Unlock instant, AI-driven research and patent intelligence for your innovation.

CPU (Central Processing Unit)-GPU (Graphic Processing Unit) collaborative query processing system and method for RDF (Resource Description Framework) graph data

A processing system and processing method technology, applied in the direction of processor architecture/configuration, etc., can solve problems such as decompression waste, speed up query, etc., to achieve the effect of high degree of parallelism, large number of threads, and high degree of overlap between communication and calculation

Active Publication Date: 2018-10-26
HUAZHONG UNIV OF SCI & TECH
View PDF3 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Some methods based on graph compression express the relationship between the six possible combinations of subjects, predicates, and objects in bitmaps according to the idea of ​​association matrix in graph theory, and use compression algorithms to reduce space, but not only need Create corresponding indexes to speed up, and frequent decompression and compression processes waste a lot of time
Some adopt the vertical partition storage method. When the body contains multiple attribute values, multiple attribute values ​​can be stored efficiently. However, when the number of predicates in the query is large and unknown, an additional index to help the predicate query needs to be added to speed up the query.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • CPU (Central Processing Unit)-GPU (Graphic Processing Unit) collaborative query processing system and method for RDF (Resource Description Framework) graph data
  • CPU (Central Processing Unit)-GPU (Graphic Processing Unit) collaborative query processing system and method for RDF (Resource Description Framework) graph data
  • CPU (Central Processing Unit)-GPU (Graphic Processing Unit) collaborative query processing system and method for RDF (Resource Description Framework) graph data

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0038] Such as figure 1 The CPU-GPU cooperative query processing system for RDF graph data shown includes a central processing unit 1, a graphics processor 2, a user interface module 3, a data storage module 4, and an algorithm database module 5.

[0039] Preferably, the user interface module 3 includes an RDF data import interface, a Sparql query interface, an entity dump interface and an RDF data dump interface. Among them, the RDF data import interface is used to import RDF graph data that needs to be processed in a wired connection with an external device. The external device may be, for example, a mobile terminal, a computer, a cloud server, etc. The Sparql query interface is used to input query statements. The query sentence can be input by connecting the Sparql query interface with devices such as keyboards and touch screens. The RDF data dump interface is used to back up and store RDF data at a specific time to prevent errors in RDF data. Through the RDF data dump inte...

Embodiment 2

[0049] This embodiment is a further supplement to Embodiment 1, and the repeated content will not be repeated.

[0050] Such as Figure 4 As shown, a method of selecting a triple pattern for each of the common variables as a representative mode, and dividing based on the data corresponding to each of the representative modes to obtain several data blocks, specifically includes the following steps:

[0051] S1: Sort the triplet patterns based on the number and / or selectivity of the public variables and set the sorted N triplet patterns as an array Pattern[N];

[0052] S2: Set the counter k=0, and set the representative mode of the selected public variable to Delegate[var], where Delegate[var] is empty, which means that the public variable var has not yet selected the representative mode;

[0053] S3: Read the k-th triplet pattern and set it to P;

[0054] S4: If there is only one public variable var in P and Delegate[var] is empty, go to step S5; if there is only one public variable var ...

Embodiment 3

[0065] This embodiment is a further supplement to Embodiment 1 and Embodiment 2, and the repeated content will not be repeated.

[0066] A method for generating Join tasks and transferring data blocks to GPU memory, specifically including the following steps:

[0067] A1: Set the counter m=0, and set the data block corresponding to the representative mode to CHUNK[P][N] to indicate that the mode P corresponds to a total of N data blocks;

[0068] A2: Under the condition that the representative mode has two public variables, set the two public variables to x and y respectively and then go to step A3; if the representative mode does not have two public variables, go to step A4;

[0069] A3: Read the nth data block CHUNK[P][n] of mode P, create the join task according to the value of the public variable y in CHUNK[P][n], and then proceed to step A5;

[0070] A4: Read the nth data block CHUNK[P][n] of mode P, and according to the interval [x of the public variable x in CHUNK[P][n] min ,x ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a CPU (Central Processing Unit)-GPU (Graphic Processing Unit) collaborative query processing system and method for RDF (Resource Description Framework) graph data. The method comprises the following steps that: on the basis of a query statement submitted from a user, carrying out analysis to obtain first information, wherein the first information comprises triple patterns,public variables and projection variables; aiming at each public variable to select one triple pattern as a representative pattern, and carrying out division on the basis of data corresponding to eachrepresentative pattern to obtain a plurality of data blocks; after the plurality of data blocks are read in sequence to generate a connection task among different data blocks, transmitting the plurality of data blocks to a GPU video memory in sequence; under a situation that the CPU detects that the data blocks on which the connection task depends are all transmitted to the GPU video memory, executing the connection task by the GPU, and generating a corresponding intermediate result; and under a condition that all connection tasks among different data blocks finish being executed and / or alldata blocks are transmitted to the GPU video memory, carrying out full connection on different intermediate results, and carrying out collection according to a way of a projection variable sequence toobtain a final result. The method has a high parallel degree and a high query speed.

Description

Technical field [0001] The present invention relates to the technical field of graph data processing, in particular to a CPU-GPU collaborative query processing system and method for RDF graph data. Background technique [0002] Resource Description Framework (RDF) has become one of the standard formats for data exchange. It provides a unified standard for describing various resources on the Web. Formally, RDF can be represented by a triple: subject, predicate, and object. [0003] SPARQL is a SQL-like query language recommended by W3C standards for querying RDF data. Formally, a SPARQL query can usually be expressed as SELECT? x? y? z WHERE{P1, P2...Pn}. among them? x? y? z is the projection variable, and P1, P2...Pn are the triple pattern. [0004] With the emergence and development of general graphics processing units (GPGPU), a new research trend is to use GPUs in database systems to accelerate query processing. Although GPU has powerful computing power, there are still...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06T1/20
CPCG06T1/20
Inventor 袁平鹏金海王磊
Owner HUAZHONG UNIV OF SCI & TECH