The tables returned by datatiermanagerimpl are returned in full for each request generating a lot of tuple garbage. It would be better to allow filtering (ideally using schema/table name lookups) as we build the result.