System and method for executing queries on multi-graphics processing unit systems

A system for performing a database query on a computer system comprising a plurality of graphics processing units, the system comprising:

• a topology identification manager for identifying via an input/output a topology of the computer system, regarding a number of the graphics processing units in the plurality, the topology identification manager for providing at an output the number of graphics processing units in the plurality;
• a query receiver having an input coupled for receiving a request comprising the database query and an express or implied specification of a data source, the query receiver for parsing the request to identify at least one component of the database query, assigning an order to each of the at least one component of the database query parsed, and for optionally providing at an output the specification of the data source and for providing at the query receiver output at least one component of the database query, and, optionally, for each of the at least one component of the database query, the order of the component assigned;
• a partition allocation manager having an input optionally coupled to the query receiver output for receiving the specification of the data source, and to the topology identification manager output for receiving the number of the graphics processing units in the plurality, the partition allocation manager for computing an allocation of a plurality of portions of the data source across each of the plurality of graphics processing units, and for providing at an output the allocation;
• a query manager at each of the graphics processing units, having an input coupled to the partition allocation manager output for receiving at least a portion of the allocation, to the query receiver for receiving the at least one component of the database query and optionally for receiving the order of the component, and coupled for receiving a subset of the plurality of portions of the data source allocated in the allocation to said graphics processing unit, the query manager for applying each of the at least one component of the database query to the subset of the plurality of portions of the data source, optionally responsive to the order optionally received at the query manager input, to produce and provide at an output at least one set of results;
• a join manager having an input coupled to the query receiver output for receiving at least one of the at least one component of the database query and to the query manager output of at least two of the plurality of graphics processing units for receiving the at least one set of results, the join manager for consolidating at least one of the at least one set of results of each of the at least two of the plurality of graphics processing units to produce a consolidated set of results, and for providing at an output the consolidated set of results.

A computer program product comprising a non-transitory computer useable medium having computer readable program code embodied therein for performing a database query on a computer system comprising a plurality of graphics processing units, the computer program product comprising computer readable program code devices configured to cause a computer system to:

• identify a topology of the computer system, regarding a number of the graphics processing units in the plurality;
• receive a request comprising the database query and an express or implied specification of a data source;
• parse the request to identify at least one component of the database query;
• assign an order to each of the at least one component of the query database parsed;
• compute an allocation of a plurality of portions of the data source across each of the plurality of graphics processing units;
• at each of the graphics processing units, apply each of the at least one component of the database query parsed to the portion of the data source allocated to the respective graphics processing unit, responsive to the order assigned, to produce at least one set of results at each of the graphics processing units;
• consolidate at least one of the at least one set of results of each of the plurality of graphics processing units by at least one central processing unit or graphics processing unit of the computer system comprising the plurality of graphics processing units to produce a consolidated set of results;
• and
• output the consolidated set of results.