To speed up processing of a query by parallelizing the execution of each individual operation, such as sort, select, project and join.
Intra-operation parallelism is achieved by executing an operation on several nodes of a multiprocessor machine. This requires that the operands have been previously partitioned across the nodes.
The set of nodes where a relation is stored is called its home. The home of an operation is the set of nodes where it is executed and it must be the home of its operands in order for the operation to access its operands. For binary operations such as join, this might imply repartitioning one of the operands.
The optimizer might even sometimes find that repartitioning both the operands is useful. Parallel optimization to exploit infra-operation parallelism can make use of some of the techniques devised for distributed databases.
Since relational operations work relations containing large sets of tuples, we can parallelize the operations by executing them in parallel on different subsets to the relations.