CN106383700A - Cloud work flow distributed executing method - Google Patents
Cloud work flow distributed executing method Download PDFInfo
- Publication number
- CN106383700A CN106383700A CN201610756284.6A CN201610756284A CN106383700A CN 106383700 A CN106383700 A CN 106383700A CN 201610756284 A CN201610756284 A CN 201610756284A CN 106383700 A CN106383700 A CN 106383700A
- Authority
- CN
- China
- Prior art keywords
- region
- activity
- execution
- time
- workflow
- 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
- G06F8/00—Arrangements for software engineering
- G06F8/20—Software design
- G06F8/22—Procedural
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a cloud work flow distributed executing method. Calculation is conducted to information such as calling time of each area when cloud work flows of multi-area service are called; different activities can be distributed to servers in different areas under cloud calculation environment and then execution is conducted, so executing time of the cloud work flow can be optimized. A comparison is made between algorithm and effects of the cloud work flow executed in the fixed area via experiments; and a result proves that the distributed executing method has faster executing speed compared with an executing strategy in the fixed area.
Description
Technical field
The present invention relates to cloud computing service field, in particular it relates to a kind of distributed execution method of cloud workflow.
Background technology
Most cloud computing service business, in order to provide service rapidly and efficiently, generally can set up in multiple regions at present
Data center, and provide a user with the service in each region.User, when starting Cloud Server, can freely select cloud to take
The regional location of the true main frame that business device reality is run on.
Simultaneously in cloud workflow, performed activity is all kinds of Web being provided on cloud service and the Internet mostly
Service, although these services can be conducted interviews in nearly all area mostly by the Internet, in the behind of the Internet, this
A little services remain on the server being deployed in specific region.When zones of different calls different services according to connection speed often
Have certain difference calling speed, when especially for there being mass data to need being transmitted of task, data passes
The defeated time may become most of movable execution time.Now on the server in what region, execution task can be greatly
The execution time of degree ground impact workflow activities.
And the development with global IT application, increasing affairs start to be related to trans-regional service call, and not
It is confined to the service of single area again.Taking scientific algorithm workflow as a example, in order to carry out a certain scientific calculation, may first to go many
The data center in individual region obtains data set, then hands over the public computational service providing to a certain laboratory to be calculated, more finally
Result is returned local.This simple calculation workflow is not only related to the service in multiple regions, wherein also has substantial amounts of
Data exchange, now using the same area server execute all tasks may due to data transfer consume substantial amounts of when
Between, and if obtaining and submit data to if the server using corresponding region, server transmission data can be reduced in a large number
Time.And for the data exchange between the Cloud Server of zones of different, due to the line optimization within cloud computing so that this
One time is compared to the service called outside cloud much faster.
For this kind of cloud workflow having across multiple regional activities, just can be by different activities in workflow be divided
The server being fitted in zones of different is improving the treatment effeciency of workflow.And how to select to close to each activity in cloud workflow
The server in suitable region, is the distributed executive problem of cloud workflow activities.
Content of the invention
For defect of the prior art, it is an object of the invention to provide a kind of cloud workflow based on dynamic programming point
Cloth executes method.
For solve above-mentioned technical problem, the present invention provide a kind of cloud workflow distributed execution method, first this
The bright distributed execution to sequential working stream proposes an optimal route derivation algorithm based on dynamic programming.Algorithm complex
For O (m × n2).Again based on propose on the basis of this algorithm be applied to ordinary circumstance comprise with or node workflow
Derivation algorithm is distributed in region, and by determining the fast area of key event to reduce the complexity of algorithm, makes this general use
The complexity of algorithm maintains O (m × n2), and solve approximate optimal path.
A kind of distributed execution method of the cloud workflow being provided according to the present invention, comprises the steps:
Step 1, calculates from hiS is called in regionjService call time T required for servicec(hi,sj) and by serial number j
Data involved by the data volume of movable desired data is from hiArea transmissions are to hjTransmission time T consuming required for regiont(hi,
hj,sj);
Wherein, hiRepresent the server zone that the activity of serial number i is allocated to, sjActivity for serial number j will be called
Service;I and the sequence number of j all expression activities, i, j=1,2 ..., n, n represent the movable number in workflow;hjRepresent serial number j
The server zone that is allocated to of activity;
Step 2, carrys out writing task stream with the form T [n] [M] of a n row M row and executes to this activity during each activity in each area
The shortest execution time of workflow when executing under domain, wherein, M is number of regions;Row represents regional, and row represent in workflow
Each activity, the activity that the time that records in the Unit Cell of form T [n] [M] represents respective column is worked as in the region execution of corresponding row
The shortest execution time of front workflow;Preserve corresponding workflow execution zone routing;T [n] [M] is initially empty and according to Tc
(hi,sj) and Tt(hi,hj,sj) fill up first row.
Preferably, in step 2, described according to Tc(hi,sj) and Tt(hi,hj,sj) method of filling up first row is:
By hiIt is set to the initiation region of workflow, make j=0, sjIt is the service that first activity of workflow is called,
hjRepresent the region performed by first activity of workflow, make hjTravel through all regions, in region m, even hjDuring=m, take Tc
(hi,sj) and Tt(hi,hj,sj) sum insert the 0th row m row.Wherein, the 0th row are first.
Preferably, in step 2, when do not run into or during node, whole workflow is wall scroll sequential working stream, before calculating
The shortest under the k of region of one activity finish the time add transmission time from this region k to target area as from k region to
The full execution time of target area, when k is traveled through each execution to obtain all regions to target area from the individual region of 1 to M
Between, take the short value in these execution times to be used as current active and finish the time in the shortest of target area:
T [n] [m]=min t [k] | t [k]=T [n-1] [k]+Tt(m,k,n)+Tc(k, n), k=1,2 ..., M }
Wherein, n is to work as prostatitis, deputy activity, and m is current line, represents region;K travels through to M from 1, and representative is traveling through
Region;Result of calculation represents current the shortest when m area executes of workflow n-th activity needing to insert T [n] [m] and holds
The row time.
Preferably, in step 2, when run into or during node, adopt with the following method:
With or node x execution time TxFor:
Tx=min { Tc(k, x) | k=1,2 ... M }
Wherein, x be with or node be located activity sequence number;Tc(k, x) represents clothes needed for x-th activity is called in k region
The time of business;Above formula as takes x-th activity of execution region the fastest to execute region as this activity, is designated as hx;
Calculate T [x] [k] as follows:
T[x][hx]=min { t [hx]|t[hx]=T [x-1] [hx]+Tt(k,hx,x)+Tc(hx, x), k=1 ... M }
Wherein, T [x] [hx] to x-th activity and x-th activity is in region h for the current workflow execution of expressionxExecution when
Between, t [hx] represent that each branch executes to hxShortest time, T [x-1] [hx] represent current workflow execution to x-th activity -1
And x-th activity -1 is in region hxThe time of execution, Tt(k,hx, x) represent and x-th movable desired data reached area from region k
Domain hxThe required time, Tc(hx, x) represent x-th activity in region hxThe time of execution, M represents region quantity;
Xth is made to arrange except hxTable cell intermediate value beyond row is just infinite, and that is, x-th activity is only in hxRegion
Execution, does not consider the possible row that x-th activity executes in other regions;
T [x] [m]=∞, m ≠ hx
Directly reach branch and converge node y, y is the activity sequence number converged representated by node;
T [y] [k] is calculated by following formula:
T [y] [k]=max T ' [p] [k] | p=1,2 ... Px, k=1,2 ... M
Wherein, T [y] [k] represents the time that the extremely movable y of current workflow execution and movable y execute in region k;Px
Represent the numbers of branches producing from x node, p travels through to P from 1x, that is, travel through each branch, T ' [p] [k] represents branch p and saves in y
Point meets at the execution time of region k;
The result obtaining be each bar forked working stream run to converge node and converge that node executes in the k of region when
Between, take wherein time-consuming branch's execution time at most as from or node x to the shortest execution time converging node y, as
Need to insert the data of T [y] [k].
Preferably, including:
Step 3, repeat step 2 by the current leu of the form time each row node of column count to the right, until reaching workflow end
Active node, the shortest execution time Tm:
Tm=min T [k] [m] | k=1,2 ... M }
And return the shortest execution time TmThe each activity of corresponding workflow executes region to obtain the shortest point of cloud workflow
Cloth execution region distribution.
Compared with prior art, beneficial effects of the present invention are as follows:With respect to execution cloud workflow under fixed area
Mode, distributed execution cloud workflow can realize faster service call speed by each activity is selected with suitable region
Degree, and the data transmission period between equilibrium region, thus shorten the execution time of whole cloud workflow.
Specific embodiment
With reference to specific embodiment, the present invention is described in detail.Following examples will be helpful to the technology of this area
Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill to this area
For personnel, without departing from the inventive concept of the premise, some changes and improvements can also be made.These broadly fall into the present invention
Protection domain.
Firstly for sequential working stream, cloud workflow execution Time Calculation such as following formula:
Tw=∑ (Tt(hi-1,hi,di)+Tc(hi,si))+σ (1)
T in formulawI.e. total cost time of cloud workflow execution.
hiThe server zone being allocated to for movable i.diData volume for movable i desired data.siWanted by movable i
The service called.
Tt(hi-1,hi,di) it is transmission time, represent diData from hi-1Area transmissions are to hiConsume required for region
Transmission time.
Tc(hi,si) it is the service call time, represent from hiS is called in regioniThe service call time required for service.
σ is remaining operation time sum of server internal.Because this is not with server zone or service region
Different and change, therefore replaced with a constant.
In this formula, service s is called in each activityiWith desired data amount diKnown quantity all can be considered as.TtWith TcThis two
Function can be estimated by prior test or usage history data.Therefore unique it needs to be determined that only each activity distributed
Region hi.By determining h for each activityiTo make whole TwMinimum.We carry out the optimum of this problem using following algorithm
Change and solve
Algorithm 1:The shortest execution time region distribution (the Seq-Workflow Activity of sequential working stream
Allocation)
Step 1, calculates from hiS is called in regionjService call time T required for servicec(hi,sj) and by serial number j
Data involved by the data volume of movable desired data is from hiArea transmissions are to hjTransmission time T consuming required for regiont(hi,
hj,sj);
Step 2, hiRepresent the server zone that the activity of serial number i is allocated to, sjActivity for serial number j will be adjusted
Service;I and the sequence number of j all expression activities, i, j=1,2 ..., n, n represent the movable number in workflow;hjRepresent sequence number
The server zone that activity for j is allocated to;
Afterwards, carry out writing task stream with the form T [n] [M] of a n row M row to execute to this activity during each activity in each area
The shortest execution time of workflow when executing under domain, wherein, M is number of regions, and n is the movable number in workflow;Row represents each
Region, row represent each activity in workflow, and the activity that the time that records in the Unit Cell of form T [n] [M] represents respective column exists
The shortest execution time of work at present stream during the region execution of corresponding row;Preserve corresponding workflow execution zone routing;T[n]
[M] is initially empty and according to Tc(hi,sj) and Tt(hi,hj,sj) fill up first row.
According to Tc(hi,sj) and Tt(hi,hj,sj) method of filling up first row is:
By hiIt is set to the initiation region of workflow, make j=0, sjIt is the service that first activity of workflow is called,
hjRepresent the region performed by first activity of workflow, make hjTravel through all regions, in region m, even hjDuring=m, take Tc
(hi,sj) and Tt(hi,hj,sj) sum insert the 0th row m row.
When do not run into or during node, whole workflow is wall scroll sequential working stream, calculates previous item activity under the k of region
The shortest finish the time and add transmission time from this region k to target area as the full execution from k region to target area
Time, k traversal to be obtained each execution time of all regions to target area from the individual region of 1 to M, to take these execution times
In short value be used as current active and finish the time in the shortest of target area:
T [n] [m]=min t [k] | t [k]=T [n-1] [k]+Tt(m,k,n)+Tc(k, n), k=1,2 ..., M }
Wherein, n is to work as prostatitis, deputy activity, and m is current line, represents region;K travels through to M from 1, and representative is traveling through
Region;Result of calculation represents current the shortest when m area executes of workflow n-th activity needing to insert T [n] [m] and holds
The row time.
Step 3, repeat step 2 by the current leu of the form time each row node of column count to the right, until reaching workflow end
Active node, the shortest execution time Tm:
Tm=min T [k] [m] | k=1,2 ... M }
And return the shortest execution time TmThe each activity of corresponding workflow executes region to obtain the shortest point of cloud workflow
Cloth execution region distribution.
Due to needing to calculate every lattice data of whole m × n form in algorithm, it is required to calculate previous when calculating every lattice data
The data of the n of row, the therefore time complexity of algorithm are O (m × n2).M is the movable number calling service in whole cloud workflow
Amount, n is all available server zone quantity.Due to the quantity very little of usual Free Region, the therefore complexity of algorithm not
Height, in TcWith TtCan be in the hope of optimum workflow the shortest execution time path in the case of known.
And for with or node cloud workflow, process only one layer with or during node, due to need iteration with
Or node is in the execution time in each region inferior division path, every time during iterative calculation individual path, complexity is identical with algorithm 1
For O (m × n2), the total time complexity of therefore all iteration is O (n × m × n2)=O (m × n3).If additionally, in branch
Exist again with or node, then need many one layer of recurrence, therefore in this time complexity, index also can rise.If one has
K layer branched structure, that complexity can reach O (m × nk+2).
And if we algorithm run into or during node, directly pass through one area of the quick determination of algorithm of a Constant Grade
Domain as with or node execution region, abandon its situation in the execution of other regions, just can eliminate this circular recursion structure,
Algorithm complex is made to return to O (m × n2).Because the transmission speed within cloud computing is quickly so that transmission between Cloud Server
Compare that the service call time is much smaller the time, the optimal path calculating when therefore most is all to be assigned to activity
Call this service region server the fastest.So we completely can using this strategy quickly select with or node divided
The execution region joined.And for other sequential organizations, still carry out dynamic programming calculating using algorithm 1.
This pair with band with or node cloud workflow approximation algorithm as described below, wherein Work flow model w
[1 ... n] is stored using recursive structure:
Algorithm 2:Approximate the shortest execution time region distribution (Workflow Activity Allocation) of workflow
Step 1 is identical with the step 1 of algorithm 1.
Step 2 is identical with the step 2 of algorithm 1.
Step 3, when run into or during node, adopt with the following method:
With or node x execution time TxFor:
Tx=min { Tc(k, x) | k=1,2 ... M }
Wherein, x be with or node be located activity sequence number;Tc(k, x) represents clothes needed for x-th activity is called in k region
The time of business;
This formula as takes x-th activity of execution region the fastest to execute region as this activity, is designated as hx;
Calculate T [x] [k] as follows:
T[x][hx]=min { t [hx]|t[hx]=T [x-1] [hx]+Tt(k,hx,x)+Tc(hx, x), k=1 ... M }
Wherein, T [x] [hx] to x-th activity and x-th activity is in region h for the current workflow execution of expressionxExecution when
Between, t [hx] represent that each branch executes to hxShortest time, T [x-1] [hx] represent current workflow execution to x-th activity -1
And x-th activity -1 is in region hxThe time of execution, Tt(k,hx, x) represent and x-th movable desired data reached area from region k
Domain hxThe required time, Tc(hx, x) represent x-th activity in region hxThe time of execution, M represents region quantity;
Xth is made to arrange except hxTable cell intermediate value beyond row is just infinite, and that is, x-th activity is only in hxRegion
Execution, does not consider the possible row that x-th activity executes in other regions;
T [x] [m]=∞, m ≠ hx
To by x corresponding with or node produce forked working stream recurrence calculated using algorithm 2;Directly reach point
Zhi Huihe node y, y are the activity sequence number converged representated by node;
T [y] [k] is calculated by following formula:
T [y] [k]=max T ' [p] [k] | p=1,2 ... Px, k=1,2 ... M
Wherein, T [y] [k] represents the time that the extremely movable y of current workflow execution and movable y execute in region k;Px
Represent the numbers of branches producing from x node, p travels through to P from 1x, that is, travel through each branch, T ' [p] [k] represents branch p and saves in y
Point meets at the execution time of region k;
The result obtaining be each bar forked working stream run to converge node and converge that node executes in the k of region when
Between, take wherein time-consuming branch's execution time at most as from or node x to the shortest execution time converging node y, as
Need to insert the data of T [y] [k].
Step 4, repeat step 2 and step 3 by the form current leu time each row node of column count to the right, until reaching work
Stream end active node, the shortest execution time Tm:
Tm=min T [k] [m] | k=1,2 ... M }
And return the shortest execution time TmThe each activity of corresponding workflow executes region to obtain the shortest point of cloud workflow
Cloth execution region distribution.
Due to or the approximate quick value taken of node, therefore all branches are only needed to recurrence and execute once,
Without repeating, each activity all only calculates once, and therefore Algorithms T-cbmplexity is still O (m × n2).And for sky
Between complexity, if calculating a movable time each, if remaining into the path till this activity, i.e. each activity needs
Preserve all shortest time execution routes under this movable regional, therefore space complexity is O (m × m × n)=O
(m2×n).
Above the specific embodiment of the present invention is described.It is to be appreciated that the invention is not limited in above-mentioned
Particular implementation, those skilled in the art can make a variety of changes within the scope of the claims or change, this not shadow
Ring the flesh and blood of the present invention.In the case of not conflicting, feature in embodiments herein and embodiment can any phase
Mutually combine.
Claims (5)
1. a kind of distributed execution method of cloud workflow is it is characterised in that comprise the steps:
Step 1, calculates from hiS is called in regionjService call time T required for servicec(hi,sj) and the activity by serial number j
Data involved by the data volume of desired data is from hiArea transmissions are to hjTransmission time T consuming required for regiont(hi,hj,
sj);
Wherein, hiRepresent the server zone that the activity of serial number i is allocated to, sjFor the activity of serial number j clothes to be called
Business;I and the sequence number of j all expression activities, i, j=1,2 ..., n, n represent the movable number in workflow;hjRepresent the work of serial number j
The dynamic server zone being allocated to;
Step 2, carrys out writing task stream with the form T [n] [M] of a n row M row and executes to this activity during each activity under each region
The shortest execution time of workflow during execution, wherein, M is number of regions;Row represents regional, and row represent each in workflow
Activity, in the Unit Cell of form T [n] [M], the record time represents the activity of the respective column current work in the region execution of corresponding row
Make the shortest execution time flowed;Preserve corresponding workflow execution zone routing;T [n] [M] is initially empty and according to Tc(hi,sj)
And Tt(hi,hj,sj) fill up first row.
2. the distributed execution method of cloud workflow according to claim 1 it is characterised in that in step 2, described
According to Tc(hi,sj) and Tt(hi,hj,sj) method of filling up first row is:
By hiIt is set to the initiation region of workflow, make j=0, sjIt is the service that first activity of workflow is called, hjGeneration
Region performed by first activity of table workflow, makes hjTravel through all regions, in region m, even hjDuring=m, take Tc(hi,
sj) and Tt(hi,hj,sj) sum insert the 0th row m row.
3. the distributed execution method of cloud workflow according to claim 1 is it is characterised in that in step 2, when not meeting
To with or during node, whole workflow is wall scroll sequential working stream, calculates the shortest under the k of region of previous item activity and finishes the time
Plus the transmission time from this region k to target area as the full execution time from k region to target area, k is traveled through from 1
To obtain each execution time of all regions to target area to M region, to take short value in these execution times to make
Finish the time for current active in the shortest of target area:
T [n] [m]=min t [k] | t [k]=T [n-1] [k]+Tt(m,k,n)+Tc(k, n), k=1,2 ..., M }
Wherein, n is to work as prostatitis, deputy activity, and m is current line, represents region;K travels through to M from 1, represents the area traveling through
Domain;Result of calculation represents the current the shortest execution when m area executes for workflow n-th activity needing to insert T [n] [m]
Time.
4. the distributed execution method of cloud workflow according to claim 1 is it is characterised in that in step 2, when running into
With or during node, adopt with the following method:
With or node x execution time TxFor:
Tx=min { Tc(k, x) | k=1,2 ... M }
Wherein, x be with or node be located activity sequence number;Tc(k, x) represent from k region call x-th movable required service when
Between;Above formula as takes x-th activity of execution region the fastest to execute region as this activity, is designated as hx;
Calculate T [x] [k] as follows:
T[x][hx]=min { t [hx]|t[hx]=T [x-1] [hx]+Tt(k,hx,x)+Tc(hx, x), k=1 ... M }
Wherein, T [x] [hx] to x-th activity and x-th activity is in region h for the current workflow execution of expressionxThe time of execution, t
[hx] represent that each branch executes to hxShortest time, T [x-1] [hx] represent current workflow execution to x-th activity -1 and the
X activity -1 is in region hxThe time of execution, Tt(k,hx, x) represent and x-th movable desired data reached region h from region kx
The required time, Tc(hx, x) represent x-th activity in region hxThe time of execution, M represents region quantity;
Xth is made to arrange except hxTable cell intermediate value beyond row is just infinite, and that is, x-th activity is only in hxRegion executes,
Do not consider the possible row that x-th activity executes in other regions;
T [x] [m]=∞, m ≠ hx
Directly reach branch and converge node y, y is the activity sequence number converged representated by node;
T [y] [k] is calculated by following formula:
T [y] [k]=max T ' [p] [k] | p=1,2 ... Px, k=1,2 ... M
Wherein, T [y] [k] represents the time that the extremely movable y of current workflow execution and movable y execute in region k;PxRepresent from
The numbers of branches that x node produces, p travels through to P from 1x, that is, travel through each branch, T ' [p] [k] represents branch p and converges in y node
Execution time in region k;
The result obtaining is each bar forked working stream and runs to converging node and converge the time that node executes in the k of region,
Take wherein time-consuming branch's execution time at most as from or node x to the shortest execution time converging node y, as need
Insert the data of T [y] [k].
5. the distributed execution method of cloud workflow according to claim 1 is it is characterised in that include:
Step 3, repeat step 2 is by the form current leu time each row node of column count to the right, movable until reaching workflow end
Node, the shortest execution time Tm:
Tm=min T [k] [m] | k=1,2 ... M }
And return the shortest execution time TmTo obtain, cloud workflow is the shortest distributed to be held in corresponding workflow each activity execution region
Row region is distributed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610756284.6A CN106383700B (en) | 2016-08-29 | 2016-08-29 | Distributed execution method of cloud workflow |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610756284.6A CN106383700B (en) | 2016-08-29 | 2016-08-29 | Distributed execution method of cloud workflow |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106383700A true CN106383700A (en) | 2017-02-08 |
CN106383700B CN106383700B (en) | 2020-07-14 |
Family
ID=57915994
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610756284.6A Active CN106383700B (en) | 2016-08-29 | 2016-08-29 | Distributed execution method of cloud workflow |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106383700B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110238458A1 (en) * | 2010-03-24 | 2011-09-29 | International Business Machines Corporation | Dynamically optimized distributed cloud computing-based business process management (bpm) system |
CN103699446A (en) * | 2013-12-31 | 2014-04-02 | 南京信息工程大学 | Quantum-behaved particle swarm optimization (QPSO) algorithm based multi-objective dynamic workflow scheduling method |
CN104536806A (en) * | 2014-12-26 | 2015-04-22 | 东南大学 | Workflow application flexible resource supplying method in cloud environment |
CN105354085A (en) * | 2015-10-30 | 2016-02-24 | 广东石油化工学院 | Scheduling method for cloud workflow jobs |
-
2016
- 2016-08-29 CN CN201610756284.6A patent/CN106383700B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110238458A1 (en) * | 2010-03-24 | 2011-09-29 | International Business Machines Corporation | Dynamically optimized distributed cloud computing-based business process management (bpm) system |
CN103699446A (en) * | 2013-12-31 | 2014-04-02 | 南京信息工程大学 | Quantum-behaved particle swarm optimization (QPSO) algorithm based multi-objective dynamic workflow scheduling method |
CN104536806A (en) * | 2014-12-26 | 2015-04-22 | 东南大学 | Workflow application flexible resource supplying method in cloud environment |
CN105354085A (en) * | 2015-10-30 | 2016-02-24 | 广东石油化工学院 | Scheduling method for cloud workflow jobs |
Non-Patent Citations (1)
Title |
---|
JIAN CAO等: "An Optimal Engine Component Placement Strategy for Cloud Workflow Service", 《2016 IEEE INTERNATIONAL CONFERENCE ON WEB SERVICES (ICWS)》 * |
Also Published As
Publication number | Publication date |
---|---|
CN106383700B (en) | 2020-07-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101494590B (en) | Optimum path selection method of communication network based on load balance | |
CN106062732B (en) | Data processing system, calculate node and the method for data processing | |
US20160173404A1 (en) | Method to jointly select cloud computing and network services and associated device | |
Steiner et al. | Computing all efficient solutions of the biobjective minimum spanning tree problem | |
CN106407207A (en) | Real-time added data updating method and apparatus | |
CN104735166B (en) | The Skyline method for service selection annealed based on MapReduce and multi-target simulation | |
CN103546583B (en) | Group intellectual perception system and group intellectual perception method | |
Blazewicz et al. | Metaheuristic approaches for the two-machine flow-shop problem with weighted late work criterion and common due date | |
Bergman et al. | Lagrangian bounds from decision diagrams | |
CN113572697A (en) | Load balancing method based on graph convolution neural network and deep reinforcement learning | |
Lopez-Pajares et al. | The disjoint multipath challenge: Multiple disjoint paths guaranteeing scalability | |
Khlopin | Necessity of limiting co-state arcs in Bolza-type infinite horizon problem | |
CN114461858A (en) | Causal relationship analysis model construction and causal relationship analysis method | |
CN109657801A (en) | Shunt method, device and the readable storage medium storing program for executing of recommender system | |
CN106383700A (en) | Cloud work flow distributed executing method | |
JP4517166B1 (en) | Method and program for reverse estimation of OD traffic volume between large zones using centroid | |
CN105045906A (en) | Estimation method and device of click rate of delivery information | |
Choo et al. | Reliable vehicle selection algorithm with dynamic mobility of vehicle in vehicular cloud system | |
Brito et al. | An exact algorithm for the stratification problem with proportional allocation | |
CN103108044A (en) | Web service combination method based on dependence graph reducing and quality of service (QoS) holding | |
CN108737130B (en) | Network flow prediction device and method based on neural network | |
CN105389337A (en) | Method for searching big data space for statistical significance mode | |
Lu et al. | Cost-efficient resource provision for multiple mobile users in fog computing | |
Hüsig et al. | Empirical analysis and classification of innovation management software | |
Zhang et al. | Enhanced branch-and-bound framework for a class of sequencing problems |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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 |