Boomerang join: a network efficient, late-materialized, distributed join technique

One or more non-transitory computer-readable media storing instructions which, when executed by one or more processors, cause a database management system performing a join operation, over first and second database objects that are horizontally partitioned across a plurality of nodes, to produce a join result materialization comprising:

• assigning a respective local join node, from the plurality of nodes, to each respective row of a plurality of rows in the first and second database objects;
• transmitting, over a network to each local join node, a respective set of join tuples, wherein each join tuple of the respective set of join tuples represents a respective row in a set of rows, of the plurality of rows, to which the local join node was assigned;
• wherein each join tuple comprises a join key from the represented row and a row identifier for the represented row;
• each local join node, of the plurality of nodes, performing a local join operation over the respective set of join tuples to produce a respective set of join result tuples;
• wherein each join result tuple comprises a first row identifier from the first database object and a second row identifier from the second database object that identify respective rows that are matched by a local join operation;
• each node of the plurality of nodes receiving a respective materialization set of join result tuples sent from one or more other nodes of the plurality of nodes;
• and
• each node of the plurality of nodes materializing a respective portion of the join result materialization based, at least in part, on the respective materialization set of join result tuples received by said node.