System and Method for Distributed SQL Join Processing in Shared-Nothing Relational Database Clusters Using Self Directed Data Streams

A method for query processing in a massively parallel processing (MPP) database system, the method comprising:

• receiving, at a data node from one or more other data nodes, a sub-tree of an execution plan tree for a join query associated with a plurality of tables of the MPP database system, the sub-tree corresponds to a next join operator below a top join operator in the execution plan tree;
• processing the sub-tree using partitions of the tables allocated to the data node;
• and
• returning sub-tree processing results to the one or more other data nodes.

An apparatus for query processing in a massively parallel processing (MPP) database system, the apparatus comprising:

• a data node comprising a processor;
• and
• a computer readable storage medium storing programming for execution by the processor, the programming including instructions to:
• receive, at the data node from a coordinator process that runs on the MPP, an execution plan tree for a join query associated with a plurality of tables of the MPP database system;
• process the execution plan tree to obtain join query results using partitions of the data tables allocated to the data node without other partitions of the data tables allocated to other data nodes of the MPP system;
• upon detecting a next join operator below a top join operator in the execution plan tree, forward a sub-tree for the next join operator to the other data nodes;
• receive sub-tree processing results from the other data nodes;
• and
• return to the coordinator process the join query results including the sub-tree processing results.