CN113886652B - Memory-priority multimode graph data storage and calculation method and system - Google Patents
Memory-priority multimode graph data storage and calculation method and system Download PDFInfo
- Publication number
- CN113886652B CN113886652B CN202111175266.6A CN202111175266A CN113886652B CN 113886652 B CN113886652 B CN 113886652B CN 202111175266 A CN202111175266 A CN 202111175266A CN 113886652 B CN113886652 B CN 113886652B
- Authority
- CN
- China
- Prior art keywords
- graph data
- index
- edge
- data
- memory
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004364 calculation method Methods 0.000 title claims abstract description 49
- 238000013500 data storage Methods 0.000 title claims abstract description 23
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000000034 method Methods 0.000 claims abstract description 25
- 238000013499 data model Methods 0.000 claims abstract description 20
- 230000008569 process Effects 0.000 claims abstract description 11
- 230000006870 function Effects 0.000 claims description 13
- 238000010276 construction Methods 0.000 claims description 6
- 230000000875 corresponding effect Effects 0.000 description 14
- 238000010586 diagram Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 3
- 238000003491 array Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 229910052754 neon Inorganic materials 0.000 description 1
- GKAOGPIIYCISHV-UHFFFAOYSA-N neon atom Chemical compound [Ne] GKAOGPIIYCISHV-UHFFFAOYSA-N 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000008093 supporting effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/901—Indexing; Data structures therefor; Storage structures
- G06F16/9024—Graphs; Linked lists
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/901—Indexing; Data structures therefor; Storage structures
- G06F16/9014—Indexing; Data structures therefor; Storage structures hash tables
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline or look ahead
- G06F9/3885—Concurrent instruction execution, e.g. pipeline or look ahead using a plurality of independent parallel functional units
- G06F9/3887—Concurrent instruction execution, e.g. pipeline or look ahead using a plurality of independent parallel functional units controlled by a single instruction for multiple data lanes [SIMD]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a memory-prioritized multimode graph data storage and calculation method and system, which are applied to the technical field of data storage and calculation, and the method comprises the following steps: vertex, edge and attribute of the graph data are configured by adopting a json file to construct a graph data model; aiming at the graph data based on the graph data model, processing the vertex data of the graph data by adopting a Hash algorithm to construct a point index, and constructing an edge index of the graph data by adopting an adjacent linked list; the point index and the edge index are stored continuously in sequence and are preferentially loaded into a memory in the initialization process; when a processing instruction for graph data is received, vectorization calculation is performed on the graph data using a SIMD instruction set. By the technical scheme, the aim of efficiently accessing data can be fulfilled, the use efficiency of the memory space is fully exerted, and the data processing speed is improved by adopting a vectorization parallel execution mode.
Description
Technical Field
The invention relates to the technical field of data storage and calculation, in particular to a memory-preferred multimode graph data storage and calculation method and a memory-preferred multimode graph data storage and calculation system.
Background
The graph is one of the main fields of computer science research, the graph can effectively solve the problems existing at present, and the storage and calculation of graph data have great potential in the development of modern application programs. The graph is based on the model expression of object incidence relation, and the knowledge is structurally stored in a mode of bordering the entity and the relation point, so the graph has natural interpretability and is highly advocated by the academic world and the industrial world. Graph data refers to data stored using a graph structure. Currently, a large amount of graph data is accumulated in the fields of communication, internet, electronic commerce, social network, internet of things, etc., and the scale thereof is huge and is increasing.
The correlation of data can generate significant business value in today's age. Whether we want to understand the relationships between users in online social networks, or between users and goods in e-commerce, or credit relationships in financial networks, the ability to understand and analyze large amounts of highly correlated data will become the core competency of the enterprise. Graph processing techniques play an important role therein.
The storage and computation of graph data is related to the technology that solves a big trend of the macro business of today: complex, dynamic relationships in the highly connected data are exploited to generate insights and competitive advantages. Graph data is the best way to represent and query connection data versus relational data. However, compared with other data technologies, the research of the related technology of the graph data is still in a starting stage, the related system is still lagged, and the method has a larger improvement space in the aspects of storage cost, calculation performance and the like and can play a larger application value.
Disclosure of Invention
In order to solve the problems, the invention provides a memory-first multi-mode graph data storage and calculation method and system, a graph data model is constructed through a configuration file, and graph data are stored in a specific index mode to improve the support effect of a calculation scene, achieve the aim of high-efficiency data access and fully exert the use efficiency of a memory space, and the query and calculation of the graph data adopt a vectorization parallel execution mode to improve the data processing speed.
In order to achieve the above object, the present invention provides a memory-prioritized multimodal map data storage and calculation method, including:
adopting json (JavaScript Object Notation) files to configure the vertexes, edges and attributes of the graph data so as to construct a graph data model;
aiming at the graph data based on the graph data model, processing the vertex data of the graph data by adopting a Hash algorithm to construct a point index, and constructing an edge index of the graph data by adopting an adjacency linked list;
the point index and the edge index are stored continuously in sequence and are preferentially loaded into a memory in an initialization process;
when a processing instruction for the graph data is received, vectorizing computations are performed on the graph data using a SIMD instruction set.
In the foregoing technical solution, preferably, in the graph data model, the associated data is used as the vertices, a connection line between the vertices is used as the edge, the edge is used to represent an association relationship between the vertices corresponding to the two ends, and the vertices and the edge have corresponding attributes respectively.
In the foregoing technical solution, preferably, in the hash algorithm, a hash function is used to convert vertex key values of the graph data into indexes of arrays, the indexes are used as point indexes, and a zipper method is used to process hash collisions.
In the foregoing technical solution, preferably, the constructing the edge index of the graph data by using the adjacency linked list specifically includes:
representing the logical sequence of edge indexes by pointer link order in the adjacency linked list, wherein the adjacency linked list is composed of a series of nodes dynamically generated in runtime, each node comprises a data field and a pointer field, the data field stores data elements, and the pointer field stores pointers pointing to the next node;
and the adjacency linked list generates an index of a reverse edge while constructing the index of the forward edge, wherein the index of the reverse edge is realized by combining a hash value with the adjacency linked list.
In the foregoing technical solution, preferably, the performing vectorization calculation on the graph data by using a SIMD instruction set specifically includes:
and according to the received processing instruction aiming at the graph data, calling corresponding functions in the SIMD instruction set in parallel to realize vectorization calculation processing of the graph data.
The invention also provides a memory-prioritized multimode graph data storage and calculation system, which applies the memory-prioritized multimode graph data storage and calculation method disclosed by any one of the above technical solutions, and comprises the following steps:
the model building module is used for adopting the json file to configure the vertex, the edge and the attribute of the graph data so as to build a graph data model;
the index construction module is used for processing the vertex data of the graph data by adopting a Hash algorithm aiming at the graph data based on the graph data model so as to construct a point index, and constructing an edge index of the graph data by adopting an adjacent linked list;
the index storage module is used for continuously storing the point index and the edge index in sequence and preferentially loading the point index and the edge index into a memory in an initialization process;
and the graph data calculation module is used for performing vectorization calculation on the graph data by utilizing a SIMD instruction set when a processing instruction aiming at the graph data is received.
In the foregoing technical solution, preferably, in the model building module, the associated data is used as the vertex, a connection line between the vertices is used as the edge, the edge represents an association relationship between the vertices corresponding to two ends, and the vertex and the edge have corresponding attributes respectively.
In the above technical solution, preferably, in the index construction module, a hash function is used to convert vertex key values of the graph data into an index of an array, the index is used as a point index, and a zipper method is used to process hash conflicts.
In the above technical solution, preferably, in the index building module, the logical order of the edge indexes is represented by the link order of the pointers in the adjacency linked list, the adjacency linked list is composed of a series of nodes dynamically generated at runtime, each node includes a data field and a pointer field, the data field stores data elements, and the pointer field stores pointers pointing to next nodes;
and generating a corresponding reverse edge index while constructing the forward edge index, wherein the reverse edge index is realized by combining a hash value and an adjacent linked list.
In the foregoing technical solution, preferably, the graph data calculation module is specifically configured to:
and according to the received processing instruction aiming at the graph data, calling corresponding functions in the SIMD instruction set in parallel to realize vectorization calculation processing of the graph data.
Compared with the prior art, the invention has the beneficial effects that: the graph data model is built through the configuration file, the graph data are stored in a specific index mode, the supporting effect of a calculation scene is improved, the purpose of high-efficiency data access is achieved, the use efficiency of a memory space is fully exerted, a vectorization parallel execution mode is adopted for query and calculation of the graph data, and the data processing speed is improved.
Drawings
FIG. 1 is a flow chart illustrating a method for storing and computing memory-prioritized multimodal image data according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a storage structure of a point index according to an embodiment of the present disclosure;
FIG. 3 is a schematic diagram of a storage structure of an edge index according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating a vectorization calculation using a SIMD instruction set according to an embodiment of the present invention;
FIG. 5 is a block diagram of a memory-first multimodal image data storage and computing system, according to an embodiment of the present invention.
In the drawings, the correspondence between each component and the reference numeral is:
11. the system comprises a model building module, 12 an index building module, 13 an index storage module, 14 a graph data calculation module.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
The invention is described in further detail below with reference to the attached drawing figures:
as shown in fig. 1, the method for storing and calculating memory-prioritized multi-mode graph data according to the present invention includes:
vertex, edge and attribute of the graph data are configured by adopting a json file to construct a graph data model;
aiming at the graph data based on the graph data model, processing the vertex data of the graph data by adopting a Hash algorithm to construct a point index, and constructing an edge index of the graph data by adopting an adjacent linked list;
the point index and the edge index are stored continuously in sequence and are preferentially loaded into a memory in the initialization process;
when a processing instruction for graph data is received, vectorization calculations are performed on the graph data using the SIMD instruction set.
In the embodiment, the graph data model is constructed through the configuration file, and the graph data is stored in a specific index mode to improve the support effect of a calculation scene, achieve the aim of high-efficiency data access, give full play to the use efficiency of a memory space, and improve the data processing speed by adopting a vectorization parallel execution mode for query and calculation of the graph data.
Specifically, the graph data model construction part stores the entity of the associated data as a vertex (vertex) and the relationship as an edge (edge) on the basis of a graph theory, and the nodes and the edges can also have attributes (properties). Preferably, the attribute graph model is composed of vertices representing entities or instances, edges being lines that will connect the vertices representing associations between the vertices, and their attributes. The vertex and the edge respectively have corresponding attributes, and each vertex can be set to determine a unique primary _ key corresponding to a group of attributes. Each edge needs to specify src _ vertex and dest _ vertex, and the two points are connected in series.
And configuring by using a json file, and respectively storing points, edges, attributes and incidence relations thereof by using the json file, so that the graph structure is expressed from a model level. Specifically, the graph relation logic is organized through directories, each graph is a single file directory, the directory is divided into point configuration and edge configuration, names of points and edges are named, and file contents are in a json format and are as follows:
and (4) linking up to completely express core concept models such as graphs, points, edges, attributes and the like.
In the above embodiment, preferably, in the hash algorithm, a hash function is used to convert vertex key values of the graph data into indexes of arrays as point indexes, and a zipper method is used to process hash collisions.
Specifically, using a hash table to store a point index, a chain address method is used to resolve the situation where a conflict is encountered. As shown in fig. 2, for example, there is a set of data {1,12,26,337,353 26,337,353 … }, and the hash algorithm employs h (key) mod 16, where the hash value f (1) of the first data 1 is 1 and is inserted after node 1, the hash value f (12) of the second data 12 is 12 and is inserted after node 12, the hash value f (26) of the third data 26 is 10 and is inserted after node 10, the 4 th data 337 is calculated to have a hash value of 1, a collision is encountered, but it is still only necessary to find the last link point insertion of the node 1, and the 5 th data 353 and the following data are the same as the above processing method.
In the foregoing embodiment, preferably, the constructing the edge index of the graph data by using the adjacency linked list specifically includes:
representing the logic sequence of the edge index by the pointer link sequence in an adjacent linked list, wherein the adjacent linked list consists of a series of nodes dynamically generated in operation, each node comprises a data field and a pointer field, the data field stores data elements, and the pointer field stores a pointer pointing to the next node;
the adjacency linked list generates an index of a reverse edge while constructing a forward edge index, and the index of the reverse edge is realized by combining a hash value and the adjacency linked list.
Specifically, unlike the sequence table, the linked list does not limit the physical storage state of the data, in other words, the physical storage location of the data elements stored using the linked list is random.
A complete linked list needs to be composed of the following parts:
head pointer: a common pointer is characterized by always pointing to the position of the first node of the linked list. It is obvious that the head pointer is used to indicate the location of the linked list, which facilitates later finding the linked list and using the data in the list.
And (3) node: the nodes in the linked list are further subdivided into head nodes, head nodes and other nodes.
A head node: it is a null node that does not store any data, usually as the first node of the linked list. The head node is not necessary for the linked list, but serves to facilitate resolution of some practical problems.
A first node: the first node in the linked list that stores data is called the head node because of the head node (i.e., the null node). The first node is only a name for storing the data node in the first linked list, and has no practical significance.
And other nodes: other nodes in the linked list.
Thus, a fully linked list structure storing {1,2,3} is shown in FIG. 3.
In the foregoing embodiment, preferably, the performing vectorization calculation on graph data by using a SIMD instruction set specifically includes:
and according to the received processing instruction aiming at the graph data, calling corresponding functions in the SIMD instruction set in parallel to realize vectorization calculation processing of the graph data.
In SIMD, i.e. Single Instruction, Multiple Data, one Instruction operates on Multiple Data.
The SIMD idea is implemented by using different types of instructions under different platform architectures:
x86: SSE instruction
ARM: NEON instruction
MIPS：MSA。
As shown in FIG. 4, the SIMD implementation method can use intrinsic provided by a compiler, and can also be implemented by embedding related header files into assembly codes in C language.
As shown in fig. 5, the present invention further provides a memory-prioritized multi-pattern data storage and calculation system, which applies the memory-prioritized multi-pattern data storage and calculation method disclosed in any of the above embodiments, including:
the model building module 11 is used for adopting json file configuration diagram data to configure the vertexes, edges and attributes of the diagram data so as to build a diagram data model;
the index construction module 12 is configured to, for the graph data based on the graph data model, process vertex data of the graph data by using a hash algorithm to construct a point index, and construct an edge index of the graph data by using an adjacency linked list;
the index storage module 13 is used for continuously storing the point index and the edge index in sequence and preferentially loading the point index and the edge index into the memory in the initialization process;
and the graph data calculation module 14 is used for performing vectorization calculation on the graph data by using the SIMD instruction set when receiving the processing instruction aiming at the graph data.
In the above embodiment, preferably, in the model building module 11, the associated data is used as a vertex, a connecting line between the vertices is used as an edge, the edge represents an association relationship between the vertices corresponding to the two ends, and the vertex and the edge have corresponding attributes respectively.
In the above embodiment, preferably, in the index building module 12, the vertex key value of the graph data is converted into an index of an array by using a hash function, the index is used as a point index, and the hash collision is handled by using a zipper method.
In the above embodiment, preferably, in the index building module 12, the logical order of the edge indexes is represented by the order of pointer links in the adjacency linked list, the adjacency linked list is composed of a series of nodes dynamically generated in runtime, each node includes a data field and a pointer field, a data element is stored in the data field, and a pointer pointing to a next node is stored in the pointer field;
and generating a corresponding reverse edge index while constructing the forward edge index, wherein the reverse edge index is realized by combining a hash value and an adjacent linked list.
In the foregoing embodiment, preferably, the graph data calculating module 14 is specifically configured to:
and according to the received processing instruction aiming at the graph data, calling corresponding functions in the SIMD instruction set in parallel to realize vectorization calculation processing of the graph data.
According to the memory-prioritized multi-mode graph data storage and calculation system disclosed in the above embodiment, functions implemented by the modules correspond to implementation methods of the steps in the memory-prioritized multi-mode graph data storage and calculation method disclosed in the above embodiment, which are not described again.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. A memory-first multimode graph data storage and calculation method is characterized by comprising the following steps:
vertex, edge and attribute of the graph data are configured by adopting a json file to construct a graph data model;
aiming at the graph data based on the graph data model, processing the vertex data of the graph data by adopting a hash algorithm, converting the vertex key value of the graph data into an array index by adopting a hash function to serve as a point index, and processing hash collision by adopting a zipper method to construct and form the point index;
adopting an adjacency linked list to construct an edge index of the graph data, specifically comprising:
representing the logical sequence of edge indexes by pointer link order in the adjacency linked list, wherein the adjacency linked list is composed of a series of nodes dynamically generated in runtime, each node comprises a data field and a pointer field, the data field stores data elements, and the pointer field stores pointers pointing to the next node;
the adjacency linked list generates an index of a reverse edge while constructing the index of the forward edge, and the index of the reverse edge is realized by combining a hash value with the adjacency linked list;
the point index and the edge index are stored continuously in sequence and are preferentially loaded into a memory in an initialization process;
when a processing instruction for the graph data is received, vectorizing computations are performed on the graph data using a SIMD instruction set.
2. The memory-first multimode graph data storage and calculation method according to claim 1, wherein in the graph data model, associated data is used as the vertices, a connecting line between the vertices is used as the edge, the edge is used to represent an association relationship between the vertices corresponding to both ends, and the vertices and the edge have corresponding attributes respectively.
3. The memory-first multi-mode graph data storage and calculation method according to claim 1, wherein the vectorizing calculation of the graph data by using a SIMD instruction set specifically comprises:
and according to the received processing instruction aiming at the graph data, calling corresponding functions in the SIMD instruction set in parallel to realize vectorization calculation processing of the graph data.
4. A memory-prioritized multimodal map data storage and calculation system, applying the memory-prioritized multimodal map data storage and calculation method according to any one of claims 1 to 3, comprising:
the model building module is used for adopting the json file to configure the vertex, the edge and the attribute of the graph data so as to build a graph data model;
the index construction module is used for processing the vertex data of the graph data by adopting a hash algorithm aiming at the graph data based on the graph data model, converting the vertex key value of the graph data into an array index by adopting a hash function to serve as a point index, and processing hash collision by adopting a zipper method to construct and form the point index;
the index building module is further configured to build an edge index of the graph data by using an adjacency linked list, and specifically includes:
representing the logical sequence of edge indexes by the link order of pointers in the adjacency linked list, wherein the adjacency linked list is formed by a series of nodes dynamically generated in runtime, each node comprises a data field and a pointer field, data elements are stored in the data field, and pointers pointing to the next node are stored in the pointer field;
generating a corresponding reverse edge index while constructing the forward edge index, wherein the reverse edge index is realized by combining a hash value and an adjacent linked list;
the index storage module is used for continuously storing the point index and the edge index in sequence and preferentially loading the point index and the edge index into a memory in an initialization process;
and the graph data calculation module is used for performing vectorization calculation on the graph data by utilizing a SIMD instruction set when a processing instruction aiming at the graph data is received.
5. The memory-first multimodal map data storage and calculation system according to claim 4, wherein in the model construction module, the associated data is used as the vertex, the connecting line between the vertices is used as the edge, the edge represents the association relationship between the vertices at the two corresponding ends, and the vertex and the edge have corresponding attributes respectively;
and generating a corresponding reverse edge index while constructing the forward edge index, wherein the reverse edge index is realized by combining a hash value and an adjacent linked list.
6. The memory-first multi-mode graph data storage and computation system of claim 4, wherein the graph data computation module is specifically configured to:
and according to the received processing instruction aiming at the graph data, calling corresponding functions in the SIMD instruction set in parallel to realize vectorization calculation processing of the graph data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111175266.6A CN113886652B (en) | 2021-10-09 | 2021-10-09 | Memory-priority multimode graph data storage and calculation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111175266.6A CN113886652B (en) | 2021-10-09 | 2021-10-09 | Memory-priority multimode graph data storage and calculation method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113886652A CN113886652A (en) | 2022-01-04 |
| CN113886652B true CN113886652B (en) | 2022-06-17 |
Family
ID=79005762
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111175266.6A Active CN113886652B (en) | 2021-10-09 | 2021-10-09 | Memory-priority multimode graph data storage and calculation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113886652B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115309947B (en) * | 2022-08-15 | 2023-03-21 | 北京欧拉认知智能科技有限公司 | Method and system for realizing online analysis engine based on graph |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544291A (en) * | 2013-10-29 | 2014-01-29 | 东北林业大学 | Mobile object continuous k-nearest neighbor (CKNN) query method based on road based road networks tree (RRN-Tree) in road network |
| CN103617276A (en) * | 2013-12-09 | 2014-03-05 | 南京大学 | Method for storing distributed hierarchical RDF data |
| CN108304519A (en) * | 2018-01-24 | 2018-07-20 | 西安交通大学 | A kind of knowledge forest construction method based on chart database |
| CN108600321A (en) * | 2018-03-26 | 2018-09-28 | 中国科学院计算技术研究所 | A kind of diagram data storage method and system based on distributed memory cloud |
| CN110737659A (en) * | 2019-09-06 | 2020-01-31 | 平安科技（深圳）有限公司 | Graph data storage and query method, device and computer readable storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016154460A1 (en) * | 2015-03-24 | 2016-09-29 | Kyndi, Inc. | Cognitive memory graph indexing, storage and retrieval |
| EP3476051A1 (en) * | 2016-07-14 | 2019-05-01 | Huawei Technologies Co., Ltd. | General purpose data compression using simd engine |
| US11222070B2 (en) * | 2020-02-27 | 2022-01-11 | Oracle International Corporation | Vectorized hash tables |
| CN113204676B (en) * | 2021-01-07 | 2021-12-10 | 北京欧拉认知智能科技有限公司 | Compression storage method based on graph structure data |
-
2021
- 2021-10-09 CN CN202111175266.6A patent/CN113886652B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544291A (en) * | 2013-10-29 | 2014-01-29 | 东北林业大学 | Mobile object continuous k-nearest neighbor (CKNN) query method based on road based road networks tree (RRN-Tree) in road network |
| CN103617276A (en) * | 2013-12-09 | 2014-03-05 | 南京大学 | Method for storing distributed hierarchical RDF data |
| CN108304519A (en) * | 2018-01-24 | 2018-07-20 | 西安交通大学 | A kind of knowledge forest construction method based on chart database |
| CN108600321A (en) * | 2018-03-26 | 2018-09-28 | 中国科学院计算技术研究所 | A kind of diagram data storage method and system based on distributed memory cloud |
| CN110737659A (en) * | 2019-09-06 | 2020-01-31 | 平安科技（深圳）有限公司 | Graph data storage and query method, device and computer readable storage medium |
Non-Patent Citations (2)
| Title |
|---|
| "H-INDEX: Hash-Indexing for Parallel Triangle Counting on GPUs";Santosh Pandey;《2019 IEEE High Performance Extreme Computing Conference (HPEC)》;20191128;论文第1-5页 * |
| "基于H5的计算机联锁操作界面设计与实现";魏化永;《信息与电脑(理论版)》;20180615(第11期);第108-111页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113886652A (en) | 2022-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021083239A1 (en) | Graph data query method and apparatus, and device and storage medium | |
| US11580441B2 (en) | Model training method and apparatus | |
| Doktorov et al. | A dressing method in mathematical physics | |
| Darte et al. | Constructive methods for scheduling uniform loop nests | |
| CN108846133B (en) | Block chain storage structure based on B-M tree, B-M tree establishment algorithm and search algorithm | |
| WO2017053904A1 (en) | Distributed graph database | |
| US20220050887A1 (en) | Blockchain-based infringement detection method, apparatus, and electronic device | |
| CN103336758A (en) | Sparse matrix storage method CSRL (Compressed Sparse Row with Local Information) and SpMV (Sparse Matrix Vector Multiplication) realization method based on same | |
| US20140244657A1 (en) | Dynamic graph system for a semantic database | |
| CN113886652B (en) | Memory-priority multimode graph data storage and calculation method and system | |
| US20230367815A1 (en) | Energy-efficient hardware-software collaborative method and apparatus for graph processing | |
| CN114064984A (en) | Sparse array linked list-based world state increment updating method and device | |
| CN109189994B (en) | CAM structure storage system for graph computation application | |
| Li et al. | Research and implementation of parallel artificial fish swarm algorithm based on ternary optical computer | |
| CN110765319B (en) | Method for improving Janusgraph path exploration performance | |
| CN109885812B (en) | Method and device for dynamically adding hotwords and readable storage medium | |
| US9092335B2 (en) | Representing a tree structure on a flat structure | |
| Hutchison et al. | Lara: A key-value algebra underlying arrays and relations | |
| CN111552840A (en) | Method for converting JSON data into tree-shaped hierarchical data | |
| Bai et al. | Para-g: Path pattern query processing on large graphs | |
| Shivashankar et al. | Efficient software for programmable visual analysis using Morse-Smale complexes | |
| Morihata et al. | A practical tree contraction algorithm for parallel skeletons on trees of unbounded degree | |
| CN111723246A (en) | Data processing method, device and storage medium | |
| Rajendran et al. | Incremental MapReduce for K-medoids clustering of big time-series data | |
| US11960483B1 (en) | Constant time data structure for single and distributed networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |