Abstract: A method and apparatus are provided for distributing a database join operation across a plurality of processing nodes. The database join operation is performed between a first table having M1 rows and a second table have M2 rows and involves forming M1×M2 combinations of rows from the first and second tables. The method includes allocating across the plurality of processing nodes the M1×M2 combinations of rows from the first and second tables. Each of the M1×M2 combinations is present on one of the nodes, and each node stores combinations corresponding to a subset of the M1 rows from the first table and a subset of the M2 rows from the second table. The method further comprises performing for each processing node the join operation in respect of all combinations of rows allocated to that processing node and collating results across all the processing nodes to complete the join operation.