, as this is a limitation of partitioning. Take the example of orders/customers. If we shard on customer_id, which is not unreasonable, then we have to make customer_id a part of the primary key in the orders table for this to work. It is understood that the limitation in partitioning to enforce the partitioning column to be part of the primary key is well founded and that this is there to ensure the uniqueness across partitions of the primary key. That said, in some cases it would be useful to lift this restriction, on a table by table basis, making the user responsible for maintaining cross partition uniqueness.