CN106446104A - Message synchronizing method and system based on graph calculation engine - Google Patents
Message synchronizing method and system based on graph calculation engine Download PDFInfo
- Publication number
- CN106446104A CN106446104A CN201610822882.9A CN201610822882A CN106446104A CN 106446104 A CN106446104 A CN 106446104A CN 201610822882 A CN201610822882 A CN 201610822882A CN 106446104 A CN106446104 A CN 106446104A
- Authority
- CN
- China
- Prior art keywords
- summit
- vertex
- message
- sent
- principal
- 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.)
- Pending
Links
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
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The invention particularly relates to a message synchronizing method and system based on a graph calculation engine. The message synchronizing method based on the graph calculation engine includes the steps that S1, messages received by a current vortex are sent to a next vortex along the side of the current vortex according to a preset rule in all sub-graphs; S2, if the next vortex has a master vortex index, the messages received by the next vortex is sent to a master vortex of the next vortex according to the master vortex index of the next vortex; S3, all the messages received by the master vortex of the next vortex are fused; S4, the fused messages are sent to all slave vortexes of the next vortex according to slave vortex indexes of the master vortex of the next vortex; S5, if a terminal condition is not met, S1 is executed. According to the message synchronizing method and system based on the graph calculation engine, the position of a target point is rapidly sought through the master vortex index and the slave vortex indexes, no calculation is needed again, and the efficiency of data synchronization among different sub-graphs is improved.
Description
Technical field
The present invention relates to diagram data technical field is and in particular to a kind of message synchronization method based on figure computing engines and be
System.
Background technology
Big figure can be advantageously applied to various pattern analysiss and mining application, for example, social network analysis, mould
Formula identification, websites collection and Social Media marketing etc..But, with needing, the volume processing figure is more and more huger, single clothes
Business device can't bear the heavy load, and therefore, a series of large-scale distributed parallel frame diagrams arise at the historic moment.Distributed parallel frame diagram is
One big figure is divided into multiple subgraphs, then these is stored separately, can be with multiple subgraph parallel runnings, greatly when therefore calculating
Big raising operation efficiency, can be applicable to the Drawing Problems of large-scale real world.
The cutting method of figure is divided into side cutting and summit cutting.While cutting through cutting edge, summit is uniformly distributed to
In each subgraph, the more multiple subgraph of cut end bay, communication and the storage overhead after cutting with cut while quantity become
Direct ratio.Compared with the cutting of side, each side is uniformly distributed in each subgraph for summit cutting, and it is multiple to allow summit to cross over
The number of the pattern zoning that subgraph, the communication of summit cutting and storage overhead are crossed over to each summit is directly proportional.Have theoretical and
Test result indicate that, the figure of real world has good summit to cut, and therefore, summit cutting is more suitable for the figure of real world
Problem.
Subgraph after cutting is respectively stored on multiple stage machine, and the information conversion of each subgraph needs consumption network to provide
Source, for especially big to some figure of real world, the quantity of subgraph is a lot, and the information exchange between subgraph also can very frequently
Numerous.
Based on existing diagram data storage organization, the information of subgraph storage includes information a little and the information on side, the letter of point
Including point numbering and point attribute, while information include while source point, terminal and side attribute.Based on existing diagram data
The unpack storage on storage organization, summit and side, increased communication overhead；And lack index information in storage organization, greatly
Reduce greatly computation rate.
Content of the invention
The message synchronization method and system based on figure computing engines providing for defect of the prior art, the present invention,
Based on a kind of new storage organization, improve the efficiency of synchrodata between different subgraphs.
In a first aspect, a kind of message synchronization method based on figure computing engines that the present invention provides, including：Step S1,
In each subgraph, the message that current vertex is received is sent to next summit by pre-defined rule along the side of described current vertex；
Step S2, if next summit described has principal vertex to index, according to the principal vertex index on next summit described, by next top described
The message that point receives is sent to the principal vertex on next summit described；Step S3, the principal vertex merging next summit described receives
The all message arriving；Step S4, the principal vertex according to next summit described from vertex index, the message after merging is sent to
Next summit described all from summit；Step S5, if not meeting end condition, returns execution step S1.
The message synchronization method based on figure computing engines that the present invention provides, during message synchronization, by principal vertex
Index and quickly find position between different subgraphs for the summit it is no longer necessary to calculate from vertex index, improve different subgraphs
Between synchrodata efficiency.
Preferably, the described message receiving current vertex is sent to down along the side of described current vertex by pre-defined rule
One summit, including：If meeting trigger condition, the message that current vertex is received is by pre-defined rule along described current vertex
While being sent to next summit.
Preferably, if described meet trigger condition, the message that current vertex is received is worked as along described by pre-defined rule
The side on front summit is sent to next summit, including：If while being activated, the message that current vertex is received is by pre-defined rule along institute
The side stating current vertex is sent to next summit.
Preferably, if side is activated, message that current vertex is received by pre-defined rule along described current vertex side
It is sent to next summit, including：If current vertex receives message, activate described current vertex；If described current vertex quilt
Activation, then according to when the activation condition described current vertex of activation is corresponding；
If the corresponding side of described current vertex is activated, message that current vertex is received is by pre-defined rule along described
The side of current vertex is sent to next summit；The state of activation of described current vertex is reset.
Preferably, also include：The attribute of the principal vertex on next summit described in information updating after described fusion；According to
The principal vertex on next summit described from vertex index, described in after updating, the attribute of the principal vertex on next summit is sent to institute
State next summit all from summit；The attribute of the principal vertex on next summit according to after renewal, updates next top described
The attribute from summit of point.
Second aspect, a kind of message synchronization system based on figure computing engines that the present invention provides, including：Message sends mould
Block, for, in each subgraph, the message that current vertex is received is sent to along the side of described current vertex by pre-defined rule
Next summit；Message summarizing module, if having principal vertex to index for next summit described, according to the main top on next summit described
Point index, the message that next summit described is received is sent to the principal vertex on next summit described；Message Fusion Module, is used for
Merge all message that the principal vertex on next summit described receives；Message synchronization module, for according to next summit described
Principal vertex from vertex index, by the message after merging be sent to next summit described all from summit；Cycle criterion module,
If for not meeting end condition, returning the function of execution message transmission module.
The message synchronization system based on figure computing engines that the present invention provides, during message synchronization, by principal vertex
Index and quickly find position between different subgraphs for the summit it is no longer necessary to calculate from vertex index, improve different subgraphs
Between synchrodata efficiency.
Preferably, described message transmission module specifically for：If meeting trigger condition, by disappearing that current vertex receives
Breath is sent to next summit by pre-defined rule along the side of described current vertex.
Preferably, described message transmission module specifically for：If while being activated, the message that current vertex is received is by pre-
Set pattern is then sent to next summit along the side of described current vertex.
Preferably, described message transmission module specifically for：If current vertex receives message, described in activation, work as sinciput
Point；If described current vertex is activated, according to when the activation condition described current vertex of activation is corresponding；If described work as sinciput
The corresponding side of point is activated, then be sent to down the message that current vertex receives along the side of described current vertex by pre-defined rule
One summit；The state of activation of described current vertex is reset.
Preferably, also include attribute update module, be used for：Next summit described in information updating after described fusion
The attribute of principal vertex；Principal vertex according to next summit described from vertex index, the master on next summit described in after updating
The attribute on summit be sent to next summit described all from summit；The genus of the principal vertex on next summit according to after renewal
Property, update the attribute from summit on next summit described.
Brief description
Fig. 1 is the schematic diagram of a complete big figure；
Fig. 2 is by three subgraphs after big figure cutting；
The storage organization based on figure computing engines that Fig. 3 is provided by the embodiment of the present invention；
The flow chart of the message synchronization method based on figure computing engines that Fig. 4 is provided by the embodiment of the present invention；
The structured flowchart of the message synchronization system based on figure computing engines that Fig. 5 is provided by the embodiment of the present invention.
Specific embodiment
Below in conjunction with accompanying drawing, the embodiment of technical solution of the present invention is described in detail.Following examples are only used for
Clearly technical scheme is described, is therefore intended only as example, and the protection of the present invention can not be limited with this
Scope.
It should be noted that unless otherwise stated, technical term used in this application or scientific terminology should be this
The ordinary meaning that bright one of ordinary skill in the art are understood.
In order to improve the data synchronization efficiency between subgraph, embodiments provide a kind of storage side of diagram data
Method, including：
One complete big figure cutting is multiple subgraphs by the mode according to summit segmentation, wherein, in each subgraph
Side is all complete, and a summit is likely to be present in two or more subgraphs；
For the summit existing in two or more subgraphs it is intended that one of subgraph is the chief subgraph on this summit,
Remaining subgraph be this summit from subgraph, this summit in chief subgraph just be principal vertex, on this summit from subgraph be
From summit.
According to predetermined storage organization, preserve the data on summit in each subgraph and side.
Fig. 1 gives a complete big figure.
Carry out cutting along the dotted line in Fig. 1 and just obtain as shown in Figure 2 three subgraph.Wherein, solid circles are principal vertex
(summit a, b, c, d as in subgraph 0), dotted line circle is from summit (summit g, the f as subgraph 0, the summit c) in subgraph 1；With
As a example the c of summit, summit c is principal vertex in subgraph 0, then subgraph 0 is the chief subgraph of summit c, and subgraph 1 and subgraph 2 are summit c's
From subgraph.
Fig. 3 gives the storage organization being stored in a manner described.In this storage organization, each subgraph is independent storage
, therefore this storage organization is applied to parallel computation.Be stored with each subgraph 3 partial contents：
Part I maps for ID.Local2Global stores all summits in subgraph, and in Global ID, (summit is schemed big
In ID), the call number of wherein all principal vertexs is less than from summit.Global2Local stores the complete of each summit in subgraph
The index of office ID and the mapping of local ident (ID in subgraph for the summit), taking the subgraph 0 (Partition0) in Fig. 3 as a example,
Under global2Local field, " 231 " are the Global ID of summit a, and " 0 " is the local ident of summit a.In a sub- in figure, if
Principal vertex and the quantity from summit are respectively m and n, if the local ident on a summit is less than m, this summit is principal vertex, otherwise
This summit is from summit.
Part II is vertex information (vertices).AttributesOfMaster is used for storing the attribute of principal vertex
Value, its sequence is identical with the order of local2Global.SlaveInfoOfMaster be used for storing principal vertex corresponding from summit
Index, the subgraph including being located from summit from vertex index is numbered and the local ident subgraph, if a summit is split,
There will be one or more be distributed in other subgraphs from summit, for example, (i.e. summit c) has two to the summit 891 in subgraph 0
Individual be respectively 4 and 3 (global2Local see in subgraph 1 and subgraph 2) in the local ident in subgraph 1 and subgraph 2 from summit,
SlaveInfoOfMaster corresponding position storage summit 891 from vertex index " 1,4 " and " 2,3 " therefore in subgraph 0.
AttributesOfSlave is used for the property value from summit for the storage.MasterInfoOfSlave is used for storage from vertex correspondence
Principal vertex indexes, and principal vertex index includes subgraph numbering and the local ident in subgraph for the principal vertex at principal vertex place, for example, exist
In subgraph 0, from summit 412, (i.e. subgraph 2, its local ident in subgraph 2 is 0 to the principal vertex of summit g), therefore in subgraph
Principal vertex index " 2,0 " on masterInfoOfSlave corresponding position storage summit 412 in 0.Attribute from summit can root
Property value according to principal vertex is updated.
Part III is side information (edges).The local ident of the source point on side is stored in localSrcIds, the terminal on side
Local ident be stored in localDstIds corresponding with localSrcIds, by summit 231 point to summit 2851 side as a example,
Summit 231 (local ident is 0) is source point, then localSrcIds (0) is 0, and summit 2851 (local ident is 1) is terminal, then
LocalDstIds (0) is 1.EdgeAttributes stores the property value on side.
Inclusion relation (the principal vertex rope being included based on the storage organization of figure computing engines a little provided in an embodiment of the present invention
Draw and from vertex index), in search procedure, impact point quickly can be found by index it is no longer necessary to calculate, improve
Search efficiency；In addition, jointly store point with side, some shared information do not need to repeat to store, and save memory space.
Based on the storage organization of above-mentioned diagram data, a kind of message based on figure computing engines provided in an embodiment of the present invention is same
One step process, concrete steps are as shown in figure 4, include：
Step S1, in each subgraph, the message that current vertex is received is sent out along the side of current vertex by pre-defined rule
Give next summit.
Step S2, if next summit has principal vertex to index, according to the principal vertex index on next summit, next summit is connect
The message receiving is sent to the principal vertex on next summit.
Wherein, if next summit does not have principal vertex to index, show that this summit is exactly principal vertex, this summit only need to receive it
The message sending over from summit.
Step S3, merges all message that the principal vertex on next summit receives.
Step S4, the principal vertex according to next summit from vertex index, the message after merging is sent to next summit
All from summit.
Step S5, if not meeting end condition, returns execution step S1.
Wherein, return the circulation that execution step S1 carries out a new round, the summit receiving message in last circulation is
For the current vertex in circulating next time.Current vertex is to meet condition in certain cyclic process, can carry out message transmission
Summit, for be different from other be unsatisfactory for message send condition summits, do not represent physical presence summit current vertex.
Message synchronization method based on figure computing engines provided in an embodiment of the present invention, the message in different subgraphs sends
(corresponding step S1), message coalescing (corresponding step S2, S3), message synchronization (corresponding step S4) etc. are all concurrent processing,
During message synchronization, indexed by principal vertex and quickly find position between different subgraphs for the summit from vertex index, no
Need again to calculate, improve the efficiency of synchrodata between different subgraphs.
Wherein, the implementation of step S1 has multiple, embodiments provides a kind of optimal way of step S1, tool
Body includes：If meeting trigger condition, the message that current vertex is received is sent to along the side of current vertex by pre-defined rule
Next summit.
Further, the optimal way of step S1 can also be：If while being activated, message that current vertex is received by
Pre-defined rule is sent to next summit along the side of current vertex.
Further, the optimal way of step S1 can also be：If current vertex receives message, sinciput is worked as in activation
Point；If current vertex is activated, according to when activation condition activation current vertex is corresponding；If current vertex corresponding side quilt
Activation, then be sent to next summit by pre-defined rule along the side of current vertex by the message that current vertex receives；Sinciput will be worked as
The state of activation of point resets.This method for optimizing, receives message for sending the trigger condition of message with summit, forms chain type anti-
Should, new information can be automatically passed to next summit by side, until meeting end condition, improves information transmitting efficiency.
Wherein, the pre-defined rule in step S1 includes any one in following rule：(1) message can only be issued from source point
Terminal, (2) message can only be sent to source point from terminal, and (3) message both can be sent to terminal from source point and can also be sent to source from terminal
Point.
Wherein, the source point when the condition being activated includes is activated and/or the terminal on side is activated.
Wherein, the end condition in step S5 includes：Do not produce new message and/or repeatedly cycle through number of times and reach one
Fixed number amount.Do not produce new information and refer to that message does not send along side, that is, do not have new summit to receive message.
Message synchronization method based on figure computing engines provided in an embodiment of the present invention, also includes attribute and updates step, tool
Body includes：The attribute of the principal vertex according to next summit of information updating after merging；Principal vertex according to next summit from top
Point index, by update after next summit principal vertex attribute be sent to next summit all from summit；After updating
The principal vertex on next summit attribute, update the attribute from summit on next summit.
Taking send message as a example, a kind of message synchronization method based on figure computing engines provided in an embodiment of the present invention, bag
Include：
Step S10, in each subgraph, the message that current vertex is received is sent out along the side of current vertex by pre-defined rule
Give next summit.
Wherein, message includes message content and representative points ID, and representative points ID is the overall situation in big in figure for the representative points
ID.
Step S20, if next summit has principal vertex to index, according to the principal vertex index on next summit, by next summit
The message receiving is sent to the principal vertex on next summit.
Wherein, if next summit does not have principal vertex to index, show that this summit is exactly principal vertex, this summit only need to receive it
The message sending over from summit.
Step S30, merges all message that the principal vertex on next summit receives.
Step S40, the principal vertex according to next summit from vertex index, the message after merging is sent to next summit
All from summit.
Step S50, if not meeting end condition, returns execution step S10.
Wherein, end condition is sent to representative points for message；Judge Global ID's whether phase of principal vertex and representative points
With；If differing, not meeting end condition, returning execution step S10；If identical, principal vertex is representative points, meets
End condition, terminates circulation.
Message synchronization method based on figure computing engines provided in an embodiment of the present invention, in message transmitting process, passes through
Principal vertex indexes and quickly finds position subgraph for the summit it is no longer necessary to calculate from vertex index, improves different sons
The efficiency of synchrodata between figure.
Wherein, the implementation of step S10 has multiple, embodiments provides a kind of optimal way of step S10,
Specifically include：If meeting trigger condition, the message that current vertex is received is sent along the side of current vertex by pre-defined rule
To next summit.
Further, the optimal way of step S10 can also be：If while being activated, message that current vertex is received
It is sent to next summit by pre-defined rule along the side of current vertex.
Further, the optimal way of step S10 can also be：If current vertex receives message, sinciput is worked as in activation
Point；If current vertex is activated, according to when activation condition activation current vertex is corresponding；If current vertex corresponding side quilt
Activation, then be sent to next summit by pre-defined rule along the side of current vertex by the message that current vertex receives；Sinciput will be worked as
The state of activation of point resets.This method for optimizing, receives message for sending the trigger condition of message with summit, forms chain type anti-
Should, new information can be automatically passed to next summit by side, until meeting end condition, improves information transmitting efficiency.
Wherein, the pre-defined rule in step S10 includes any one in following rule：(1) message can only be sent out from source point
To terminal, (2) message can only be sent to source point from terminal, and (3) message both can be sent to terminal from source point and can also be sent to from terminal
Source point.In practical application, the direction that message sends can set according to the actual demand of user, the such as genus according to source point and terminal
Property value, the particular content of message judge the rule that sends from which kind of message.
Wherein, the source point when the condition being activated includes is activated and/or the terminal on side is activated.
Message synchronization method based on figure computing engines provided in an embodiment of the present invention, also includes attribute and updates step, tool
Body includes：The attribute of the principal vertex according to next summit of information updating after merging；Principal vertex according to next summit from top
Point index, by update after next summit principal vertex attribute be sent to next summit all from summit；After updating
The principal vertex on next summit attribute, update the attribute from summit on next summit.
Taking the process of following Message Processing as a example illustrate the application of the embodiment of the present invention it is assumed that we want to judge big in figure
The distances that individual specific point is put to other, distance can consider that direction can not also consider.Taking the summit 891 in Fig. 1 as a example, examine
Consider during direction and can reach which point (going out) and which is put and can reach 891 (coming in) it is considered to direction in point summit 891
When, the processing mode of two kinds of situations is similar to.By do not consider direction and consider in direction the first situation as a example it is assumed that summit
891 starting points sending for message：
Situation 1：Do not consider the direction of message transmission, its processing procedure is：
(1) initialize the attribute on all summits first, if the ID on summit is 891, property value is set to 0, be otherwise set to-
1.Said process is by each subgraph parallel processing summit of oneself.
(2) message transmitting process：For each edge, if judging attribute >=0 of source point and the attribute of terminal<0, then to end
Point sends a message, and message content is the property value+1 of source point；Else if the attribute of attribute >=0 of terminal and source point<0,
Then send a message to terminal, message content is the property value+1 of terminal.
With the size of source point and the property value of terminal as foundation in this embodiment, judge whether to carry out message along certain side
Transmission, in practical application, can be set according to the actual demand of user.
(3) message coalescing process：Message transmitting process is carried out in multiple subgraphs parallel, therefore once can produce and much disappear
Breath, the message being sent to same vertices all converges toward its principal vertex.Principal vertex may receive multiple messages, and principal vertex will receive
Multiple messages merge, because, in the present embodiment, often in wheel iteration, the content of message is all identical, so principal vertex receives
One can at will be chosen during multiple message.For example, as in Fig. 1, first round iteration summit 891 can apicad 231,2343,
6427th, 442,9712 totally 5 point transmission contents be 1 message；Second wheel iteration, in subgraph 0, all can in summit 231 and summit 9712
Apicad 412 transmission contents are 2 message, and meanwhile, in subgraph 2, summit 9712 can 412 transmission contents be apicad also 2 disappear
Breath, message is merged by the summit 412 in subgraph 0 and in subgraph 2.
(4) attribute renewal process：After each summit receives message, property value is updated to message content, and will update
Property synchronization is to from summit.
(5) return to second step and be circulated iteration, until second step does not have new information to produce.
For situation 1, in conjunction with Fig. 1, whole message sends, merges, synchronous process is as follows：
The first round：Summit 891 apicad 231,2343,6427,442,9712 each transmission content in totally 5 summits be 1 disappear
Breath, the attribute of oneself is updated to 1 by this 5 summits；
Second wheel：Summit 231 apicad 327,2851,412 each transmission contents be 2 message, summit 2343 apicad 327
Transmission content is 2 message, and summit 442 apicad 63 sends the message that content are 2, summit 9712 apicad 412,2081
Each one content of transmission is 2 message, and after receiving each summit message coalescing of message, Update attribute is 2；
Third round：The summit receiving message in second wheel does not have next summit can send message, that is, do not have new information,
So iteration terminates.
Final result is：The property value on summit 891 is 0, and summit 231,2343,6427,442,9712 is 1 (i.e. summit 891
These points just can be reached through a step), the property value on other each summits is 2.
Situation 2：Summit 891 can reach the situation of which point, and process is as follows：
(1) initialize the attribute on all summits first, if the ID on summit is 891, property value is set to 0, be otherwise set to-
1.Said process is by each subgraph parallel processing summit of oneself.
(2) message transmitting process：For each edge, if judging attribute >=0 of source point and the attribute of terminal<0, then to end
Point sends a message, and message content is the property value+1 of source point.
With the size of source point and the property value of terminal as foundation in this embodiment, judge whether to carry out message along certain side
Transmission, in practical application, can be set according to the actual demand of user.
(3) message coalescing process：Message transmitting process is carried out in multiple subgraphs parallel, therefore once can produce and much disappear
Breath, the message being sent to same vertices all converges toward its principal vertex.Principal vertex may receive multiple messages, and principal vertex will receive
Multiple messages merge, because, in the present embodiment, often in wheel iteration, the content of message is all identical, so principal vertex receives
One can at will be chosen during multiple message.
(4) attribute renewal process：After each summit receives message, property value is updated to message content.And by update
Property synchronization is to from summit.
(5) return to second step and be circulated iteration, until second step does not have new information to produce.
For situation 2, in conjunction with Fig. 1, whole message sends, merges, synchronous process is as follows：
The first round：Summit 891 apicad 2343,6427,442,9712 each transmission contents be 1 message, this 4 summits will
The attribute of oneself is updated to 1；
Second wheel：Summit 442 apicad 63 sends the message that content are 2, and summit 9712 apicad 2081 sends one
Bar content is 2 message, and it is 2 that summit 63 and summit 2081 receive Update attribute after each summit message coalescing of message；
Third round：The summit receiving message in second wheel does not have next summit can send message, that is, do not have new information,
So iteration terminates.
Final result is：Summit 891 property value is 0, and the property value on summit 2343,6427,442,9712 is 1 (i.e. summit
891 just can reach these points through a step), the property value on summit 63 and summit 2081 is 2, and the property value on other summits is -1
(i.e. summit 891 can not reach these points).
Based on the above-mentioned message synchronization method identical inventive concept based on figure computing engines, the embodiment of the present invention provide
A kind of message synchronization system based on figure computing engines, as shown in figure 5, include：Message transmission module 101, at each
In subgraph, the message that current vertex is received is sent to next summit by pre-defined rule along the side of current vertex；Message collects
Module 102, if having principal vertex to index for next summit, according to the principal vertex index on next summit, next summit is received
To message be sent to the principal vertex on next summit；Message Fusion Module 103, the principal vertex for merging next summit receives
All message；Message synchronization module 104, for the principal vertex according to next summit from vertex index, disappearing after merging
Breath be sent to next summit all from summit；Cycle criterion module 105, if for not meeting end condition, return execution
The function of message transmission module.
Message synchronization system based on figure computing engines provided in an embodiment of the present invention, during message synchronization, passes through
Principal vertex indexes and quickly finds position between different subgraphs for the summit it is no longer necessary to calculate from vertex index, improves not
The efficiency of synchrodata between with subgraph.
Wherein, message transmission module 101 specifically for：If meeting trigger condition, message that current vertex is received
It is sent to next summit by pre-defined rule along the side of current vertex.
Further, message transmission module 101 specifically for：If while being activated, message that current vertex is received by
Pre-defined rule is sent to next summit along the side of current vertex.
Further, message transmission module 101 specifically for：If current vertex receives message, activate current vertex；
If current vertex is activated, according to when activation condition activation current vertex is corresponding；If the corresponding side of current vertex is swashed
Live, then the message that current vertex receives is sent to next summit by pre-defined rule along the side of current vertex；By current vertex
State of activation reset.
Wherein, also include attribute update module, be used for：The genus of the principal vertex according to next summit of information updating after merging
Property；Principal vertex according to next summit from vertex index, the attribute of the principal vertex on next summit after updating is sent to down
One summit all from summit；The attribute of the principal vertex according to next summit after updating, update next summit from summit
Attribute.
Finally it should be noted that：Various embodiments above only in order to technical scheme to be described, is not intended to limit；To the greatest extent
Pipe has been described in detail to the present invention with reference to foregoing embodiments, it will be understood by those within the art that：Its according to
So the technical scheme described in foregoing embodiments can be modified, or wherein some or all of technical characteristic is entered
Row equivalent；And these modifications or replacement, do not make the essence of appropriate technical solution depart from various embodiments of the present invention technology
The scope of scheme, it all should be covered in the middle of the claim of the present invention and the scope of description.
Claims (10)
1. a kind of message synchronization method based on figure computing engines is it is characterised in that include：
Step S1, in each subgraph, the message that current vertex is received is sent out along the side of described current vertex by pre-defined rule
Give next summit；
Step S2, if next summit described has principal vertex to index, according to the principal vertex index on next summit described, under described
The message that one summit receives is sent to the principal vertex on next summit described；
Step S3, merges all message that the principal vertex on next summit described receives；
Step S4, the principal vertex according to next summit described from vertex index, by the message after merging be sent to described next
Summit all from summit；
Step S5, if not meeting end condition, returns execution step S1.
2. method according to claim 1 is it is characterised in that the described message receiving current vertex is by pre-defined rule
Side along described current vertex is sent to next summit, including：If meeting trigger condition, message that current vertex is received
It is sent to next summit by pre-defined rule along the side of described current vertex.
If 3. method according to claim 2, it is characterised in that described meet trigger condition, current vertex is received
To message be sent to next summit by pre-defined rule along the side of described current vertex, including：If while being activated, by current vertex
The message receiving is sent to next summit by pre-defined rule along the side of described current vertex.
If 4. method according to claim 3 is it is characterised in that side is activated, message that current vertex is received by
Pre-defined rule is sent to next summit along the side of described current vertex, including：
If current vertex receives message, activate described current vertex；
If described current vertex is activated, according to when the activation condition described current vertex of activation is corresponding；
If the corresponding side of described current vertex is activated, message that current vertex is received is by pre-defined rule along described current
The side on summit is sent to next summit；
The state of activation of described current vertex is reset.
5. method according to any one of claim 1 to 4 is it is characterised in that also include：
The attribute of the principal vertex on next summit described in information updating after described fusion；
Principal vertex according to next summit described from vertex index, the attribute of the principal vertex on next summit described in after updating
Be sent to next summit described all from summit；
The attribute of the principal vertex on next summit according to after renewal, updates the attribute from summit on next summit described.
6. a kind of message synchronization system based on figure computing engines is it is characterised in that include：
Message transmission module, in each subgraph, message that current vertex is received is by pre-defined rule along described current
The side on summit is sent to next summit；
Message summarizing module, if having principal vertex to index for next summit described, according to the principal vertex rope on next summit described
Draw, the message that next summit described is received is sent to the principal vertex on next summit described；
Message Fusion Module, all message that the principal vertex for merging next summit described receives；
Message synchronization module, for the principal vertex according to next summit described from vertex index, the message after merging is sent
To next summit described all from summit；
Cycle criterion module, if for not meeting end condition, return the function of execution message transmission module.
7. system according to claim 6 it is characterised in that described message transmission module specifically for：If meeting triggering
Condition, then be sent to next summit by pre-defined rule along the side of described current vertex by the message that current vertex receives.
8. system according to claim 7 it is characterised in that described message transmission module specifically for：If while being activated,
The message that current vertex is received is sent to next summit by pre-defined rule along the side of described current vertex.
9. system according to claim 8 it is characterised in that described message transmission module specifically for：
If current vertex receives message, activate described current vertex；
If described current vertex is activated, according to when the activation condition described current vertex of activation is corresponding；
If the corresponding side of described current vertex is activated, message that current vertex is received is by pre-defined rule along described current
The side on summit is sent to next summit；
The state of activation of described current vertex is reset.
10. the system according to any one of claim 6 to 9, it is characterised in that also including attribute update module, is used for：
The attribute of the principal vertex on next summit described in information updating after described fusion；
Principal vertex according to next summit described from vertex index, the attribute of the principal vertex on next summit described in after updating
Be sent to next summit described all from summit；
The attribute of the principal vertex on next summit according to after renewal, updates the attribute from summit on next summit described.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610822882.9A CN106446104A (en) | 2016-09-13 | 2016-09-13 | Message synchronizing method and system based on graph calculation engine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610822882.9A CN106446104A (en) | 2016-09-13 | 2016-09-13 | Message synchronizing method and system based on graph calculation engine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106446104A true CN106446104A (en) | 2017-02-22 |
Family
ID=58169072
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610822882.9A Pending CN106446104A (en) | 2016-09-13 | 2016-09-13 | Message synchronizing method and system based on graph calculation engine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106446104A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090019096A1 (en) * | 2003-06-18 | 2009-01-15 | International Business Machines Corporation | System and article of manufacture for mirroring data at storage locations |
| CN104063507A (en) * | 2014-07-09 | 2014-09-24 | 时趣互动（北京）科技有限公司 | Graph computation method and engine |
-
2016
- 2016-09-13 CN CN201610822882.9A patent/CN106446104A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090019096A1 (en) * | 2003-06-18 | 2009-01-15 | International Business Machines Corporation | System and article of manufacture for mirroring data at storage locations |
| CN104063507A (en) * | 2014-07-09 | 2014-09-24 | 时趣互动（北京）科技有限公司 | Graph computation method and engine |
Non-Patent Citations (2)
| Title |
|---|
| 小猫1324: "《"http://wenku.baidu.com/view/1b59212e8bd63186bdebbc82.html"》", 25 July 2016 * |
| 秦续业: "面向大规模复杂数据处理的超图并行计算框架", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106649391B (en) | Method and device for processing graph data | |
| CN109146447A (en) | The transparent sharding method of block chain, apparatus and system | |
| CN104298541A (en) | Data distribution algorithm and data distribution device for cloud storage system | |
| CN109697500B (en) | Data processing method and device, electronic equipment and storage medium | |
| CN103699618B (en) | The generation method and system of data sheet | |
| CN102752198A (en) | Multi-core message forwarding method, multi-core processor and network equipment | |
| CN104915251A (en) | Task scheduling method and device | |
| CN109190326B (en) | Method and device for generating process flow chart | |
| CN104572811A (en) | Intelligent recommendation method for bus route, device and equipment | |
| CN106033442B (en) | A kind of parallel breadth first search method based on shared drive architecture | |
| US20130138686A1 (en) | Device and method for arranging query | |
| CN105991478A (en) | Server resource distribution method and system | |
| CN102916879A (en) | Rapid route convergence method | |
| CN105701128B (en) | A kind of optimization method and device of query statement | |
| CN106156245B (en) | Line element merging method and device in electronic map | |
| CN102420797A (en) | Topology mapping method and system | |
| CN108920105A (en) | Diagram data distributed storage method and device based on community structure | |
| CN108563697A (en) | A kind of data processing method, device and storage medium | |
| CN104283966A (en) | Data distribution algorithm and device of cloud storage system | |
| CN105719118B (en) | Multi-target logistics scheduling method and system based on graph theory | |
| CN115470236A (en) | Multi-subgraph matching method, device and equipment | |
| CN104301228A (en) | Method and system for realizing searching module in network processor | |
| CN106446104A (en) | Message synchronizing method and system based on graph calculation engine | |
| CN104572774A (en) | Searching method and device | |
| CN116227566A (en) | Calculation graph visualization method, device, equipment and medium applied to AI chip |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170222 |