CN106845642B - A kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule - Google Patents
A kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule Download PDFInfo
- Publication number
- CN106845642B CN106845642B CN201710046208.0A CN201710046208A CN106845642B CN 106845642 B CN106845642 B CN 106845642B CN 201710046208 A CN201710046208 A CN 201710046208A CN 106845642 B CN106845642 B CN 106845642B
- Authority
- CN
- China
- Prior art keywords
- population
- individual
- pareto
- locating
- entropy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Operations Research (AREA)
- Biomedical Technology (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Marketing (AREA)
- Economics (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule, can be improved global detection and the part producing capacity of multi-target evolution method.The described method includes: S1, according to the number of Pareto solution and Pareto entropy detection population locating Evolving State during evolution, according to the population detected locating Evolving State during evolution, adaptively constraint condition is handled using corresponding individual assessment strategy, and the individual in population is ranked up, wherein, in individual assessment strategy, constraint condition is handled using constraint violation processing method;S2, according to individual ranking results, select individual to carry out genetic manipulation from population, obtain sub- population, wherein when carrying out genetic manipulation, according to population during evolution locating for Evolving State evolution parameter is adaptively adjusted.The present invention is suitable for solving the multi-objective optimization question of belt restraining, and can be applied to workflow schedule technical field in cloud computing environment.
Description
Technical field
The present invention relates to the multiple-objection optimizations for solving the problems, such as belt restraining, and are applied to cloud workflow schedule technical field,
Particularly relate to a kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule.
Background technique
Workflow schedule (referred to as: cloud workflow schedule) under cloud environment is to find suitable cloud resource to execute workflow
Task, and meet the QoS requirement of user.Cloud workflow schedule problem is the multi-objective optimization question of a belt restraining, more
Problems can be effectively treated in target evolution algorithm.It but is come simply by static penalty mostly
Constraint condition is handled, is easy to cause Premature Convergence in this way, even into infeasible search space, such as:
The prior art one, by assessing multiple target grain using Pareto (Pareto) entropy information and Pareto difference entropy information
The diversity and Evolving State of population in swarm optimization, and carry out Design evolution strategy as feedback information, so that algorithm
With better convergence and diversity.
The prior art two is improved on the basis of NSGA-II algorithm, by the entire Pareto of discretization it is optimal before
The reference point of some good distributions is found on edge, using these points as the direction of search in the evolutionary process of algorithm, is found and offer
With reference to the solution near the associated Pareto optimal solution of point set or Pareto optimal solution, the disaggregation made, which has, more preferably to be restrained
Property and diversity.
The prior art one and the prior art two, although the non-dominant disaggregation that algorithm can be made to obtain has better convergence
And diversity, but it is directed to unconfined multi-objective optimization question.It is one for the workflow schedule problem under cloud environment
The multi-objective optimization question of a belt restraining, when handling problems, common method is used based on multi-objective Evolutionary Algorithm
Constrained Optimization is converted unconstrained optimization problem by static penalty, if penalty is too small, it is some it is non-can
The fitness value of row solution will be above most of feasible solution, and population is likely to evolve towards a non-feasible search space;But
If penalty is too big, many more preferably individuals will be excluded, so as to cause Premature Convergence, to sum up, existing multiple target
Evolution algorithm is easily trapped into local optimum when handling constraint condition using static penalty.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of adaptive multi-target evolutions of belt restraining cloud workflow schedule
Method is easy with solving multi-objective Evolutionary Algorithm present in the prior art when using static penalty processing constraint condition
The problem of falling into local optimum.
In order to solve the above technical problems, the embodiment of the present invention provides a kind of adaptive more mesh of belt restraining cloud workflow schedule
Mark evolvement method, comprising:
S1, according to the number of Pareto solution and Pareto entropy detection population locating Evolving State during evolution, root
According to the population detected locating Evolving State during evolution, adaptively about using corresponding individual assessment strategy processing
Beam condition, and the individual in population is ranked up, wherein in individual assessment strategy, using constraint violation processing method come
Handle constraint condition;
S2, according to individual ranking results, select individual to carry out genetic manipulation from population, obtain sub- population, wherein into
When row genetic manipulation, according to population, evolution parameter is adaptively adjusted in locating Evolving State during evolution.
Further, it is described according to the number of Pareto solution and Pareto entropy detection population during evolution it is locating into
Change state includes:
If not having Pareto solution in population, locating Evolving State is original state to population during evolution.
Further, it is described according to the number of Pareto solution and Pareto entropy detection population during evolution it is locating into
Change state includes:
If the number that Pareto is solved in population is less than Population Size, locating Evolving State is population during evolution
Convergence state;Or,
If the number that Pareto is solved in population is equal to Population Size, and population is in t+1 iteration, the number that Pareto is solved
Changed, then locating Evolving State is convergence state to population during evolution;Or,
If the number that Pareto is solved in population is equal to Population Size, in t+1 iteration, the number of Pareto solution does not have population
It changes, andThen locating Evolving State is to receive to population during evolution
Hold back state;
Wherein, △ Entropy (t+1) indicates that the poor entropy of the t+1 times iteration and the t times iteration, M indicate of optimization aim
Number, △ Entropymax-diverIndicate maximum poor entropy.
Further, it is described according to the number of Pareto solution and Pareto entropy detection population during evolution it is locating into
Change state includes:
If the number that Pareto is solved in population is equal to Population Size, in t+1 iteration, the number of Pareto solution does not have population
It changes, andThen locating Evolving State is more to population during evolution
Sample state;Or,
If the number that Pareto is solved in population is equal to Population Size, in t+1 iteration, the number of Pareto solution does not have population
It changes, Pareto entropy does not also change, then locating Evolving State is maturity state to population during evolution;
Wherein, △ Entropy (t+1) indicates that the poor entropy of the t+1 times iteration and the t times iteration, M indicate of optimization aim
Number, △ Entropymax-diverIndicate maximum poor entropy.
Further, the △ Entropymax-diverIt indicates are as follows:
Wherein, EntropymaxIndicate Pareto entropy of the Pareto solution to be optimally distributed when, EntropyminIndicate Pareto
Pareto entropy when solution is worst distribution, M indicate the number of optimization aim, and K indicates the grid division under every one-dimensional optimization aim,
Cellk,m(t) the t times iteration is indicatedIn individual amount,It isThe Based on Integer Labelling being mapped in PCCS,Indicate the
M-th of lattice coordinate components of k Pareto solution,Indicate that k-th of Pareto solves the value of corresponding m-th of optimization aim, l table
Show EntropyminIn the case of only one individual that grid coordinate.
Further, the population that the basis detects locating Evolving State during evolution, is adaptive selected
Corresponding individual assessment strategy handles constraint condition, and is ranked up to the individual in population, comprising:
If the Evolving State locating for detecting population during evolution is original state, according to the first fitness function
Assess the ideal adaptation angle value of population, wherein first fitness function indicates are as follows:
The individual in population is ranked up according to the size of ideal adaptation angle value;
Wherein, Fa(xi) indicate individual i fitness value, xiIndividual i is represented, a indicates Fa(xi) it is after having added penalty
Fitness value,It is normalized constraint violation.
Further, the population that the basis detects locating Evolving State during evolution, is adaptive selected
Corresponding individual assessment strategy handles constraint condition, and is ranked up to the individual in population, comprising:
If the Evolving State locating for detecting population during evolution is convergence state, according to the second fitness function
Assess the ideal adaptation angle value of population, wherein second fitness function indicates are as follows:
The individual in population is ranked up according to the size of ideal adaptation angle value;
Wherein,Indicate the fitness value of k-th of target of individual i, xiIndividual i is represented, a is indicatedIt is to add
Fitness value after penalty,Fitness value after indicating the normalization of k-th of target of individual i,Table
Show normalized constraint violation, rfIndicate possible ratios.
Further, the population that the basis detects locating Evolving State during evolution, adaptively utilizes
Corresponding individual assessment strategy, is ranked up the individual in population and includes:
If the Evolving State locating for detecting population during evolution is diversified state or maturity state, based on about
The dominated Sorting principle of beam is ranked up the individual in population.
Further, the evolution parameter includes: crossover probability and mutation probability;
It is described when carrying out genetic manipulation, according to population during evolution locating Evolving State be adaptively adjusted into
Changing parameter includes:
When carrying out genetic manipulation, in genetic manipulation crossover probability and mutation probability carry out adaptive adjustment,
In, the crossover probability p of the t times iterationc(t) adjustment Rule Expression are as follows:
The mutation probability p of the t times iterationm(t) adjustment Rule Expression are as follows:
Wherein,<>indicates a holding pc(t) and pm(t) function among given boundary, works as pc(t) and pm
(t) it is lower than lower boundary, lower border value is given to them, works as pc(t) and pm(t) it is higher than coboundary, upper boundary values is given to it
, pc(t-1) crossover probability of the t-1 times iteration, p are indicatedm(t-1) mutation probability of the t-1 times iteration, △ np (t) table are indicated
Show that the variation number of the Pareto number of individuals in the t times iteration, △ Entropy (t-1) indicate the t-1 times iteration and change for the t-2 times
The poor entropy in generation, ciAnd mjExpression adjusting step, i, j=1,2,3,4.
The advantageous effects of the above technical solutions of the present invention are as follows:
In above scheme, according to the locating evolution during evolution of the number of Pareto solution and Pareto entropy detection population
State adaptively assesses plan using corresponding individual according to the population detected locating Evolving State during evolution
Constraint condition is slightly handled, and the individual in population is ranked up, wherein in individual assessment strategy, at constraint violation
Reason method handles constraint condition, can effectively handle constraint condition;According to individual ranking results, selected from population individual into
Row genetic manipulation obtains sub- population, wherein when carrying out genetic manipulation, according to population locating evolution shape during evolution
Evolution parameter is adaptively adjusted in state, and multi-target evolution method is made to have good convergence and diversity, is solving cloud work
There is preferably performance when flowing the optimization problem of scheduling, so that can also have preferably global detection and part under constraint condition
Producing capacity.
Detailed description of the invention
Fig. 1 is the process of the adaptive multi-target evolution method of belt restraining cloud workflow schedule provided in an embodiment of the present invention
Schematic diagram;
Fig. 2 is the detailed of the adaptive multi-target evolution method of belt restraining cloud workflow schedule provided in an embodiment of the present invention
Flow diagram;
Fig. 3 indicates Epigenomics workflow total executory cost TEC and degree of unbalancedness DI under 1 constraint condition of time limit
Equilibrium solution;
Fig. 4 indicates Epigenomics workflow total executory cost TEC and degree of unbalancedness DI under 2 constraint condition of time limit
Equilibrium solution;
Fig. 5 indicates Epigenomics workflow total executory cost TEC and degree of unbalancedness DI under 3 constraint condition of time limit
Equilibrium solution;
Fig. 6 indicates Epigenomics workflow total executory cost TEC and degree of unbalancedness DI under 4 constraint condition of time limit
Equilibrium solution;
The equilibrium of Fig. 7 expression Inspiral workflow total executory cost TEC and degree of unbalancedness DI under 1 constraint condition of time limit
Solution;
The equilibrium of Fig. 8 shows Inspiral workflows under 2 constraint condition of time limit total executory cost TEC and degree of unbalancedness DI
Solution;
The equilibrium of Fig. 9 expression Inspiral workflow total executory cost TEC and degree of unbalancedness DI under 3 constraint condition of time limit
Solution;
Figure 10 indicate Inspiral workflow under 4 constraint condition of time limit total executory cost TEC and degree of unbalancedness DI it is equal
Weighing apparatus solution.
Specific embodiment
To keep the technical problem to be solved in the present invention, technical solution and advantage clearer, below in conjunction with attached drawing and tool
Body embodiment is described in detail.
The present invention is easy to fall into for existing multi-objective Evolutionary Algorithm when using static penalty processing constraint condition
The problem of entering local optimum provides a kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule.
As shown in Figure 1, the adaptive multi-target evolution method of belt restraining cloud workflow schedule provided in an embodiment of the present invention,
Include:
S1, according to the number of Pareto solution and Pareto entropy detection population locating Evolving State during evolution, root
According to the population detected locating Evolving State during evolution, adaptively about using corresponding individual assessment strategy processing
Beam condition, and the individual in population is ranked up, wherein in individual assessment strategy, using constraint violation processing method come
Handle constraint condition;
S2, according to individual ranking results, select individual to carry out genetic manipulation from population, obtain sub- population, wherein into
When row genetic manipulation, according to population, evolution parameter is adaptively adjusted in locating Evolving State during evolution.
The adaptive multi-target evolution method of belt restraining cloud workflow schedule described in the embodiment of the present invention, according to Pareto
Number and Pareto entropy the detection population of solution locating Evolving State during evolution, were evolving according to the population detected
Locating Evolving State in journey adaptively handles constraint condition using corresponding individual assessment strategy, and in population
Body is ranked up, wherein in individual assessment strategy, constraint condition is handled using constraint violation processing method, it can be effective
Handle constraint condition;According to individual ranking results, selects individual to carry out genetic manipulation from population, obtain sub- population, wherein
When carrying out genetic manipulation, according to population, evolution parameter is adaptively adjusted in locating Evolving State during evolution, makes more mesh
Marking evolvement method has good convergence and diversity, has preferably property when solving the optimization problem of cloud workflow schedule
Can, so that can also have preferably global detection and local producing capacity under constraint condition.
The adaptive multi-target evolution method of belt restraining cloud workflow schedule provided in an embodiment of the present invention is properly termed as base
In NSGA-II algorithm (the Pareto Entropy based on NSGA- with adaptive individual assessment strategy of Pareto entropy
II with adaptive individual-assessment scheme, ai-NSGA-II-PE), ai-NSGA-II-PE with
Multi-objective optimization algorithm (Non-dominated Sorting the Genetic Algorithm-II, NSGA- of non-dominated ranking
II based on), as shown in Fig. 2, the adaptive multi-target evolution side of belt restraining cloud workflow schedule provided in an embodiment of the present invention
Method can specifically include:
S201, initialization population, including the number of iterations, population scale etc.;
S202, according to the number of Pareto solution and Pareto entropy detection population locating Evolving State during evolution,
According to the population detected locating Evolving State during evolution, adaptively handled using corresponding individual assessment strategy
Constraint condition, and the individual in population is ranked up;Wherein, the number and the detection kind of Pareto entropy according to Pareto solution
Group's Evolving State locating in NSGA-II evolutionary process includes:
(1) if not having Pareto solution in population, defining population, locating Evolving State is initial shape during evolution
State;
(2) following 3 kinds of situations, defining population, locating Evolving State is convergence state during evolution:
A, the number that Pareto is solved in population is less than Population Size;
B, the number that Pareto is solved in population is equal to Population Size, but population is in t+1 iteration, the number that Pareto is solved
Changed;
C, the number that Pareto is solved in population is equal to Population Size, and in t+1 iteration, the number of Pareto solution does not have population
It changes, andWherein, △ Entropy (t+1) indicate the t+1 time iteration with
The poor entropy of the t times iteration, M indicate the number of optimization aim, △ Entropymax-diverIndicate maximum poor entropy.
In the present embodiment, Pareto entropy, poor entropy, maximum difference entropy △ Entropy in order to better understandmax-diver, right
Pareto entropy, poor entropy, maximum difference entropy △ Entropymax-diverIt is described in detail:
In the present embodiment, can by parallel lattice coordinate system (Parallel Cell Coordinate System,
PCCS Pareto entropy and its poor entropy) described.M is the number of optimization aim, and the grid division under every one-dimensional optimization aim is K,Indicate that k-th of Pareto solves the value of corresponding m-th of optimization aim,Being mapped to one according to formula (1) has K × M
The two-dimensional surface grid of grid,It isThe Based on Integer Labelling being mapped in PCCS indicates m-th of lattice of k-th of Pareto solution
Coordinate components.The coordinate components of solution in each cartesian coordinate system may be mapped to a certain in two-dimensional surface grid
In a grid.
It enablesWhereinFor the function that rounds up, that is, the smallest positive integral for being not less than x is returned to,WithIt is the maximum value and minimum value for m-th of target in current Pareto solution set respectively,
IfIt willIt is set as 1.
In the present embodiment, the Pareto entropy Entropy (t) of the t times iteration can be indicated are as follows:
Wherein, Cellk,m(t) it indicatesIn individual amount, the poor entropy △ Entropy (t) of the t times and t-1 iteration can
To indicate are as follows:
△ Entropy (t)=Entropy (t)-Entropy (t-1)
Maximum difference entropy △ Entropymax-diverThe difference for referring to entropy under a kind of extreme case, i.e., in diversified state
When, new explanation and it is old solution be in same dominance hierarchy, but have better crowding when, old solution can be replaced, and be madeWithIt is aobvious
Variation is write, solution is caused to become worst distribution from optimal distribution (i.e. the coordinate vector of each target occupies a grid in PCCS)
(i.e. the coordinate vector of each target has K-1 to squeeze in a grid in PCCS, remaining to occupy at another), at this time
Changing value (the △ Entropy of Pareto entropymax-diver) can indicate are as follows:
Wherein, EntropymaxIndicate Pareto entropy of the Pareto solution to be optimally distributed when, EntropyminIndicate Pareto
Pareto entropy when solution is worst distribution, M indicate the number of optimization aim, and K indicates the grid division under every one-dimensional optimization aim,
Cellk,m(t) the t times iteration is indicatedIn individual amount,It isThe Based on Integer Labelling being mapped in PCCS,Indicate the
M-th of lattice coordinate components of k Pareto solution,Indicate that k-th of Pareto solves the value of corresponding m-th of optimization aim, l table
Show EntropyminIn the case of only one individual that grid coordinate.
(3) if the number that Pareto is solved in population is equal to Population Size, population is in t+1 iteration, of Pareto solution
There is no variations for number, andThen define population locating evolution during evolution
State is diversified state;
(4) if the number that Pareto is solved in population is equal to Population Size, population is in t+1 iteration, of Pareto solution
There is no variation, Pareto entropys also not to change for number, then defining population, locating Evolving State is mature shape during evolution
State.
In the present embodiment, then, according to the population detected locating Evolving State during evolution, adaptive land productivity
Constraint condition is handled with corresponding individual assessment strategy, and the individual in population is ranked up.In adaptive individual assessment ring
Section, mainly judges the quality of an individual by two aspects, and one is target value that it is obtained in objective function, another
It is whether the constrained objective value obtained under constraint condition is more than concludeed a contract or treaty beam.
In the present embodiment, individual is assessed using different individual assessment strategies in different Evolving States, to individual into
Row sequence, in individual assessment strategy, constraint condition is handled using constraint violation processing method according to population in evolutionary process
In locating different Evolving States, constraint condition effectively can be handled using different constraint violation processing methods, made
More preferably feasible individual remains;It is specific:
(1) if the Evolving State locating for detecting population during evolution is original state, according to the first fitness
The ideal adaptation angle value of function evaluation population, wherein first fitness function indicates are as follows:
The individual in population is ranked up according to the size of ideal adaptation angle value;
Wherein, Fa(xi) indicate individual i fitness value, xiIndividual i is represented, a indicates Fa(xi) it is after having added penalty
Fitness value,It is normalized constraint violation, i.e., by constrained objective value, (constrained objective value is obtained under constraint condition
To) be more than concludeed a contract or treaty beam part be normalized, then according to normalized constraint violation come give individual sequence so that
Remain the individual for possessing small constraint violation.
(2) if the Evolving State locating for detecting population during evolution is convergence state, i.e., at least one in population
After a feasible solution (feasible solution is referred to as feasible individual), then the ideal adaptation of population is assessed according to the second fitness function
Angle value, wherein second fitness function indicates are as follows:
The individual in population is ranked up according to the size of ideal adaptation angle value;
Wherein,Indicate the fitness value of k-th of target of individual i, xiIndividual i is represented, a is indicatedIt is to add
Fitness value after penalty,Fitness value after indicating the normalization of k-th of target of individual i,Table
Show normalized constraint violation, rfIndicate possible ratios, rfFor the ratio of feasible individual quantity and population scale.
(3) if the Evolving State locating for detecting population during evolution is diversified state or maturity state, this reality
Example is applied using another individual assessment strategy come the individual that sorts, according to defining 1, the dominated Sorting principle based on constraint is as follows, makes
Group hunting must be planted and possess the feasible individual of more preferable crowding, and prevent infeasible solutions from entering population:
Define 1: if any one following condition is very, to illustrate individual S1Than individual S2It is outstanding:
(1) if S1It is feasible solution and S2It is infeasible solutions;
(2)S1、S2It is all infeasible solutions, S1There is smaller constraint violation;
(3)S1、S2It is all feasible solution, S1Compare S2There is higher non-dominant grade;Or, working as S1And S2There is identical domination etc.
Grade, S1Compare S2There is more excellent crowding, individual is ranked up by non-dominant grade, wherein non-dominant grade sequence and gathers around
It is similar with employed in NSGA-II algorithm to squeeze degree calculating;
Non-dominant grade sequence refers to: if one or more optimization target values of some individual i are superior to other one
Individual j, and other optimization target values are equal, then individual i dominates individual j.It is non-dominant to individual distribution according to dominance relation
Grade, the low individual of the high individual dominance hierarchy of grade, if individual m cannot be dominated by any individual, i.e., obtained solution
Optimal, then individual m is referred to as Pareto optimal solution or non-domination solution.
Crowding, for being ranked up to the individual in identical non-dominant grade, basic thought is according to all
The optimization target values of body calculate and the adjacent two individual European space distances of individual.
S203 selects individual to carry out genetic manipulation, obtains sub- population according to individual ranking results from population.Possess row
There is the higher individual of sequence bigger probability to be selected into sub-group.
It is that individual is selected to carry out heredity using different individual assessment strategies in different Evolving States in the present embodiment
Operation.
When according to population, locating Evolving State carries out corresponding genetic manipulation during evolution, population is according to locating for it
Evolving State adaptively carry out the adjustment of evolution parameter, i.e., in genetic manipulation crossover probability and mutation probability carry out from
The adjustment of adaptation, to improve global detection and local producing capacity.
In the present embodiment, crossover probability and mutation probability in genetic manipulation can be according to population institutes during evolution
The Evolving State at place, the number of Pareto solution and Pareto entropy adaptively adjust, wherein the crossover probability of the t times iteration
pc(t) adjustment Rule Expression are as follows:
The mutation probability p of the t times iterationm(t) adjustment Rule Expression are as follows:
Wherein,<>indicates a holding pc(t) and pm(t) function among given boundary, works as pc(t) and pm
(t) it is lower than lower boundary, lower border value is given to them, works as pc(t) and pm(t) it is higher than coboundary, upper boundary values is given to it
, pc(t-1) crossover probability of the t-1 times iteration, p are indicatedm(t-1) mutation probability of the t-1 times iteration, △ np (t) table are indicated
Show that the variation number of the Pareto number of individuals in the t times iteration, △ Entropy (t-1) indicate the t-1 times iteration and change for the t-2 times
The poor entropy in generation, ciAnd mjExpression adjusting step, i, j=1,2,3,4;Wherein, ciAnd mjIt indicates are as follows:
Wherein, gmaxIt is expressed as maximum number of iterations.
S204, according to father population and sub- population locating Evolving State during evolution, adaptive polo placement father population and
The ideal adaptation angle value of sub- population, the same S202 of specific steps.
S205 determines the individual sequence of father population and sub- population according to the ideal adaptation angle value of father population and sub- population.
In the present embodiment, the individual of father population and sub- population is ranked up from small to large according to ideal adaptation angle value,
In, when sequence, father population and sub- population are placed on to be operated together.
S206 selects top n individual as new population, calculates new population from the father population and sub- population after sequence
Pareto entropy and detection Population status, return execute S202, and new population participates in next iteration, preset until the number of iterations is equal to
Maximum number of iterations, wherein the value of N is determines according to actual conditions.
In the present embodiment, from the father population and sub- population after sequence, select top n individual as new population.When sequence,
Father population and sub- population are placed on to be operated together, and choosing individual is also the top n for selecting two populations (father population and sub- population)
Body.
In the case where calculating sensitive cloud computing environment, to solve the problems, such as its workflow schedule, such as drag is established:
Optimization aim (Minimize) is to minimize total executory cost (Total Execution Cost, TEC) and injustice
Weighing apparatus degree (Degree of Imbalance, DI), constraint condition (Subject to) are total execution time (Total Execution
Time, TET) it is less than time limit dw, it is expressed as
Minimize:TEC
DI
Subject to:TET≤dw
Wherein, | VM | indicate the number of virtual machine,Indicate the virtual machine of execution task i,It is virtual machine unit
Time cost,It isThe actual run time of execution task i, tiExpression task i, τ are virtual machine processed in units energy
Power,Transmission time between expression task i and task j, subscript ei,jWhat expression task i and task j was connected to,It isData conversion cost, the T in i ∈ T, j ∈ T refers to all task subscripts, TmaxAnd TminRespectively indicate all virtual machines
Maximum runing time and minimum runing time, TavgIt is the average operating time of virtual machine, V is set of tasks,It indicates
The end time of task i.
In the present embodiment,The actual run time of execution task iIt indicates are as follows:
Wherein,For the size of task i,It isProcessing capacity.
Transmission time in the present embodiment, between task i and task jIt indicates are as follows:
Wherein,WithRespectively indicate virtual machine VMiAnd VMjBetween data transmission bandwidth, subscript ei,jIt indicates to appoint
Business i and task j is connected to,It is the output data size generated by task i, if task i and task j is same virtual
Machine executes, then transmission time is 0;
In the present embodiment, at the beginning of task iIt indicates are as follows:
Wherein,It isIt rents and terminates the time,Refer to the virtual machine of execution task i, taExpression task a,
parent(ti) indicate all female tasks of task i,The end time of expression task a,Expression task a and task i
Between transmission time, subscript ea,iWhat expression task a and task i was connected to.
In the present embodiment, the end time of task iIt indicates are as follows:
Wherein,At the beginning of expression task i,It indicatesThe actual run time of execution task i.
In the present embodiment, 4 time limit value are defined when testing, when these time limit value are in most fast and most run slowly
Between between, most fast runing time by heterogeneous computing environment earliest time complete algorithm (Heterogeneous Earliest
Finish Time, HEFT) traffic control stream obtained, and most slow runing time executes all tasks by a virtual machine and is obtained
, the time limit gradually becomes stringent with the increase of number.
The present embodiment uses two kinds of workflows of Epigenomics and Inspiral and is tested, wherein
Epigenomics workflow is a kind of highly pipelined application program, plurality of line concurrently to independent data block into
Row operation;Inspiral is made of several sub- workflows.Test result such as Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7, Fig. 8, Fig. 9, Figure 10
It is shown.
Embodiment one
In the present embodiment, it is the time limit 1 that time limit value, which is respectively set, and in the time limit 2, in the time limit 3, the time limit 4 chooses research-on-research flow model
Epigenomics, by ai-NSGA-II-PE and NSGA-II algorithm (the ParetoEntropy based on based on Pareto entropy
NSGA-II, NSGA-II-PE) (NSGA-II-PE uses adaptive adjustment evolution parameter proposed by the invention, but uses tradition
Static penalty handles constraint condition), NSGA-II, the multi-objective Evolutionary Algorithm (Multi-Objective based on decomposition
Evolutionary Algorithm based on Decomposition, MOEA/D), strength Pareto evolutionary algorithm
(Strength Pareto Evolutionary Algorithm 2, SPEA2), multi-objective particle (Multi-
Objective Particle Swarm Optimization, MOPSO) it is compared respectively;Simulation result such as Fig. 3, Fig. 4, figure
5, Fig. 6, by Fig. 3, Fig. 4, Fig. 5, Fig. 6 as it can be seen that compared to other algorithms, ai-NSGA-II-PE energy under stringent constraint condition
Find global detection and local the exploitation effect more preferably forward position Pareto.
Embodiment two
In the present embodiment, it is the time limit 1 that time limit value, which is respectively set, and in the time limit 2, in the time limit 3, the time limit 4 chooses research-on-research flow model
Inspiral, by ai-NSGA-II-PE and NSGA-II-PE (use adaptive adjustment evolution parameter proposed by the invention, but
Use traditional static penalty to handle constraint condition), NSGA-II, MOEA/D, SPEA2, MOPSO evolution algorithm carries out respectively
Compare;Simulation result such as Fig. 7, Fig. 8, Fig. 9, Figure 10, by Fig. 7, Fig. 8, Fig. 9, Figure 10 as it can be seen that compared to other algorithms, ai-
NSGA-II-PE can find global detection and local the exploitation effect more preferably forward position Pareto under stringent constraint condition.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, without departing from the principles of the present invention, it can also make several improvements and retouch, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (9)
1. a kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule characterized by comprising
S1, belt restraining cloud workflow schedule model is established:
Minimize:TEC
DI
Subject to:TET≤dw
Wherein, Minimize indicates optimization aim, and TEC indicates total executory cost, and DI indicates degree of unbalancedness, and TEC and DI are 2 excellent
Change target, Subject to indicates constraint condition, and TET indicates total and executes time, dwIndicate the time limit, | VM | indicate of virtual machine
Number,Indicate the virtual machine of execution task i,It is virtual machine long-run cost rate,It isExecution task i
Actual run time, tiExpression task i, τ are virtual machine unit capacities,Biography between expression task i and task j
Defeated time, subscript ei,jWhat expression task i and task j was connected to,It isData conversion cost, in i ∈ T, j ∈ T
T refer to all task subscripts, TmaxAnd TminThe maximum runing time and minimum runing time of all virtual machines are respectively indicated,
TavgIt is the average operating time of virtual machine, V is set of tasks,The end time of expression task i;
Using the corresponding relationship between Pareto solution and the value of optimization aim, detected according to the number of Pareto solution and Pareto entropy
Population locating Evolving State during evolution, according to the population detected locating Evolving State during evolution, from
Constraint condition adaptively is handled using corresponding individual assessment strategy, and the individual in population is ranked up, wherein in individual
In assessment strategy, constraint condition is handled using constraint violation processing method;
S2, according to individual ranking results, select individual to carry out genetic manipulation from population, obtain sub- population, wherein lost
When passing operation, according to population, evolution parameter is adaptively adjusted in locating Evolving State during evolution, until the number of iterations
Equal to preset maximum number of iterations.
2. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 1, which is characterized in that
It is described that according to the number of Pareto solution and Pareto entropy detection population, locating Evolving State includes: during evolution
If not having Pareto solution in population, locating Evolving State is original state to population during evolution.
3. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 1, which is characterized in that
It is described that according to the number of Pareto solution and Pareto entropy detection population, locating Evolving State includes: during evolution
If the number that Pareto is solved in population is less than Population Size, locating Evolving State is convergence to population during evolution
State;Or,
If the number that Pareto is solved in population is equal to Population Size, and population, in t+1 iteration, the number of Pareto solution occurs
Variation, then locating Evolving State is convergence state to population during evolution;Or,
If the number that Pareto is solved in population is equal to Population Size, population in t+1 iteration, do not send out by the number of Pareto solution
Changing, andThen locating Evolving State is convergence shape to population during evolution
State;
Wherein, Δ Entropy (t+1) indicates that the poor entropy of the t+1 times iteration and the t times iteration, M indicate the number of optimization aim,
ΔEntropymax-diverIndicate maximum poor entropy.
4. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 1, which is characterized in that
It is described that according to the number of Pareto solution and Pareto entropy detection population, locating Evolving State includes: during evolution
If the number that Pareto is solved in population is equal to Population Size, population in t+1 iteration, do not send out by the number of Pareto solution
Changing, andThen locating Evolving State is diversification to population during evolution
State;Or,
If the number that Pareto is solved in population is equal to Population Size, population in t+1 iteration, do not send out by the number of Pareto solution
Changing, Pareto entropy also do not change, then locating Evolving State is maturity state to population during evolution;
Wherein, Δ Entropy (t+1) indicates that the poor entropy of the t+1 times iteration and the t times iteration, M indicate the number of optimization aim,
ΔEntropymax-diverIndicate maximum poor entropy.
5. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 3 or 4, feature exist
In the Δ Entropymax-diverIt indicates are as follows:
Wherein, EntropymaxIndicate Pareto entropy of the Pareto solution to be optimally distributed when, EntropyminIndicate that Pareto solution is
Pareto entropy when worst distribution, M indicate the number of optimization aim, and K indicates the grid division under every one-dimensional optimization aim,
Cellk,m(t) the t times iteration is indicatedIn individual amount,It isThe Based on Integer Labelling being mapped in PCCS,Indicate the
M-th of lattice coordinate components of k Pareto solution,Indicate that k-th of Pareto solves the value of corresponding m-th of optimization aim, l table
Show EntropyminIn the case of only one individual that grid coordinate.
6. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 1, which is characterized in that
The population that the basis detects locating Evolving State during evolution, is adaptive selected corresponding individual assessment strategy
Constraint condition is handled, and the individual in population is ranked up, comprising:
If the Evolving State locating for detecting population during evolution is original state, assessed according to the first fitness function
The ideal adaptation angle value of population, wherein first fitness function indicates are as follows:
The individual in population is ranked up according to the size of ideal adaptation angle value;
Wherein, Fa(xi) indicate individual i fitness value, xiIndividual i is represented, a indicates Fa(xi) it is suitable after having added penalty
Angle value is answered,It is normalized constraint violation.
7. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 1, which is characterized in that
The population that the basis detects locating Evolving State during evolution, is adaptive selected corresponding individual assessment strategy
Constraint condition is handled, and the individual in population is ranked up, comprising:
If the Evolving State locating for detecting population during evolution is convergence state, assessed according to the second fitness function
The ideal adaptation angle value of population, wherein second fitness function indicates are as follows:
The individual in population is ranked up according to the size of ideal adaptation angle value;
Wherein,Indicate the fitness value of k-th of optimization aim of individual i, xiIndividual i is represented, a is indicatedIt is to add
Fitness value after penalty,Fitness value after indicating the normalization of k-th of optimization aim of individual i,Indicate normalized constraint violation, rfIndicate possible ratios.
8. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 1, which is characterized in that
The population that the basis detects locating Evolving State during evolution, adaptively assesses plan using corresponding individual
Slightly, the individual in population is ranked up and includes:
If the Evolving State locating for detecting population during evolution is diversified state or maturity state, based on constraint
Dominated Sorting principle is ranked up the individual in population.
9. the adaptive multi-target evolution method of belt restraining cloud workflow schedule according to claim 1, which is characterized in that
The evolution parameter includes: crossover probability and mutation probability;
It is described when carrying out genetic manipulation, according to population, evolution ginseng is adaptively adjusted in locating Evolving State during evolution
Number includes:
When carrying out genetic manipulation, in genetic manipulation crossover probability and mutation probability carry out adaptive adjustment, wherein
The crossover probability p of t iterationc(t) adjustment Rule Expression are as follows:
The mutation probability p of the t times iterationm(t) adjustment Rule Expression are as follows:
Wherein,<>indicates a holding pc(t) and pm(t) function among given boundary, works as pc(t) and pm(t) low
In lower boundary, lower border value is given to them, works as pc(t) and pm(t) it is higher than coboundary, upper boundary values is given to them, pc
(t-1) crossover probability of the t-1 times iteration, p are indicatedm(t-1) indicate that the mutation probability of the t-1 times iteration, Δ np (t) indicate
The variation number of Pareto number of individuals in the t times iteration, Δ Entropy (t-1) indicate the t-1 times iteration and the t-2 times iteration
Poor entropy, ciAnd mjExpression adjusting step, i, j=1,2,3,4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710046208.0A CN106845642B (en) | 2017-01-22 | 2017-01-22 | A kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710046208.0A CN106845642B (en) | 2017-01-22 | 2017-01-22 | A kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106845642A CN106845642A (en) | 2017-06-13 |
CN106845642B true CN106845642B (en) | 2019-03-29 |
Family
ID=59120017
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710046208.0A Active CN106845642B (en) | 2017-01-22 | 2017-01-22 | A kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106845642B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108650030B (en) * | 2018-05-07 | 2021-05-11 | 上海电力学院 | Water surface multi-sink node deployment method of underwater wireless sensor network |
CN109361725B (en) * | 2018-08-14 | 2020-12-08 | 西安电子科技大学 | Internet of vehicles cloud system resource allocation method based on multi-target genetic algorithm |
US11379728B2 (en) | 2019-01-09 | 2022-07-05 | Red Hat, Inc. | Modified genetic recombination operator for cloud optimization |
CN109992355B (en) * | 2019-01-30 | 2021-04-13 | 北京理工大学 | Multi-target cloud workflow scheduling method based on improved non-dominated genetic algorithm |
CN110119558B (en) * | 2019-04-30 | 2023-11-14 | 广州大学 | Circuit parameter optimization method based on differential optimization algorithm |
CN110111146B (en) * | 2019-04-30 | 2023-07-11 | 东华大学 | Multi-objective optimization solving method oriented to cloud instance selection |
CN112907057B (en) * | 2021-02-06 | 2022-11-11 | 北京理工大学 | Production scheduling optimization method and system based on improved MOPSO algorithm |
CN113419854B (en) * | 2021-06-23 | 2023-09-22 | 东北大学 | Unbalanced multi-objective optimization-oriented cloud resource scheduling method |
CN113947705B (en) * | 2021-10-25 | 2022-08-12 | 四川大学 | Target detection method based on evolutionary neural network under constraint condition |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103019822A (en) * | 2012-12-07 | 2013-04-03 | 北京邮电大学 | Large-scale processing task scheduling method for income driving under cloud environment |
CN105046408A (en) * | 2015-06-25 | 2015-11-11 | 国网山东省电力公司 | Configurable workflow realization method and system |
CN105354085A (en) * | 2015-10-30 | 2016-02-24 | 广东石油化工学院 | Scheduling method for cloud workflow jobs |
CN105790266A (en) * | 2016-04-27 | 2016-07-20 | 华东交通大学 | Microgrid parallel multi-target robust optimization scheduling integrated control method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8762522B2 (en) * | 2011-04-19 | 2014-06-24 | Cisco Technology | Coordinating data center compute and thermal load based on environmental data forecasts |
-
2017
- 2017-01-22 CN CN201710046208.0A patent/CN106845642B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103019822A (en) * | 2012-12-07 | 2013-04-03 | 北京邮电大学 | Large-scale processing task scheduling method for income driving under cloud environment |
CN105046408A (en) * | 2015-06-25 | 2015-11-11 | 国网山东省电力公司 | Configurable workflow realization method and system |
CN105354085A (en) * | 2015-10-30 | 2016-02-24 | 广东石油化工学院 | Scheduling method for cloud workflow jobs |
CN105790266A (en) * | 2016-04-27 | 2016-07-20 | 华东交通大学 | Microgrid parallel multi-target robust optimization scheduling integrated control method |
Non-Patent Citations (4)
Title |
---|
A Survey on Virtual Machine Scheduling in Cloud Computing;Li Liu等;《IEEE》;20161231;全文 |
Small is better:avoiding latency trap in virtualized data centers;Yunjing Xu;《ACM》;20131031;全文 |
云环境中基于分解的多目标工作流调度算法;李克武等;《计算机工程与科学》;20160831;第38卷(第8期);全文 |
求多目标优化问题Pareto最优解集的方法;王海军等;《大学数学》;20081031;第24卷(第5期);全文 |
Also Published As
Publication number | Publication date |
---|---|
CN106845642A (en) | 2017-06-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106845642B (en) | A kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule | |
CN111065103B (en) | Multi-objective optimization wireless sensor network node deployment method | |
Liaw et al. | Evolutionary manytasking optimization based on symbiosis in biocoenosis | |
CN106484512B (en) | The dispatching method of computing unit | |
Sindhya et al. | A hybrid framework for evolutionary multi-objective optimization | |
CN106155791B (en) | A kind of workflow task dispatching method under distributed environment | |
CN108133260A (en) | The workflow schedule method of multi-objective particle swarm optimization based on real-time status monitoring | |
CN104932938B (en) | A kind of cloud resource dispatching method based on genetic algorithm | |
Eskandari et al. | FastPGA: A dynamic population sizing approach for solving expensive multiobjective optimization problems | |
CN105929690B (en) | A kind of Flexible Workshop Robust Scheduling method based on decomposition multi-objective Evolutionary Algorithm | |
CN109670650A (en) | The method for solving of Cascade Reservoirs scheduling model based on multi-objective optimization algorithm | |
CN112685138B (en) | Multi-workflow scheduling method based on multi-population hybrid intelligent optimization in cloud environment | |
CN110909787A (en) | Method and system for multi-objective batch scheduling optimization based on clustering evolutionary algorithm | |
CN108564592A (en) | Based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic | |
CN106960244A (en) | A kind of genetic algorithm and the evolution algorithm of particle cluster algorithm Parallel Fusion | |
CN107016461A (en) | One kind mixing multi-target evolution method | |
CN102621887B (en) | Garment sewing process analysis and combination method | |
CN109298930A (en) | A kind of cloud workflow schedule method and device based on multiple-objection optimization | |
Dorronsoro et al. | Multi-objective robust static mapping of independent tasks on grids | |
CN116627241A (en) | Method, system, equipment and storage medium for optimizing energy consumption of server | |
KR101580202B1 (en) | Svm-bta updating apparatus and method for large scale dataset | |
Korejo et al. | Multi-population methods with adaptive mutation for multi-modal optimization problems | |
Ma et al. | NSGA-II with local search for multi-objective application deployment in multi-cloud | |
CN109460862A (en) | The method that meta-heuristic algorithms based on MAB solve multi-objective optimization question | |
CN107545323A (en) | Logistics Distribution Method, device, equipment and the storage medium of commodity |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20211105 Address after: 1109, Beichen times building, No. 8, Beichen East Road, Chaoyang District, Beijing 100020 Patentee after: Beijing mingyida Technology Co.,Ltd. Address before: 100083 No. 30, Haidian District, Beijing, Xueyuan Road Patentee before: University OF SCIENCE AND TECHNOLOGY BEIJING Patentee before: BEIJING INSTITUTE OF TECHNOLOGY |
|
TR01 | Transfer of patent right |