Mapping query operations in database systems to hardware based query accelerators

Patent metadata

Field	Value
Publication number	US-2016012107-A1
Application number	US-201414328933-A
Country	US
Kind code	A1
Filing date	Jul 11, 2014
Priority date	Jul 11, 2014
Publication date	Jan 14, 2016
Grant date	—

Abstract

Methods and arrangements for mapping a query operation to an accelerator are provided. The method includes receiving, by a processor, a query operation and determining the design logic of the query operation, receiving a configuration of one or more available accelerators and a design logic of each of the one or more available accelerators, and determining if the query operation can be offloaded to one or more of the one or more available accelerators. Based on a determination that the query operation can be offloaded to one or more of the one or more available accelerators, the method also includes creating software structures to interface with a selected accelerator and executing the query operation on the selected accelerator. Based on a determination that the query operation cannot be offloaded to one or more of the one or more available accelerators, the method further includes executing the query operation in software.

First claim

What is claimed is:

1 . A method for mapping a query operation to an accelerator comprising:
receiving, by a processor, a query operation and determining the design logic of the query operation; receiving a configuration of one or more available accelerators and a design logic of each of the one or more available accelerators; determining if the query operation can be offloaded to one or more of the one or more available accelerators; based on a determination that the query operation can be offloaded to one or more of the one or more available accelerators, creating software structures to interface with a selected accelerator and executing the query operation on the selected accelerator; and based on a determination that the query operation cannot be offloaded to one or more of the one or more available accelerators, executing the query operation in software.

2 . The method of claim 1 , further comprising:
based on a determination that the query operation cannot be offloaded to one or more of the one or more available accelerators, determining if the query operation can be restructured; and based on a determination that the query operation can be restructured, restructuring the query operation and creating software structures to interface with a selected accelerator and executing the restructured query operation on the selected accelerator.

3 . The method of claim 2 , further comprising:
based on a determination that the query operation cannot be restructured, executing the query operation in software.

4 . The method of claim 2 , wherein the determination whether the query operation can be restructured is based on the design logic of the query operation.

5 . The method of claim 2 , wherein restructuring the query operation comprises transforming a type of query operation.

6 . The method of claim 2 , wherein restructuring of the query operation comprises partitioning the query operation into the restructured query operation with multiple query operation.

7 . The method of claim 6 , further comprising executing the multiple query operations of the restructured query operation on different accelerators simultaneously.

8 . The method of claim 6 , further comprising executing the multiple query operations of the restructured query operation on the selected accelerators sequentially.

9 . The method of claim 1 , wherein the determination whether the query operation can be offloaded is based on one or more of the design logic of the query operation, the configuration of the one or more available accelerators and the design logic of each of the one or more available accelerators.

10 . A method for mapping a query operation to a hardware implementation on an accelerator comprising:
receiving, by a processor, a database query from an application, wherein the database query comprises one or more query operations; determining a configuration of one or more accelerators and a design logic for each of the of one or more accelerators; identifying a candidate query operation from the one or more query operations which can be offloaded to one of the one or more accelerators; identifying one or more database structures holding information corresponding to the candidate query operation; determining attributes of the one of the one or more accelerators selected to implement the candidate query operation; creating an intermediate data structure representing the candidate query operation based on the attributes; and transforming the intermediate data structure into data structures that will be provided to the one of the one or more accelerators.

11 . The method of claim 10 , further comprising determining if one or more of the query operations not identified as candidate query operation can be restructured.

12 . The method of claim 11 , further comprising based on a determination that one or more of the query operations not identified as candidate query operations can be restructured, restructuring the one or more of the query operations and identifying the restructured query operations as candidate query operations.

13 . The method of claim 11 , wherein the determination whether the one or more of the query operations can be restructured is based on design logic of the one or more the query operation.

14 . The method of claim 12 , wherein restructuring the one or more query operations comprises transforming a type of the one or more of the query operations.

15 . The method of claim 12 , wherein restructuring the one or more query operations comprises partitioning the one or more query operations into multiple query operations.

16 . The method of claim 10 , wherein identifying a candidate query operation is based on one or more of the design logic of the query operation, the configuration of the one or more available accelerators and the design logic of each of the one or more available accelerators.

17 . A method for evaluating performance gains from offloading query processing operations to one or more accelerators, the method comprising:
receiving, by a processor, a workload of queries; receiving design logic of query operations of the workload of queries and a configuration of elements in accelerators; determining a percentage the query operations in the workload of queries that can be offloaded to the accelerators; and estimating the performance gains achievable with the accelerators for the given workload of queries.

Assignees

• IBM

Inventors

• Asaad Sameh
• Dube Parijat
• Iyer Balakrishna R
• Min Hong
• Sukhwani Bharat
• Thoennes Mathew S

Classifications

• G06F16/2453 Primary

  Query optimisation · CPC title

• G06F16/24569

  Query processing with adaptation to specific hardware, e.g. adapted for using GPUs or SSDs · CPC title

• G06F17/30312

  Physics · mapped topic

• G06F17/30442 Primary

  Physics · mapped topic