CN109656798A - Vertex reordering-based big data processing capability test method for supercomputer - Google Patents
Vertex reordering-based big data processing capability test method for supercomputer Download PDFInfo
- Publication number
- CN109656798A CN109656798A CN201811600894.2A CN201811600894A CN109656798A CN 109656798 A CN109656798 A CN 109656798A CN 201811600894 A CN201811600894 A CN 201811600894A CN 109656798 A CN109656798 A CN 109656798A
- Authority
- CN
- China
- Prior art keywords
- vertex
- degree
- current
- son
- cur
- 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
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3409—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment
- G06F11/3419—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment by assessing time
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3409—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for testing the big data processing capacity of a super computer based on vertex reordering, and aims to improve the speed of testing the big data processing capacity of the super computer. The technical scheme includes that a graph is generated, an adjacency matrix of the graph is constructed, vertexes in the graph are sequenced based on vertex degrees, BFS searching is conducted on the graph by utilizing a sequenced vertex set, the characteristic that the vertex with high degrees has high probability of edge association is utilized in the same-layer traversal, child nodes of nodes in the current-layer vertex set are searched in a traversal mode, the vertex with high degrees is detected in a traversal mode preferentially, and invalid access traversal is reduced to the maximum extent. The invention can improve the hit rate of the edge relation between the nodes, reduce the invalid access times, avoid unnecessary access to the memory to the maximum extent, accelerate the traversal of the graph and improve the test speed of the big data processing capacity of the supercomputer.
Description
Technical field
It is espespecially a kind of to be surpassed based on what vertex was reordered the present invention relates to supercomputer big data processing capacity test method
Grade computer big data processing capacity test method.
Background technique
Graph structure is one of most important data structure in big data application, is widely used in various fields,
Such as social media, bioinformatics, astrophysics, artificial intelligence, data mining.The common feature of these applications is several
According to amount greatly and structure is complicated, often can achieve billions of a sides and many trillion node, this cause data store and calculating
There is higher demand in terms of power.Supercomputer is mainly used for numerical value calculating, and most of HPC benchmark tests are all to calculate power
As measurement standard, the HPL used such as Top 500；In the big data era that data-intensive applications rise extensively, Graph
500 important supplement as Top500 is the new benchmark for testing supercomputer computing capability.Graph 500
With the big data of quantity TEPS (Traversed Edge Per Second) the Lai Hengliang supercomputer on side in traversing graph per second
Processing capacity, and the pretreatment before traversal is not counted in time-consuming.
Graph500 benchmark is generated by figure, figure is established, BFS search exports four parts and form with verifying, result,
As shown in Figure 1.
(1) figure generates: program generates a series of side tuple informations by Kronecker diagram generator, the scale of figure by with
Parameter SCALE, edegefactor of family input determines, wherein SCALE indicates the vertex scale of figure, edegefactor instruction
The par on each vertex connection side, N=2sCALEIndicate that the vertex number of input figure, M=edgefactor*N indicate input
The number of edges mesh of figure.
(2) figure is established: the vertex that figure generation phase generates and side information are converted to the data of any expression figure by this process
Structure stores figure information using the adjacency matrix of figure in standard graph500.
(3) BFS (Breadth-First Search) search and verifying: a root vertex is generated at random, and as source
Point carries out BFS search to entire figure, records the forerunner vertex on each vertex, and output spanning tree records BFS and search as search result
Rope time t, and the obtained BFS spanning tree of verification search whether with original image information matches.The process will recycle 64 times, and right respectively
Each BFS searches for part timing.
(4) result exports: for the number of edges TEPS of Graph500 traversal per second come the execution performance of measurement procedures, TEPS=is raw
At the number of edges M of figure divided by BFS search time t, that is, 64 times searching loop calculates separately TEPS=M/t, then takes 64 TEPS's
Foundation of the average value as Graph500 final test and ranking.
Scheming G=(V, E) includes vertex set V and line set E, usually using viThe vertex that number is i in expression figure uses
Vertex is to (vi,vj) indicate vertex i to the side of vertex j.(vi,vj) ∈ E, 0≤i≤NV- 1,0≤j≤NV- 1, NVFor vertex in V
Number.G usually indicates with adjacency matrix A, the i-th row A in AiFor adjacency list.As shown in Fig. 2, the figure G shaped like Fig. 2 (a) is available
The adjacency matrix A of Fig. 2 (b) indicates, the elements A of the i-th row jth column in AijIndicate side (vi,vj).Indicate that there are this usually using 1
The side of sample, 0 indicates that such side is not present.
In conclusion needing once successively to traverse vertex all in V in BFS search in Graph500, examination and root knot
Relationship between point.Therefore, memory bandwidth is the key factor for influencing the performance of Graph500.Also, the measurement of Graph500
Only using BFS search time as measure time according in, the diagram data before not limiting BFS prepares and pretreatment time.
Experiment shows that Graph500 test performance is primarily limited to memory size and memory bandwidth, and bandwidth is higher, and performance is also got over
It is good, unnecessary memory access how is reduced when memory bandwidth is constant, improves effective memory access frequency, promotes Graph500 test performance
Technical problem as those skilled in the art's urgent need to resolve.
Summary of the invention
The technical problem to be solved in the present invention is that: time-consuming and degree is not counted in using the pretreatment before BFS traversal
Have the characteristics that the probability of frontier juncture system is high between high vertex and root node, proposes a kind of supercomputing reordered based on vertex
Machine big data processing capacity test method reduces memory access number, avoids invalid memory access to greatest extent, makes pair compared to Graph500
Graph traversal accelerates, and improves supercomputer big data processing capacity test speed.
The specific technical proposal is:
The first step, figure generate.Random graph structure G=(V, E) is generated using Kronecker diagram generator, V is vertex set
It closes, wherein including NVA vertex, NVFor positive integer；E is line set；It include N in EESide, NEFor positive integer；
The adjacency matrix A of second step, building storage figure G.AijThere is no side, A between=0 expression vertex i and vertex jij=1
Indicate there is side, 0≤i≤N between vertex i and vertex jV- 1,0≤j≤NV- 1, i and j are positive integer；
Third step pre-processes V based on degree of vertex, and the specific method is as follows:
3.1. each vertex in V is traversed, and records each degree of vertex, obtains in degree of vertex set D, D i-th yuan
Plain deg (vi) indicate vertex viDegree, that is, have deg (vi) a vertex and vertex viBetween have side；
3.2. it sorts to the vertex in V: descending sort being carried out to the element in D using Bubble Sort Algorithm, is sorted
Degree of vertex binary group set D2 afterwards,I-th of element in D2
< vi,deg(vi) > expression vertex viDegree be deg (vi), and meetVertex in V is ranked up according to D2, is sorted
Vertex set Deg afterwards,First element v in Deg0Corresponding vertex degree is maximum
Vertex, second element v1Corresponding vertex degree is only smaller than or is equal to the maximum vertex of degree, and the identical vertex of degree weighs side by side
It lists again,The smallest vertex of corresponding vertex degree；
4th step carries out BFS search to figure G using the vertex set Deg after sequence, and the specific method is as follows:
4.1. data structure definition, the specific method is as follows:
4.1.1. not visited vertex set V is definedns=V；
4.1.2. set D-tmp=D2 among the degree of vertex of BFS search is defined；
4.1.3. set Deg-tmp=Deg among the degree of vertex after definition sequence；
4.1.4. the vertex set being accessed is defined
4.1.5. defining current layer vertex set
4.1.6. current level of child nodes set is defined
4.1.7. defining child node setIndicate vertex viChild node set；
4.1.8. selecting a vertex v at random in VrAs root vertex, i.e. source summit, r=0,1 ..., NV；
4.1.9. the set of the child node set of root vertex is enabledSonrIn element be set；
4.1.10. by vertex vrIt is added in the vertex set being accessed, Vs=Vs+{Vr}；
4.1.11. by vertex vrIt is added in current vertex set, i.e. Cur=Cur+ { Vr}；
4.2. it loops through, one cycle exports a spanning tree, recycles 64 times, exports 64 spanning trees, specific method
It is as follows:
4.2.1. defining cyclic variable k=0；
4.2.2. obtaining system time t1；
If 4.2.3. k < 64, turn 4.3；Otherwise, turn the 5th step；
4.3. same layer traverses, and haves the characteristics that the associated probability in side is also high using the high vertex of degree, in traversal search Cur
The child node of node, the specific method is as follows:
4.3.1. enabling
4.3.2. if4.3.3 is executed, otherwise, the search of this layer is completed, and is turned 4.4 pairs of next layers and is traversed；
4.3.3. appoint in Cur and take a vertex vi, it is denoted as current root node vcs, cs=0,1 ..., NV；
4.3.4. v is deleted from Curcs, i.e. Cur=Cur- { vcs}；
4.3.5. ifIt executes 4.3.6 first traversal and checks the high vertex of degree, otherwise, current root section
Point search finishes, and turns 4.3.20；
4.3.6. inquiry D2 set, finds binary group, confirm vcsDegree be deg
(vcs)；
4.3.7. from not visited vertex set VnsIt is middle to delete current root node vcs, i.e. Vns=Vns-{vcs}；
4.3.8. defining cyclic variable m=0；
If 4.3.9. m < deg (vcs), it executes 4.3.10 and has otherwise found all sides of current vertex totally, turn
4.3.16 v is checkedcsAdjacent node, i.e. other element vertex in Cur；
4.3.10. first vertex is selected from Deg-tmp, is enabled as vj, i.e., the current highest vertex of degree；
4.3.11. adjacency matrix A is inquired, if Aij=1, indicate vertex viWith vertex vjBetween have a side, execute 4.3.12, it is no
Then, turn 4.3.14；
4.3.12. the existing associated vertex in side, i.e. Deg-tmp=Deg-tmp- { v are deleted from set Deg-tmpj}；
If 4.3.13. vj∈Vns, by vertex vjFrom VnsMiddle deletion, i.e. Vns=Vns-{vj, directly turn 4.3.14；Otherwise, it says
It is brightIt does not need from VnsMiddle deletion, turns 4.3.14；
4.3.14. updating current root node vcs, i.e. vertex viChild node set, that is, Soni=Soni+{vj}；
4.3.15. the child node set of current layer is updated, that is, L-Son=L-Son+ { vj}；
4.3.16. Son is added in the child node set of current root node in the form of set elementrIn, i.e. Sonr=Sonr
+{Soni}；
4.3.17.m=m+1；
4.3.18. by vertex vjFrom VnsMiddle deletion, i.e. Vns=Vns-{vj}；
4.3.19. ifTurn 4.3.9, otherwise, the vertex traversal of all not visited mistakes finishes, and turns
4.3.20；
4.3.20. root node v before being deleted from current vertex set Curcs, i.e. Cur=Cur- { vcs}；
4.3.21. turn 4.3.1；
4.4. interlayer traverses, and the specific method is as follows:
4.4.1. current layer vertex set is emptied, current vertex set is reset
4.4.2. current level of child nodes set L-Son is assigned to current vertex set, i.e. Cur=L-Son；
4.4.3. obtaining system time t2；
4.4.4. the time t=t of the heuristic traversal search of record kth time2-t1；
4.4.5. turn 4.3.2；
4.5. root vertex set Son is exportedr, SonrAs kth wheel circulation is with vertex vrBFS's as root vertex
Spanning tree rootk-r；
4.6. test performance is calculated.Calculate the BFS traversal test performance value of current spanning tree
4.7. turn 4.2.2；
5th step, the evaluation of estimate for calculating figure test performance, i.e., the BFS traversal test performance value average value of 64 spanning treesObtain test result.The Large Scale Graphs processing capacity of the higher surface supercomputer of TEPS value
Stronger, Graph500 ranking is also more forward, while also reflecting that the supercomputer is more suitable to be handled with big data.
6th step terminates.
Following technical effect can achieve using the present invention:
It is ranked up 1. third step of the present invention is based on degree of vertex opposite vertexes, pre-processes, can be kept away to greatest extent as BFS
Exempt from the invalid memory access of BFS, promotes the hit rate that BFS is effectively traversed, optimize BFS traversal efficiency；
2. the 4th step of the invention improves the hit rate of frontier juncture system between node based on the BFS search that vertex is reordered, reduce
Invalid memory access number avoids unnecessary memory access to greatest extent, accelerates to improve supercomputer big data to graph traversal
Processing capacity test speed.
Detailed description of the invention
Fig. 1 is Graph500 test benchmark program flow diagram；
Fig. 2 is that the adjacency matrix of figure indicates schematic diagram；Fig. 2 (a) is an oriented no weight graph；Fig. 2 (b) is the neighbour of Fig. 2 (a)
Connect matrix.
Fig. 3 is overview flow chart of the present invention.
Specific embodiment
Fig. 3 is overview flow chart of the present invention, and step of the invention is as follows:
The first step, figure generate.Random graph structure G=(V, E) is generated using Kronecker diagram generator, V is vertex set
It closes, wherein including NVA vertex, NVFor positive integer；E is line set；
The adjacency matrix A of second step, building storage figure G.AijThere is no side, A between=0 expression vertex i and vertex jij=1
Indicate there is side, 0≤i≤N between vertex i and vertex jV- 1,0≤j≤NV- 1, i and j are positive integer；
Third step pre-processes V based on degree of vertex, and the specific method is as follows:
3.1. each vertex in V is traversed, and records each degree of vertex, obtains degree of vertex set D,
I-th of element deg (v in Di) indicate vertex viDegree, that is, have deg (vi) a vertex and vertex viBetween have side；
3.2. it sorts to the vertex in V: descending sort being carried out to the element in D using Bubble Sort Algorithm, is sorted
Degree of vertex binary group set D2 afterwards,, i-th yuan in D2
Plain < vi,deg(vi) > expression vertex viDegree be deg (vi), and meetVertex in V is ranked up according to D2, is sorted
Vertex set Deg afterwards,First element v in Deg0Corresponding vertex degree is maximum
Vertex, second element v1Corresponding vertex degree is only smaller than or is equal to the maximum vertex of degree, and the identical vertex of degree weighs side by side
It lists again,The smallest vertex of corresponding vertex degree；
4th step carries out BFS search to figure G using the vertex set Deg after sequence, and the specific method is as follows:
4.1. data structure definition, the specific method is as follows:
4.1.12. not visited vertex set V is definedns=V；
4.1.13. set D-tmp=D2 among the degree of vertex of BFS search is defined；
4.1.14. set Deg-tmp=Deg among the degree of vertex after definition sequence；
4.1.15. defining the vertex set being accessed
4.1.16. defining current layer vertex set
4.1.17. defining current level of child nodes set
4.1.18. defining child node setIndicate vertex viChild node set；
4.1.19. a vertex v is selected at random in VrAs root vertex, i.e. source summit, r=0,1 ..., NV；
4.1.20. the set of the child node set of root vertex is enabledSonrIn element be set；
4.1.21. by vertex vrIt is added in the vertex set being accessed, Vs=Vs+{Vr}；
4.1.22. by vertex vrIt is added in current vertex set, i.e. Cur=Cur+ { Vr}；
4.2. it loops through, one cycle exports a spanning tree, recycles 64 times, exports 64 spanning trees, specific method
It is as follows:
4.2.4. defining cyclic variable k=0；
4.2.5. obtaining system time t1；
If 4.2.6. k < 64, turn 4.3；Otherwise, turn the 5th step；
4.3. same layer traverses, and haves the characteristics that the associated probability in side is also high using the high vertex of degree, in traversal search Cur
The child node of node, the specific method is as follows:
4.3.1. enabling
4.3.2. if4.3.3 is executed, otherwise, the search of this layer is completed, and is turned 4.4 pairs of next layers and is traversed；
4.3.3. appoint in Cur and take a vertex vi, it is denoted as current root node vcs, cs=0,1 ..., NV；
4.3.4. v is deleted from Curcs, i.e. Cur=Cur- { vcs}；
4.3.5. ifIt executes 4.3.6 first traversal and checks the high vertex of degree, otherwise, current root section
Point search finishes, and turns 4.3.20；
4.3.6. inquiry D2 set, finds binary group, confirm vcsDegree be deg
(vcs)；
4.3.7. from not visited vertex set VnsIt is middle to delete current root node vcs, i.e. Vns=Vns-{vcs}；
4.3.8. defining cyclic variable m=0；
If 4.3.9. m < deg (vcs), it executes 4.3.10 and has otherwise found all sides of current vertex totally, turn
4.3.16 v is checkedcsAdjacent node, i.e. other element vertex in Cur；
4.3.10. first vertex is selected from Deg-tmp, is enabled as vj, i.e., the current highest vertex of degree；
4.3.11. adjacency matrix A is inquired, if Aij=1, indicate vertex viWith vertex vjBetween have a side, execute 4.3.12, it is no
Then, turn 4.3.14；
4.3.12. the existing associated vertex in side, i.e. Deg-tmp=Deg-tmp- { v are deleted from set Deg-tmpj}；
If 4.3.13. vj∈Vns, by vertex vjFrom VnsMiddle deletion, i.e. Vns=Vns-{vj, directly turn 4.3.14；Otherwise, it says
It is brightIt does not need from VnsMiddle deletion, turns 4.3.14；
4.3.14. updating current root node vcs, i.e. vertex viChild node set, that is, Soni=Soni+{vj}；
4.3.15. the child node set of current layer is updated, that is, L-Son=L-Son+ { vj}；
4.3.16. Son is added in the child node set of current root node in the form of set elementrIn, i.e. Sonr=Sonr
+{Soni}；
4.3.17.m=m+1；
4.3.18. by vertex vjFrom VnsMiddle deletion, i.e. Vns=Vns-{vj}；
4.3.19. ifTurn 4.3.9, otherwise, the vertex traversal of all not visited mistakes finishes, and turns
4.3.20；
4.3.20. root node v before being deleted from current vertex set Curcs, i.e. Cur=Cur- { vcs}；
4.3.21. turn 4.3.1；
4.4. interlayer traverses, and the specific method is as follows:
4.4.1. current layer vertex set is emptied, current vertex set is reset
4.4.2. current level of child nodes set L-Son is assigned to current vertex set, i.e. Cur=L-Son；
4.4.3. obtaining system time t2；
4.4.4. the time t=t of the heuristic traversal search of record kth time2-t1；
4.4.5. turn 4.3.2；
4.5. root vertex set Son is exportedr, SonrAs kth wheel circulation is with vertex vrBFS's as root vertex
Spanning tree rootk-r；
4.6. test performance is calculated.Calculate the BFS traversal test performance value of current spanning tree
4.7. turn 4.2.2；
5th step, the evaluation of estimate for calculating figure test performance, i.e., the BFS traversal test performance value average value of 64 spanning treesObtain test result.The Large Scale Graphs processing capacity of the higher surface supercomputer of TEPS value
Stronger, Graph500 ranking is also more forward, while also reflecting that the supercomputer is more suitable to be handled with big data.
6th step terminates.
Claims (3)
1. a kind of supercomputer big data processing capacity test method to be reordered based on vertex, it is characterised in that including following
Step:
The first step generates random graph structure G=(V, E), and V is vertex set, wherein including NVA vertex, NVFor positive integer；E is
Line set includes N in EESide, NEFor positive integer；
The adjacency matrix A, A of second step, building storage figure GijThere is no side, A between=0 expression vertex i and vertex jij=1 indicates top
There are side, 0≤i≤N between point i and vertex jV- 1,0≤j≤NV- 1, i and j are positive integer；
Third step pre-processes V based on degree of vertex, and the specific method is as follows:
3.1. each vertex in V is traversed, and records each degree of vertex, obtains i-th of element in degree of vertex set D, D
deg(vi) indicate vertex viDegree, that is, have deg (vi) a vertex and vertex viBetween have side；
3.2. it sorts to the vertex in V: descending sort being carried out to the element in D, the degree of vertex binary group collection after being sorted
D2 is closed,I-th in D2
A element < vi,deg(vi) > expression vertex viDegree be deg (vi), and meetVertex in V is ranked up according to D2, is sorted
Vertex set Deg afterwards,First element v in Deg0Corresponding vertex degree is maximum
Vertex, second element v1Corresponding vertex degree is only smaller than or is equal to the maximum vertex of degree, and the identical vertex of degree weighs side by side
It lists again,The smallest vertex of corresponding vertex degree；
4th step carries out BFS search to figure G using the vertex set Deg after sequence, and the specific method is as follows:
4.1. data structure definition, the specific method is as follows:
4.1.1. not visited vertex set V is definedns=V；
4.1.2. set D-tmp=D2 among the degree of vertex of BFS search is defined；
4.1.3. set Deg-tmp=Deg among the degree of vertex after definition sequence；
4.1.4. the vertex set being accessed is defined
4.1.5. defining current layer vertex set
4.1.6. current level of child nodes set is defined
4.1.7. defining child node setIndicate vertex viChild node set；
4.1.8. a vertex v is selected at random in VrAs root vertex, i.e. source summit, r=0,1 ..., NV；
4.1.9. the set of the child node set of root vertex is enabledSonrIn element be set；
4.1.10. by vertex vrIt is added in the vertex set being accessed, Vs=Vs+{Vr}；
4.1.11. by vertex vrIt is added in current vertex set, i.e. Cur=Cur+ { Vr}；
4.2. it looping through, one cycle exports a spanning tree, and it recycles 64 times, exports 64 spanning trees, the specific method is as follows:
4.2.1. defining cyclic variable k=0；
4.2.2. obtaining system time t1；
If 4.2.3. k < 64, turn 4.3；Otherwise, turn the 5th step；
4.3. same layer traverses, and haves the characteristics that the associated probability in side is also high using the high vertex of degree, traversal search Cur interior joint
Child node, the specific method is as follows:
4.3.1. enabling
4.3.2. if4.3.3 is executed, otherwise, the search of this layer is completed, and is turned 4.4 pairs of next layers and is traversed；
4.3.3. appoint in Cur and take a vertex vi, it is denoted as current root node vcs, cs=0,1 ..., NV；
4.3.4. v is deleted from Curcs, i.e. Cur=Cur- { vcs}；
4.3.5. ifIt executes 4.3.6 first traversal and checks the high vertex of degree, otherwise, current root node is searched
Rope finishes, and turns 4.3.20；
4.3.6. inquiry D2 set, finds binary group, confirm vcsDegree be deg
(vcs)；
4.3.7. from not visited vertex set VnsIt is middle to delete current root node vcs, i.e. Vns=Vns-{vcs}；
4.3.8. defining cyclic variable m=0；
If 4.3.9. m < deg (vcs), it executes 4.3.10 and has otherwise found all sides of current vertex totally, turned 4.3.16
Check vcsAdjacent node, i.e. other element vertex in Cur；
4.3.10. first vertex is selected from Deg-tmp, is enabled as vj, i.e., the current highest vertex of degree；
4.3.11. adjacency matrix A is inquired, if Aij=1, indicate vertex viWith vertex vjBetween have a side, execute 4.3.12, otherwise, turn
4.3.14；
4.3.12. the existing associated vertex in side, i.e. Deg-tmp=Deg-tmp- { v are deleted from set Deg-tmpj}；
If 4.3.13. vj∈Vns, by vertex vjFrom VnsMiddle deletion, i.e. Vns=Vns-{vj, directly turn 4.3.14；Otherwise, explanationIt does not need from VnsMiddle deletion, turns 4.3.14；
4.3.14. updating current root node vcs, i.e. vertex viChild node set, that is, Soni=Soni+{vj}；
4.3.15. the child node set of current layer is updated, that is, L-Son=L-Son+ { vj}；
4.3.16. Son is added in the child node set of current root node in the form of set elementrIn, i.e. Sonr=Sonr+
{Soni}；
4.3.17.m=m+1；
4.3.18. by vertex vjFrom VnsMiddle deletion, i.e. Vns=Vns-{vj}；
4.3.19. ifTurn 4.3.9, otherwise, the vertex traversal of all not visited mistakes finishes, and turns 4.3.20；
4.3.20. root node v before being deleted from current vertex set Curcs, i.e. Cur=Cur- { vcs}；
4.3.21. turn 4.3.1；
4.4. interlayer traverses, and the specific method is as follows:
4.4.1. current layer vertex set is emptied, current vertex set is reset
4.4.2. current level of child nodes set L-Son is assigned to current vertex set, i.e. Cur=L-Son；
4.4.3. obtaining system time t2；
4.4.4. the time t=t of the heuristic traversal search of record kth time2-t1；
4.4.5. turn 4.3.2；
4.5. root vertex set Son is exportedr, SonrAs kth wheel circulation is with vertex vrThe generation of BFS as root vertex
Set rootk-r；
4.6. the BFS traversal test performance value of current spanning tree is calculated
4.7. turn 4.2.2；
5th step, the evaluation of estimate for calculating figure test performance, i.e., the BFS traversal test performance value average value of 64 spanning treesObtain test result；
6th step terminates.
2. the supercomputer big data processing capacity test method to be reordered as described in claim 1 based on vertex, special
Sign is using Kronecker diagram generator generation figure G.
3. the supercomputer big data processing capacity test method to be reordered as described in claim 1 based on vertex, special
Sign is to carry out the element in D to use Bubble Sort Algorithm when descending sort.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811600894.2A CN109656798B (en) | 2018-12-26 | 2018-12-26 | Vertex reordering-based big data processing capability test method for supercomputer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811600894.2A CN109656798B (en) | 2018-12-26 | 2018-12-26 | Vertex reordering-based big data processing capability test method for supercomputer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109656798A true CN109656798A (en) | 2019-04-19 |
| CN109656798B CN109656798B (en) | 2022-02-01 |
Family
ID=66116766
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811600894.2A Active CN109656798B (en) | 2018-12-26 | 2018-12-26 | Vertex reordering-based big data processing capability test method for supercomputer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109656798B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111090651A (en) * | 2019-12-18 | 2020-05-01 | 深圳前海微众银行股份有限公司 | Data source processing method, device and equipment and readable storage medium |
| CN111881327A (en) * | 2020-07-30 | 2020-11-03 | 中国人民解放军国防科技大学 | Big data processing capacity testing method based on vertex reordering and priority caching |
| CN112165405A (en) * | 2020-10-13 | 2021-01-01 | 中国人民解放军国防科技大学 | Method for testing big data processing capacity of supercomputer based on network topological structure |
| CN112883241A (en) * | 2021-03-19 | 2021-06-01 | 中国人民解放军国防科技大学 | Supercomputer benchmark test acceleration method based on connected component generation optimization |
| WO2022082860A1 (en) * | 2020-10-21 | 2022-04-28 | 深圳大学 | Lightweight and efficient graph vertex rearrangement method |
| WO2023236239A1 (en) * | 2022-06-09 | 2023-12-14 | 深圳计算科学研究院 | Multi-round sampling based data screening rule validation method, and apparatus thereof |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180160281A1 (en) * | 2015-06-05 | 2018-06-07 | Technische Universität Kaiserslautern | Automated determination of network motifs |
| CN108520275A (en) * | 2017-06-28 | 2018-09-11 | 浙江大学 | A kind of regular system of link information based on adjacency matrix, figure Feature Extraction System, figure categorizing system and method |
-
2018
- 2018-12-26 CN CN201811600894.2A patent/CN109656798B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180160281A1 (en) * | 2015-06-05 | 2018-06-07 | Technische Universität Kaiserslautern | Automated determination of network motifs |
| CN108520275A (en) * | 2017-06-28 | 2018-09-11 | 浙江大学 | A kind of regular system of link information based on adjacency matrix, figure Feature Extraction System, figure categorizing system and method |
Non-Patent Citations (1)
| Title |
|---|
| 赵波: "基于PageRank的计算机性能评价方法", 《计算机工程》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111090651A (en) * | 2019-12-18 | 2020-05-01 | 深圳前海微众银行股份有限公司 | Data source processing method, device and equipment and readable storage medium |
| CN111090651B (en) * | 2019-12-18 | 2024-03-29 | 深圳前海微众银行股份有限公司 | Data source processing method, device, equipment and readable storage medium |
| CN111881327A (en) * | 2020-07-30 | 2020-11-03 | 中国人民解放军国防科技大学 | Big data processing capacity testing method based on vertex reordering and priority caching |
| CN112165405A (en) * | 2020-10-13 | 2021-01-01 | 中国人民解放军国防科技大学 | Method for testing big data processing capacity of supercomputer based on network topological structure |
| CN112165405B (en) * | 2020-10-13 | 2022-04-22 | 中国人民解放军国防科技大学 | Method for testing big data processing capacity of supercomputer based on network topological structure |
| WO2022082860A1 (en) * | 2020-10-21 | 2022-04-28 | 深圳大学 | Lightweight and efficient graph vertex rearrangement method |
| CN112883241A (en) * | 2021-03-19 | 2021-06-01 | 中国人民解放军国防科技大学 | Supercomputer benchmark test acceleration method based on connected component generation optimization |
| WO2023236239A1 (en) * | 2022-06-09 | 2023-12-14 | 深圳计算科学研究院 | Multi-round sampling based data screening rule validation method, and apparatus thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109656798B (en) | 2022-02-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109656798A (en) | Vertex reordering-based big data processing capability test method for supercomputer | |
| Jiang et al. | Hop doubling label indexing for point-to-point distance querying on scale-free networks | |
| Mehlhorn et al. | External-memory breadth-first search with sublinear I/O | |
| Liu et al. | Finding top-k optimal sequenced routes | |
| Blelloch et al. | Parallel write-efficient algorithms and data structures for computational geometry | |
| CN102333036B (en) | Method and system for realizing high-speed routing lookup | |
| CN108287840A (en) | A kind of data storage and query method based on matrix Hash | |
| Sacharidis et al. | Topologically sorted skylines for partially ordered domains | |
| Maltry et al. | A critical analysis of recursive model indexes | |
| US20220374733A1 (en) | Data packet classification method and system based on convolutional neural network | |
| JP6418431B2 (en) | Method for efficient one-to-one coupling | |
| CN112860692A (en) | Database table structure conversion method and device and electronic equipment thereof | |
| CN107133335A (en) | A kind of repetition record detection method based on participle and index technology | |
| CN109684185A (en) | Heuristic traversal-based big data processing capacity test method for supercomputer | |
| CN108052743B (en) | Method and system for determining step approach centrality | |
| CN107807793B (en) | The storage of data copy isomery and access method in distributed computer storage system | |
| CN103646035A (en) | Information search method based on heuristic method | |
| Singh et al. | High average-utility itemsets mining: a survey | |
| CN108595624A (en) | A kind of large-scale distributed functional dependence discovery method | |
| CN103761298A (en) | Distributed-architecture-based entity matching method | |
| Setayesh et al. | Presentation of an Extended Version of the PageRank Algorithm to Rank Web Pages Inspired by Ant Colony Algorithm | |
| CN111881327A (en) | Big data processing capacity testing method based on vertex reordering and priority caching | |
| CN108280176A (en) | Data mining optimization method based on MapReduce | |
| Rohlf | A new approach to the computation of the Jardine-Sibson Bk clusters | |
| CN105354243A (en) | Merge clustering-based parallel frequent probability subgraph searching 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 |