CN109254725A - A kind of disk figure processing method and system based on subgraph building - Google Patents
A kind of disk figure processing method and system based on subgraph building Download PDFInfo
- Publication number
- CN109254725A CN109254725A CN201810838033.1A CN201810838033A CN109254725A CN 109254725 A CN109254725 A CN 109254725A CN 201810838033 A CN201810838033 A CN 201810838033A CN 109254725 A CN109254725 A CN 109254725A
- Authority
- CN
- China
- Prior art keywords
- vertex
- outgoing
- emitting edge
- subgraph
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0668—Interfaces specially adapted for storage systems adopting a particular infrastructure
- G06F3/0671—In-line storage system
- G06F3/0673—Single storage device
- G06F3/0674—Disk device
- G06F3/0676—Magnetic disk device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0602—Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
- G06F3/0604—Improving or facilitating administration, e.g. storage management
- G06F3/0607—Improving or facilitating administration, e.g. storage management by facilitating the process of upgrading existing storage systems, e.g. for improving compatibility between host and storage device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
- G06F3/0655—Vertical data movement, i.e. input-output transfer; data movement between one or more hosts and one or more storage devices
- G06F3/0659—Command handling arrangements, e.g. command buffers, queues, command scheduling
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing For Digital Recording And Reproducing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses it is a kind of based on subgraph building disk figure processing method and system, comprising: by diagram data tissue be two-way side block structure；Start iteration diagram to calculate；Diagram data is loaded into memory；The memory subgraph centered on vertex is constructed by efficient subgraph construction method；Subgraph is updated；The diagram data of update is write back into disk；Judge whether to reach the condition of convergence；Terminate iteration diagram to calculate.Disk figure processing method proposed by the present invention based on subgraph building is by continuously organizing together the vertex for needing to access in subgraph building process and side, ensure that internal storage access locality is fully used in subgraph building process, the high subgraph building overhead issues in present disk figure processing system have been well solved, the overall performance of system has been obviously improved.
Description
Technical field
The invention belongs to computer big data processing technology fields, more particularly, to a kind of magnetic based on subgraph building
Disk figure processing method and system.
Background technique
Under current big data background, shows and more and more large-scale graph data is analyzed, handled and is excavated
Application demand.In recent years, the figure processing system based on disk was obtained extensively due to the features such as its is at low cost, scalability is strong
Concern, such as the GraphChi of Carnegie Mellon University, the X-Stream of Ecole Polytechnique Federale de Lausanne, Tsinghua University
GridGraph etc..Diagram data is divided into several subgraphs by the processing of these disk figures, and loads and handle every time one from disk
A subgraph.In addition, these systems mostly use greatly the computation model centered on vertex, before handling each subgraph, need
Sub-graph data structure of the building centered on vertex in memory.It is each top in subgraph that this subgraph building process, which requires system,
Point addition emitting edge and outgoing side.However, due to side source summit or purpose vertex is discontinuous is stored in memory, building
The process of figure would generally bring very big expense.For example, GraphChi is when executing PageRank nomography, when subgraph constructs
Between the main reason for accounting for the 60% of the whole service time, leading to high subgraph building expense be that subgraph building process results in largely
Memory random read-write.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides a kind of disk figures based on subgraph building
Thus processing method and system solve the high subgraph in existing disk figure processing system and construct the larger technical problem of expense.
To achieve the above object, according to one aspect of the present invention, provide it is a kind of based on subgraph building disk figure at
Reason method, comprising:
It is that P disjoint vertex sections are respectively created one for each vertex section by the vertex partition for inputting figure
Emitting edge block structure and an outgoing side block structure, to utilize vertex in emitting edge block structure storage corresponding vertex section
Emitting edge, using the outgoing when block structure stores the outgoing on vertex in corresponding vertex section, then in the input figure
Side traversed, according to the source summit of each edge and purpose vertex, will when corresponding emitting edge block and outgoing is being written block,
In, the value of P is necessary to ensure that each incident block or is emitted capacity of the size less than memory of block；
Set the initialization value on each vertex and it is selected enliven vertex after, according to the vertex section of division, load current vertex
The vertex in section and the emitting edge block in current vertex section and outgoing side block obtain target diagram data into memory；
The emitting edge queue that every emitting edge in the target diagram data is written to the emitting edge purpose vertex, by every
It is emitted the outgoing side queue of the source summit when the outgoing is written, constructs the memory subgraph centered on vertex；
Data are concomitantly read from respective emitting edge in each vertex in the memory subgraph, and according to the data of reading
Opposite vertexes value is updated, and the vertex being updated then is write back disk.
Preferably, in the emitting edge block structure while by while purpose vertex be ranked up, in outgoing side block
In structure while by while source summit be ranked up.
Preferably, the vertex of enlivening is the vertex that vertex value is updated in last round of iteration.
Preferably, the emitting edge on the emitting edge purpose vertex is written in every emitting edge by the target diagram data
The outgoing side queue of every outgoing source summit when the outgoing is written is constructed the memory subgraph centered on vertex, packet by queue
It includes:
For every emitting edge in the target diagram data, with accessing the memory on the purpose vertex of the emitting edge first
Then the emitting edge is added to the emitting edge queue on its purpose vertex by location；
For every outgoing side in the target diagram data, the memory address of the source summit on the outgoing side is accessed first,
Then by outgoing queue in the outgoing for being added to its source summit.
Preferably, after the vertex that will be updated writes back memory, the method also includes:
Judge whether to reach the default condition of convergence, if not reaching the default condition of convergence, executes the load and work as
The vertex in preceding vertex section and the emitting edge block in current vertex section and outgoing side block obtain target diagram data into memory, if reaching
To the default condition of convergence, then terminate iteration diagram calculating.
It is another aspect of this invention to provide that providing a kind of disk figure processing system based on subgraph building, comprising:
Two-way side block structure constructs module, and the vertex partition for that will input figure is P disjoint vertex sections, for
An emitting edge block structure and an outgoing side block structure is respectively created in each vertex section, to utilize the emitting edge agllutination
Structure stores the emitting edge on vertex in corresponding vertex section, utilizes vertex in the outgoing side block structure storage corresponding vertex section
It is emitted side, then the side in the input figure is traversed, according to the source summit of each edge and purpose vertex, phase is written into side
The emitting edge block and outgoing side block answered, wherein the value of P is necessary to ensure that each incident block or is emitted the size of block less than memory
Capacity；
Loading module, for set each vertex initialization value and it is selected enliven vertex after, according to the vertex section of division,
It loads the vertex in current vertex section and the emitting edge block in current vertex section and outgoing side block obtains target figure number into memory
According to；
Memory subgraph constructs module, for the emitting edge purpose top to be written in every emitting edge in the target diagram data
The emitting edge queue of point constructs the outgoing side queue of every outgoing source summit when the outgoing is written centered on vertex
Memory subgraph；
Subgraph update module, for number concomitantly to be read from respective emitting edge in each vertex in the memory subgraph
According to, and be updated according to the data opposite vertexes value of reading, the vertex being updated then is write back into disk.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect:
Disk figure processing method and system proposed by the present invention based on subgraph building in subgraph building process by will need
The vertex and side to be accessed continuously organize together, it is ensured that internal storage access locality obtains abundant benefit in subgraph building process
With the high subgraph well solved in present disk figure processing system constructs overhead issues, has been obviously improved the whole of system
Body performance.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of disk figure processing method based on subgraph building provided in an embodiment of the present invention；
Fig. 2 is the overall process schematic diagram that a kind of system provided in an embodiment of the present invention handles some section subgraph；
Fig. 3 (a) is a kind of schematic diagram of digraph G provided in an embodiment of the present invention；
Fig. 3 (b) is provided in an embodiment of the present invention a kind of organizes digraph G for the schematic diagram of dibaryon block structure；
Fig. 4 be it is provided in an embodiment of the present invention it is a kind of by the vertex in vertex section 1 and side in digraph G be built into memory
The schematic diagram of figure.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
It is existing its object is to reduce the present invention provides a kind of disk figure processing method and system based on subgraph building
High subgraph in disk figure processing system constructs expense.
It is provided in an embodiment of the present invention based on subgraph building disk figure processing method, process as shown in Figure 1, include with
Lower step:
(1) it is two-way side block structure by diagram data tissue: is first P disjoint areas by the vertex partition for inputting figure
Between.For each vertex section, an emitting edge block and an outgoing side block structure is respectively created, stores this section vertex respectively
Emitting edge and outgoing side；Side in input figure is traversed, according to the source summit of each edge and purpose vertex, is written into
Corresponding emitting edge block and outgoing side block.
Wherein, the value of P is necessary to ensure that each incident block or is emitted capacity of the size less than memory of block.
(2) start iteration diagram to calculate: according to the setting of the requirement of nomography and user, setting the initialization value on vertex and choosing
Surely vertex is enlivened, starts to be iterated processing；
In embodiments of the present invention, enlivening vertex is the vertex that vertex value is updated in last round of iteration.
(3) load diagram data into memory: according to step (1) divide vertex section, load current interval vertex and
Corresponding emitting edge block and outgoing side block are into memory；
In embodiments of the present invention, it ensure that using the streaming reading of diagram data to disk in the loading procedure of step (3)
Sequential access, thus ensure that external memory IO maximum utilize.
(4) the memory subgraph centered on vertex: the figure number loaded according to step (3) is constructed by subgraph construction method
According to, every emitting edge is written to the emitting edge queue on its purpose vertex, by every outgoing in the outgoing that its source summit is written team
Column, are built into a memory subgraph centered on vertex；
In embodiments of the present invention, the purpose top of every emitting edge is successively accessed in the subgraph construction method of step (4)
The source summit of point and every outgoing side, since all emitting edges are ranked up according to purpose vertex, all outgoing sides are
It is ranked up according to source summit, this construction method is capable of the locality of abundant internal storage access, and building subgraph can be greatly reduced
Expense.
(5) be updated to subgraph: the memory subgraph being built into based on step (4) uses the mode centered on vertex
It is updated.Under this update mode, data are concomitantly read from the emitting edge of oneself in each vertex, according to the data of reading
Opposite vertexes value is updated.
In embodiments of the present invention, vertex data and side data separating are used when step (5) is updated memory subgraph
Method, in this case, only vertex value can be updated, and number of edges will not be updated according to the topological structure of figure is then reflected,
This method significantly reduces the data volume of disk read-write, improves the overall performance of system.
(6) diagram data of update is write back into disk: the vertex being updated is write back into disk, it is ensured that the one of overall calculation result
Cause property；
(7) judge whether to reach the condition of convergence, if reached, execute step (8), otherwise enter next iteration, execute
Step (3).Wherein, the condition of convergence is by user preset.
(8) terminate iteration diagram to calculate, export calculated result.
Fig. 2 is the overall process schematic diagram that system handles some section subgraph.System is first by the top in a certain vertex section
Point data and number of edges are loaded into memory according to (with two-way side block structure Dual-Chunks storage).Then being constructed by subgraph will
These data are built into the memory subgraph centered on vertex.After the completion of memory building, system is according to centered on vertex
Update mode is updated subgraph.Finally, updated diagram data is write back disk by system.
Fig. 3 is the schematic diagram organized digraph G as two-way side block structure.Include 6 tops in the digraph G of Fig. 3 (a)
Point is divided into the section (1,3) and (4,6) of the sizes such as two.The emitting edge on each section vertex and outgoing side are respectively stored in
In corresponding emitting edge block (in-chunk) and outgoing side block (out-chunk).As shown in Fig. 3 (b), the incidence in vertex section 1
It is respectively stored in in-chunk (1) and out-chunk (1) while with outgoing, the emitting edge and outgoing side difference in vertex section 2
It is stored in in-chunk (2) and out-chunk (2).In emitting edge block while according to while purpose vertex be ranked up, be emitted
In block while be ranked up according to the source summit on side.It is stored which ensures that the vertex order for needing to access in subgraph building
In memory.For example, the side storage order in in-chunk (1) is (3,1), (1,2), (5,2), (2,3), (4,3), (5,3).
Side storage order in Out-chunk (2) is (4,3), (4,6), (5,2), (5,3), (5,4), (6,5).
Fig. 4 is the schematic diagram that the vertex in vertex section 1 and side in digraph G are built into memory subgraph.System concurrency
Access emitting edge loaded into memory and outgoing side.For every emitting edge, the memory address on its purpose vertex is accessed first,
Then this edge is added to the emitting edge queue on its purpose vertex.For every outgoing side, the interior of its source summit is accessed first
Address is deposited, then this edge is added to the outgoing side queue of its source summit.As shown in figure 4, successively accessing the interior of vertex 1,2,3
Address is deposited, and adds corresponding emitting edge and outgoing side.Since all emitting edges are ranked up by purpose vertex, all goes out
It penetrates side to be ranked up by source summit, in the emitting edge and outgoing side for adding vertex, the access order of opposite vertexes is completely continuous
's.Therefore, the locality of internal storage access is fully utilized, and the number of memory random read-write is significantly reduced, and constructs subgraph
Expense can be greatly reduced.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (6)
1. a kind of disk figure processing method based on subgraph building characterized by comprising
It is that an incidence is respectively created for each vertex section in P disjoint vertex sections by the vertex partition for inputting figure
The block structure in block structure and an outgoing, to utilize the incidence on vertex in emitting edge block structure storage corresponding vertex section
Side, using the outgoing when block structure stores the outgoing on vertex in corresponding vertex section, then to the side in the input figure
It is traversed, according to the source summit of each edge and purpose vertex, by the block when corresponding emitting edge block and outgoing is written, wherein P
Value be necessary to ensure that each incident block or be emitted the size of block and be less than the capacity of memory；
Set the initialization value on each vertex and it is selected enliven vertex after, according to the vertex section of division, load current vertex section
Vertex and current vertex section emitting edge block and outgoing side block target diagram data is obtained into memory；
Every is emitted by the emitting edge queue that every emitting edge in the target diagram data is written to the emitting edge purpose vertex
The outgoing side queue of source summit when the outgoing is written constructs the memory subgraph centered on vertex；
Data are concomitantly read from respective emitting edge in each vertex in the memory subgraph, and according to the data of reading to top
Point value is updated, and the vertex being updated then is write back disk.
2. the method according to claim 1, wherein the purpose top when pressing in the emitting edge block structure
Point is ranked up, the outgoing in block structure while be ranked up by the source summit on side.
3. the method according to claim 1, wherein the vertex of enlivening is vertex value quilt in last round of iteration
The vertex of update.
4. according to the method described in claim 2, it is characterized in that, every emitting edge by the target diagram data is write
The emitting edge queue for entering the emitting edge purpose vertex, by the outgoing side queue of every outgoing source summit when the outgoing is written, structure
Build the memory subgraph centered on vertex, comprising:
For every emitting edge in the target diagram data, the memory address on the purpose vertex of the emitting edge is accessed first, so
The emitting edge is added to the emitting edge queue on its purpose vertex afterwards；
For every outgoing side in the target diagram data, the memory address of the source summit on the outgoing side is accessed first, then
By outgoing queue in the outgoing for being added to its source summit.
5. the method according to claim 1, which is characterized in that in the vertex that will be updated writes back
After depositing, the method also includes:
Judge whether to reach the default condition of convergence, if not reaching the default condition of convergence, executes the load and work as sinciput
The vertex in point section and the emitting edge block in current vertex section and outgoing side block obtain target diagram data into memory, if reaching institute
The default condition of convergence is stated, then terminates iteration diagram calculating.
6. a kind of disk figure processing system based on subgraph building characterized by comprising
Two-way side block structure constructs module, and the vertex partition for that will input figure is P disjoint vertex sections, for each
An emitting edge block structure and an outgoing side block structure is respectively created, to deposit using the emitting edge block structure in vertex section
The emitting edge for storing up vertex in corresponding vertex section utilizes the outgoing on vertex in the outgoing side block structure storage corresponding vertex section
Then side traverses the side in the input figure, according to the source summit of each edge and purpose vertex, side is written corresponding
Emitting edge block and outgoing side block, wherein the value of P is necessary to ensure that each incident block or is emitted appearance of the size less than memory of block
Amount；
Loading module, for set each vertex initialization value and it is selected enliven vertex after, according to the vertex section of division, load
The vertex in current vertex section and the emitting edge block in current vertex section and outgoing side block obtain target diagram data into memory；
Memory subgraph constructs module, for the emitting edge purpose vertex to be written in every emitting edge in the target diagram data
The outgoing side queue of every outgoing source summit when the outgoing is written is constructed the memory centered on vertex by emitting edge queue
Subgraph；
Subgraph update module, for data concomitantly to be read from respective emitting edge in each vertex in the memory subgraph,
And be updated according to the data opposite vertexes value of reading, the vertex being updated then is write back into disk.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810838033.1A CN109254725B (en) | 2018-07-26 | 2018-07-26 | Disk graph processing method and system based on subgraph construction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810838033.1A CN109254725B (en) | 2018-07-26 | 2018-07-26 | Disk graph processing method and system based on subgraph construction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109254725A true CN109254725A (en) | 2019-01-22 |
| CN109254725B CN109254725B (en) | 2020-05-19 |
Family
ID=65050043
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810838033.1A Active CN109254725B (en) | 2018-07-26 | 2018-07-26 | Disk graph processing method and system based on subgraph construction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109254725B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110187968A (en) * | 2019-05-22 | 2019-08-30 | 上海交通大学 | Diagram data under heterogeneous computing environment handles accelerated method |
| CN112988064A (en) * | 2021-02-09 | 2021-06-18 | 华中科技大学 | Concurrent multitasking-oriented disk image processing method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104952032A (en) * | 2015-06-19 | 2015-09-30 | 清华大学 | Graph processing method and device as well as rasterization representation and storage method |
| CN105224528A (en) * | 2014-05-27 | 2016-01-06 | 华为技术有限公司 | The large data processing method calculated based on figure and device |
| CN105590321A (en) * | 2015-12-24 | 2016-05-18 | 华中科技大学 | Block-based subgraph construction and distributed graph processing method |
| CN105653204A (en) * | 2015-12-24 | 2016-06-08 | 华中科技大学 | Distributed graph calculation method based on disk |
| US9400767B2 (en) * | 2013-12-17 | 2016-07-26 | International Business Machines Corporation | Subgraph-based distributed graph processing |
| US20160350662A1 (en) * | 2015-05-28 | 2016-12-01 | GraphSQL, Inc. | System and method for real-time graph-based recommendations |
| CN108132838A (en) * | 2016-11-30 | 2018-06-08 | 华为技术有限公司 | A kind of method, apparatus and system of diagram data processing |
-
2018
- 2018-07-26 CN CN201810838033.1A patent/CN109254725B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9400767B2 (en) * | 2013-12-17 | 2016-07-26 | International Business Machines Corporation | Subgraph-based distributed graph processing |
| CN105224528A (en) * | 2014-05-27 | 2016-01-06 | 华为技术有限公司 | The large data processing method calculated based on figure and device |
| US20160350662A1 (en) * | 2015-05-28 | 2016-12-01 | GraphSQL, Inc. | System and method for real-time graph-based recommendations |
| CN104952032A (en) * | 2015-06-19 | 2015-09-30 | 清华大学 | Graph processing method and device as well as rasterization representation and storage method |
| CN105590321A (en) * | 2015-12-24 | 2016-05-18 | 华中科技大学 | Block-based subgraph construction and distributed graph processing method |
| CN105653204A (en) * | 2015-12-24 | 2016-06-08 | 华中科技大学 | Distributed graph calculation method based on disk |
| CN108132838A (en) * | 2016-11-30 | 2018-06-08 | 华为技术有限公司 | A kind of method, apparatus and system of diagram data processing |
Non-Patent Citations (4)
| Title |
|---|
| YONGLI CHENG ET AL.: "LCC-Graph: A high-performance graph-processing framework with lowcommunication costs", 《2016 IEEE》 * |
| 张海威等: "一种基于自适应结构概要的有向标签图子图匹配查询算法", 《计算机学报》 * |
| 李金吉等: "流式处理的异步图处理框架", 《软件学报》 * |
| 汪修能: "基于活跃顶点的外存图计算系统I/O访问策略", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110187968A (en) * | 2019-05-22 | 2019-08-30 | 上海交通大学 | Diagram data under heterogeneous computing environment handles accelerated method |
| CN112988064A (en) * | 2021-02-09 | 2021-06-18 | 华中科技大学 | Concurrent multitasking-oriented disk image processing method |
| CN112988064B (en) * | 2021-02-09 | 2022-11-08 | 华中科技大学 | Concurrent multitask-oriented disk graph processing method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109254725B (en) | 2020-05-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10817514B2 (en) | Splitting of a join operation to allow parallelization | |
| Checconi et al. | Traversing trillions of edges in real time: Graph exploration on large-scale parallel machines | |
| US8364909B2 (en) | Determining a conflict in accessing shared resources using a reduced number of cycles | |
| CN104809179B (en) | The device and method for accessing Hash table | |
| CN104952032B (en) | Processing method, device and the rasterizing of figure represent and storage method | |
| Mattila et al. | Comparison of implementations of the lattice-Boltzmann method | |
| CN102855121B (en) | Branching processing method and system | |
| CN105653204A (en) | Distributed graph calculation method based on disk | |
| JP2010539609A5 (en) | ||
| Wang et al. | BENU: Distributed subgraph enumeration with backtracking-based framework | |
| JP6877435B2 (en) | Database operation method and equipment | |
| Giannakouris et al. | MuSQLE: Distributed SQL query execution over multiple engine environments | |
| CN109254725A (en) | A kind of disk figure processing method and system based on subgraph building | |
| CN109522428A (en) | A kind of external memory access method of the figure computing system based on index positioning | |
| CN110083312A (en) | Disk expansion method, device and computer equipment | |
| CN107678446A (en) | A kind of unmanned plane for extensive target complex cooperates with target assignment method | |
| CN106569896A (en) | Data distribution and parallel processing method and system | |
| Sun et al. | GraphMP: An efficient semi-external-memory big graph processing system on a single machine | |
| CN109189994A (en) | A kind of CAM structure storage system calculating application towards figure | |
| CN104376047B (en) | A kind of big table join methods based on HBase | |
| CN101846978B (en) | Reliability analyzing method based on GSPN reliability model | |
| Qin et al. | The performance of sql-on-hadoop systems-an experimental study | |
| Kipf et al. | Adaptive geospatial joins for modern hardware | |
| CN106775450B (en) | A kind of data distribution method in mixing storage system | |
| CN105045879A (en) | Parallel data processing method |
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 |