Method and device for inquiring data in database

Abstract

The application discloses a method and a device for inquiring data in a database. The method comprises the following steps of receiving an inquiry instruction, determining the state of each block index in a block index collection which is preliminarily stored in a video memory of a GPU (Graphics Processing Unit), starting one or more GPU processes in the case that each block index in the video memory of the GPU is in an asynchronous state, and according to inquiry conditions, filtering the block indexes in the video memory of the GPU to obtain a first block index inquiry result; determining a final data inquiry result according to the first block index inquiry result. According to the method, a CPU (Central Processing Unit) preliminarily generates the block index collection corresponding to the database; as the volume of data of the block index collection is smaller than the volume of original data or partitioned data, all data of the block index collection can be copied and stored in a global memory of the GPU; when the block indexes in the global memory of the GPU are all in the asynchronous state, the inquiry is implemented by directly using the GPU processes; therefore the course of copying the partitioned data from the memory multiple times in the prior art is avoided; the inquiry time is shortened; the inquiry efficiency is improved.

Description

technical field [0001] The present invention relates to the technical field of databases, and more specifically relates to a data query method and device in a database. Background technique [0002] In analytical databases, in order to speed up data processing, multi-process / multi-thread concurrency technology is usually used. Although the number of CPU cores is increasing, its growth rate can no longer meet the needs of data processing. The general-purpose graphics processing unit (Graphics Processing Units, referred to as GPU) provides higher concurrent processing capabilities and greater memory bandwidth, and is suitable for highly concurrent, computationally intensive, and simple-control computing applications. [0003] In the process of using the GPU to execute data query instructions in the prior art, since the size of the GPU's Global Memory (global memory, also called video memory) is limited, it is necessary to match the query condition with the query condition aft...

AI Technical Summary

Problems solved by technology

[0003] In the process of using the GPU to execute data query instructions in the prior art, since the size of the GPU's Global Memory (global memory, also called video memory) is limited, it is necessary to match the query condition with the query condition after starting the GPU thread for each short query. The corresponding partition data is imported into the global memory of the GPU. When there are multiple query conditions, the repeated copy process will be repeated multiple times, so the query process is very cumbersome, and the delay in copying data from the memory to the global memory of the GPU Longer, therefore, greatly reduces the efficiency of using the GPU to execute query instructions

Images