US 9,311,359 B2Granted

Join operation partitioning

AssigneeInternational Business Machines Corporation

InventorsGopi K. Attaluri, Vijayshankar Raman

Life of this patent

Estimating the term…

Loading patent…

Abstract

Partitioned join operations are performed between a first database object and a second database object by determining an agent group for an agent in response to the agent receiving rows of the second database object to process; partitioning the rows to determine a target hash table for each row and adding the partitioned rows to work to be performed by the agent group; and distributing the work for the group to agents of the group by assigning to a single agent all the rows associated with a particular hash table to perform a join operation on the assigned rows. Each partition is assigned a first counter value indicating an upper bound of a task id range that is most recently assigned to an agent in the agent group for processing, and a second counter value indicating the highest task id that has been processed for that partition.

Claims18

Dependent claims

An apparatus for performing partitioned join operations between a first database object and a second database object, comprising:

a processor including:
a determining module configured to:
determine an agent group for an agent in response to the agent receiving rows of the second database object to process, wherein the rows of the first database object are partitioned into a plurality of hash tables;
and
determine an amount of the received rows of the second database object and adding the agent to the agent group in response to the determined amount of the received rows of the second database object exceeding a predetermined quantity of rows;
a partitioning module configured to partition the received rows of the second database object to determine a target hash table from among the plurality of hash tables for each received row and adding the partitioned rows to a pool of rows representing work to be performed by a plurality of agents of the agent group;
and
a distributing module configured to distribute the work for the group to the plurality of agents of the agent group by assigning to a single agent for each of the plurality of hash tables all the rows of the pool associated with that particular hash table to perform a join operation on the assigned rows.

5 more elements

A computer program product comprising:

a computer readable storage device encoded with instructions that, when executed by a processor, cause the processor to perform partitioned join operations between a first database object and a second database object by performing the functions of:
determining an agent group for an agent in response to the agent receiving rows of the second database object to process, wherein the rows of the first database object are partitioned into a plurality of hash tables;
determining an amount of the received rows of the second database object and adding the agent to the agent group in response to the determined amount of the received rows of the second database object exceeding a predetermined quantity of rows;
partitioning the received rows of the second database object to determine a target hash table from among the plurality of hash tables for each received row and adding the partitioned rows to a pool of rows representing work to be performed by a plurality of agents of the agent group;
and
distributing the work for the group to the plurality of agents of the agent group by assigning to a single agent for each of the plurality of hash tables all the rows of the pool associated with that particular hash table to perform a join operation on the assigned rows.

3 more elements

Dependencies are inferred from the claim text.