CN106484512B - The dispatching method of computing unit - Google Patents
The dispatching method of computing unit Download PDFInfo
- Publication number
- CN106484512B CN106484512B CN201610875838.4A CN201610875838A CN106484512B CN 106484512 B CN106484512 B CN 106484512B CN 201610875838 A CN201610875838 A CN 201610875838A CN 106484512 B CN106484512 B CN 106484512B
- Authority
- CN
- China
- Prior art keywords
- computing unit
- physical machine
- individual
- resource
- algorithm
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
- G06F9/5088—Techniques for rebalancing the load in a distributed system involving task migration
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Genetics & Genomics (AREA)
- Physiology (AREA)
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to the dispatching methods of computing unit, comprising: S1. inputs physical machine resource；S2. the computing unit resource of user demand is inputted；S3. formulation description is carried out to optimization aim by data model；S4. a kind of computing unit deployment scheme is obtained using group genetic algorithm, it is overall minimum that the program makes power consumption, wasting of resources situation, SLA violate rate；S5. computing unit deployment scheme corresponding with optimal solution is exported.The dispatching method of computing unit of the present invention solves the uncertain problem of multiple optimization aim influence degrees during multiple-objection optimization by improved group genetic algorithm, has more practicability compared to single object optimization.The convergence rate that scheduling calculates obviously is accelerated, makes server cluster while significantly energy saving, also ensures running quality, significantly improve the harmony and computational efficiency of server resource.
Description
Technical field
The present invention relates to computing unit dispatching technique under cloud platform and a kind of dispatching methods of artificial intelligence, concretely
It is a kind of dispatching method of computing unit that can optimize multiple target simultaneously.
Background technique
Computing unit scheduling in distributed computing refers to one group of calculating list for applying for user according to a certain dispatching algorithm
Member is mapped on physical machine (i.e. server or physical node), and at the same time to meet necessary constraint condition.Cloud data center
The consuming of different mappings bring electric energy, resource utilization, user experience and the cloud provider income of computing unit and physical machine
Deng.Therefore, it is extremely important to design a kind of efficient scheduling algorithm.As data center's scale constantly expands, consumed electricity
Energy rapid growth, more and more researchs concentrate on the focus of computing unit scheduling problem in energy saving, existing way one
As be to be realized by way of aggregate server, computing unit is placed in the physical machine of negligible amounts by genetic algorithm,
The quantity of activation physical machine is minimized to reach energy-efficient target.This mode is strictly effective in terms of energy conservation.However, working as
Task in physical machine will cause physical machine overload when excessively assembling, the decline of user application performance is brought poor
User experience.So considering that QoS of customer cannot be had ignored while saving power consumption.Meanwhile improving physical machine
The balanced of resource utilizes, and improves data center's efficiency, reduces the effective way of the wasting of resources.
Genetic algorithm solution computing unit scheduling problem at least has following disadvantage at present: 1) algorithm performance relatively relies on
The selection of parameter, the improper superiority and inferiority that will seriously affect solution of parameter selection.2) crossover probability and mutation probability are in Evolution of Population mistake
It immobilizes in journey, seriously affect the convergence of population and easily causes precocious and globally optimal solution is not achieved.3) it is handed in algorithm
Fork operation and mutation operator randomly choose the gene on chromosome, have blindness, influence convergence speed of the algorithm.4) it adapts to
Function is spent when evaluating multi-objective optimization question, and multiple targets are linearly summed and are converted to single-objective problem, and it is multiple in practice
The fact that target is uncertain to the influence degree of problem is not consistent.
Summary of the invention
The present invention provides a kind of dispatching methods of computing unit, to optimize a variety of of the server calculated for cloud data
Operation data makes server cluster that can also guarantee running quality while energy saving, and it is harmonious to improve server resource
And computational efficiency, and practicability is had more than existing single object optimization.
The dispatching method of computing unit of the present invention, step include:
S1. it inputs physical machine resource: determining the resource category and its appearance that the quantity of physical machine and each physical machine are possessed
Amount, such as CPU, memory, bandwidth, storage etc. are all considered as the resource that physical machine possesses；
S2. the computing unit resource of user demand is inputted；
S3. formulation description is carried out to optimization aim by data model；
S4. computing unit is scheduled by group genetic algorithm: first determine gene coding mode, then by with
Machine method generates multiple groups computing unit deployment scheme, and each computing unit deployment scheme is corresponded to one in group genetic algorithm
Mapping relations between computing unit and physical machine are converted into group genetic algorithm by individual in the way of block encoding
Gene coding, all group of individuals corresponding with computing unit deployment scheme are at the initial population in group genetic algorithm；Pass through again
Calculate the fitness value of each individual in the initial population, be genetic to by the selection of the select probability of each individual it is follow-on
Individual carries out crossing operation and mutation operator to corresponding physical machine according to the crossover probability and mutation probability of each individual, it
After generate progeny population, in progeny population fitness value it is the smallest individual be group genetic algorithm optimal solution；
S5. computing unit deployment scheme corresponding with optimal solution is exported.
By the improvement to existing group genetic algorithm, keep computing unit more balanced in the scheduling of physical machine, makes entire
The whole energy consumption of physical machine cluster is minimum, and guarantees good computational efficiency.
Further, computing unit resource described in step S2 includes the number of requests and each calculating list of computing unit
The resource category and resource capacity of member.Due to being usually using network attached storage, in resource in distributed computing
Type generally only focuses on the CPU and memory of physical machine, does not consider disk, but can do corresponding adjustment according to the actual situation.
On the basis of the above, a kind of preferred method is in step s 4, by fuzzy logic theory to maximize minimum
The mode of satisfaction calculates the fitness value of each individual in the initial population.And it can also be calculated by roulette algorithm
The select probability of each individual, and individual crossover probability and mutation probability are determined with adaptive algorithm, calculate individual
In each physical machine evaluation of estimate, institute's evaluation values ascending order is arranged, the maximum physical machine of evaluation of estimate is selected to carry out crossing operation and change
Xor.
The fitness value of individual is calculated according to fitness function, and fitness function is the choosing selected the superior and eliminated the inferior in genetic algorithm
Standard is selected, to overcome the problems, such as that multiple optimization aims pass through the uncertainty of influence degree to be solved in practical application
Fuzzy logic theory carries out integration and quantitative evaluation to multiple optimization aims, using maximization min-satisfaction degree fuzzy logic model
Fuzzy processing is carried out to three optimization aims.Then use Selecting operation to a by genetic algorithm mimic biology evolution principle
Body is selected the superior and eliminated the inferior, and the probability that the high individual of fitness is genetic to filial generation is larger, and the low individual of fitness is genetic to son
The probability in generation is smaller.If but only the high individual progress heredity of selection fitness can fall into locally optimal solution, therefore the present invention uses
The individual of roulette algorithms selection heredity.
To optimize last scheduling scheme, after generating progeny population in step S4, current iteration time is first judged
Whether number reaches the maximum times of algorithm setting, stops iteration if having reached, and executes step S5；It is returned if not up to maximum number
It is back to the selection and is genetic to follow-on individual step circulation execution.
Specifically, optimization aim described in step S3 includes the minimum power consumption of physical machine, minimizes resource benefit
Rate is violated with rate and minimum SLA (service-level agreement), can also according to actual needs include other corresponding optimization mesh
Mark.
The dispatching method of computing unit of the present invention solves multiple-objection optimization process by improved group genetic algorithm
In multiple optimization aim influence degrees uncertain problem, compared to single object optimization have more practicability.Obviously plus
The convergence rate that fast scheduling calculates, makes server cluster while significantly energy saving, also ensures running quality, show
Write the harmony and computational efficiency for improving server resource.
Specific embodiment with reference to embodiments is described in further detail above content of the invention again.
But the range that this should not be interpreted as to the above-mentioned theme of the present invention is only limitted to example below.Think not departing from the above-mentioned technology of the present invention
In the case of thinking, the various replacements or change made according to ordinary skill knowledge and customary means should all be included in this hair
In bright range.
Detailed description of the invention
Fig. 1 is the flow chart of the dispatching method of computing unit of the present invention.
Fig. 2 is the flow chart of group genetic algorithm flow chart in Fig. 1.
Fig. 3 is the schematic diagram of block encoding mode in the present invention.
Specific embodiment
The dispatching method of computing unit of the present invention as shown in Figure 1, step include:
S1. physical machine resource, the physical machine quantity including cloud data center, the available resources of any one physical machine are inputted
Kind (such as CPU, memory, bandwidth, storage) class number scale is made, then Hi,jIndicate i-th physical machine PiJ class resource capacity.
S2. the computing unit resource of input user application, the quantity including computing unit, the need of any one computing unit
Resource is asked to be denoted as Ri,j, indicate i-th of computing unit ViTo the quantity required of jth class resource.Computing unit application in the present invention
CPU (processor) and memory are only focused in resource, does not consider disk, because generally being deposited using network attached in distributed computing
It stores up (network-attached storage, NAS), storage can be used as an individual module.
S3. formulation modeling.The purpose of the present embodiment is to be realized by a kind of computing unit dispatching algorithm by one group of calculating
Unit balanced and reasonable is mapped in the physical machine of cloud data center, so that reaching most while meeting physical machine resource constraint
Smallization power consumption minimizes the wasting of resources, minimizes the target that SLA (service-level agreement) violates rate.It is calculated in design scheduling
Before method, need to carry out formulation description to three kinds of optimization aims, specific as follows:
S31 power consumption: power consumption refers to the power consumption of data center's physical machine, and wherein CPU is to physical machine power consumption
Influence contribution maximum occupy most energy consumption, the present invention only considers influence of the CPU to power consumption, according to existing document it is found that
When physical machine changes from zero load (0%) to full load (100%) this section, physical machine power consumption and CPU are using linearly
Relationship, and the 67% of electric energy is consumed when physical machine consumed electric energy in zero load is full load, the power consumption P of physical machineiIt can
It calculates are as follows:Wherein, PmaxIndicate power consumption when physical machine full load, PidleIndicate physical machine
Power consumption when idle,For the cpu busy percentage of i-th physical machine.
S32. resource utilization: various resources (such as the money such as CPU, memory, memory space of each physical machine of data center
Source) utilization rate should keep balancing as far as possible, avoid resulting in waste of resources because of wooden pail effect, reduce resource utilization, physical machine
Resource waste rate RiIt indicates are as follows:
Wherein,For the ratio of i-th physical machine remaining cpu resource and total cpu resource, i.e. CPU surplus ratio,Indicating the memory surplus ratio of the physical machine, the value range of ε is 0~0.001,For the cpu busy percentage of the physical machine,For the memory usage of the physical machine.
S33.SLA violates rate: in distributed computing, cloud provider is needed and user signs SLA, wherein service quality is defined,
The service quality of the items such as user-pay, computing unit distribution is generally considered with user's application performance, and cloud provider must be
User's distribution just can guarantee its performance with the comparable resource of demand.SLA violates the function that rate is defined as cpu busy percentage, and SLA is violated
The evaluation function f of rateiIt indicates are as follows:Wherein,For the cpu busy percentage of i-th physical machine,For the max-thresholds of cpu busy percentage.
The optimization aim of computing unit scheduling indicates are as follows: minimizes power consumptionMinimize resource wave
TakeIt minimizes SLA and violates rate
S4. as shown in Fig. 2, being asked using the computing unit scheduling that improved group genetic algorithm solves the multiple-objection optimization
Topic.The following steps are included:
S41. gene encodes.Mapping feature in conjunction with the computing unit in computing unit scheduling problem and between physical machine is adopted
With block encoding mode.The gene that physical machine is regarded as to chromosome in genetic algorithm sees the computing unit for being deployed in physical machine
Work is the value of gene.Such as block encoding schematic diagram shown in Fig. 3, it is assumed that 7 computing units have been deployed to 3 physics
On machine, a physical machine represents the coding of a chromosome, and a computing unit number represents the genic value on homologue.
When chromosome coding is arranged, in addition to the constraint condition that must satisfy the problem of being solved, it is necessary to meet any one gene
It can only appear on a chromosome, all genes must all map on chromosome, but the dye of not gene can occur
Colour solid.Such as the coding of chromosome is ABC in Fig. 3, according to mentioned above principle setting chromosome coding mode be A:123, B:45,
C:67.Thus can solve in traditional code mode can that is, when intersection and variation to the indefinite problem of grouping information
It is transparent to server info to be operated to computing unit.And traditional binary coding mode be will each individual with one
Go here and there regular length binary number representation, intersected, mutation operation when can only be by individual as unit of operated, cannot be anti-
Reflect the structure feature and specific information of required problem.
S42. initial population is generated.The computing unit deployment scheme that quantification is generated using random algorithm, i.e., by user
On the computing unit Random Maps to physical machine of request, precondition is that the resource capacity of physical machine has to be larger than computing unit and asks
The resource capacity asked.The mapping relations between computing unit and physical machine are converted into being grouped according to the block encoding mode of S41
Gene coding in genetic algorithm, each computing unit deployment scheme correspond to the individual in algorithm, all computing unit portions
The corresponding all individuals of management side case constitute the initial population of group genetic algorithm；
S43. fitness value is calculated.The fitness value of individual is calculated according to fitness function, and fitness function is hereditary calculation
The selection criteria of the method survival of the fittest, for overcome the problems, such as multiple optimization aims in practical problem to influence degree uncertainty,
Integration and quantitative evaluation are carried out to multiple optimization aims by fuzzy logic theory, using maximization min-satisfaction degree fuzzy logic
Model carries out Fuzzy processing to three optimization aims.Specific steps are as follows:
T431. the optimal solution of each single goal is solved, it is first in order to which whether clear last solution shows well in each optimization aim
First need to solve the optimal solution of each single goal.
Firstly, the minimum value of physical machine quantity is denoted as Lmin。
Wherein, wherein Rall-cpu, Rall-memRespectively indicate all computing unit applications of data center cpu resource sum total and
Memory source summation, Hcpu, HmemRespectively indicate the cpu resource and memory source that separate unit physical machine is possessed.Physical machine quantity
Maximum value is denoted as Lmax=N, wherein N indicates the number of computing unit.
The minimum value of power consumption indicates are as follows: W*=Lmin·Pmax；The maximum value of power consumption indicates are as follows: Wmax=Lmax·
Pmax, PmaxIndicate power consumption when physical machine full load,.
The minimum value of the wasting of resources indicates are as follows:Wherein Hcpu-all, Hmem-allIt respectively indicates
All physical machine cpu resource summations of data center and memory source summation.The maximum value of the wasting of resources indicates are as follows:Wherein Ri,cpu, Ri,memRespectively indicate the cpu resource and memory money of i-th of computing unit application
Source, Hi,cpu, Hi,memRespectively indicate the cpu resource and memory source of i-th physical machine.
SLA violates the upper limit of rate and lower limit violates the evaluation function f of rate according to SLAiIt determines.
T432. membership function is constructed.Membership function reacts the superiority and inferiority degree of optimization aim, and the smaller expression solution of degree of membership is more not
The fact that energy receives, and cannot receive completely for 0 expression, and being 1 is exactly ideal value, general can not occur.
The membership function of three optimization aims indicates are as follows:
Wherein, μ (fi(x)) angle value that is subordinate to of i-th of optimization aim is indicated, x indicates computing unit deployment scheme, W*+ δ1It is
The maximum value of power consumption, similarly, R*+ δ2, F*+ δ3Respectively indicate the maximum value of the wasting of resources and maximum value that SLA is violated, δi
Indicate the difference of ideal value (minimum value) and maximum value, F* is the minimum value that SLA is violated.
T433. fitness function is constructed.Fitness function determines the direction of Evolution of Population, therefore the letter in genetic algorithm
Whether number is suitable for directly determining whether the last solution of algorithm is optimal solution.Adaptation is determined using min-satisfaction degree method is maximized
Function is spent, this method makes all objective functions have degree of membership as high as possible, and fitness function indicates are as follows: μ (x)=min { μ1
(x),μ2(x),μ3(x) }, μiIt (x) is the membership function value of three optimization aims.Then multiple-objection optimization computing unit scheduling turns
Turning to makes the maximized single-object problem of μ (x):
Wherein s.t. indicates constraint condition.
S44. Selecting operation.Genetic algorithm mimic biology evolution principle selects the superior and eliminates the inferior to individual using Selecting operation,
The probability that the high individual of fitness is genetic to filial generation is larger, and the probability that the low individual of fitness is genetic to filial generation is smaller.But
If only the high individual of selection fitness, which carries out heredity, can fall into locally optimal solution, so here using the heredity of roulette algorithms selection
Individual.
Firstly, calculating the fitness value μ (x of all individuals in population according to step S2i), 1≤i≤m, m are population scale,
The fitness summation of population is calculated, then individual xiSelect probability be p (xi), illustrate the choosing of i-th kind of computing unit deployment scheme
Select probability, p (xi) are as follows:
The accumulation probability of individual is calculated,
The random number r for being generated (0,1) at random using roulette algorithm is matched each to determine with the genetic probability of individual
Whether individual is inherited by filial generation.If qi-1< r≤qi, then i-th of individual will be genetic to the next generation.This Selecting operation can
Avoid the defect that locally optimal solution is fallen into caused by carrying out heredity because of the high individual of only selection fitness.
S45. crossing operation, it is new that crossing operation is that two chromosomes are formed with certain crossover probability switching part gene
Chromosome is the key step that genetic algorithm generates new individual, it is therefore an objective to it is desirable that outstanding gene can be genetic to filial generation
In, crossing operation step are as follows:
T451. adaptive polo placement crossover probability.Crossover probability has the performance and efficiency of entire algorithm particularly important
It influences, also most important to convergence, the new individual of the bigger generation of crossover probability will be faster, while to hereditary pattern
Destruction a possibility that also can be higher；On the other hand, crossover probability is smaller, entire search process can be allowed to become slowly, or even stop
It is stagnant not before.Determine that crossover probability, crossover probability change automatically with fitness using adaptive method, when each ideal adaptation of population
When degree reaches unanimity or tends to local optimum, increases crossover probability, when group's fitness is more dispersed, the solution is enable to protect
Shield enters the next generation, and the adaptive method for determining crossover probability guarantees the receipts of genetic algorithm while keeping population diversity
Holding back property.
Crossover probability PcIt indicates are as follows:
Wherein, fmaxIndicate maximum fitness value in population, favgIndicate the average fitness value of all individuals in population,
fcBiggish fitness value in two individuals to be intersected, k1, k2The value between (0,1) is taken, is adjusted in calculating process.
According to the population at individual that step S44 is generated, the individual in population is matched two-by-two randomly, successively selects one pair of them
Body generates random number r, if Pc> r then carries out crossing operation to the individual chosen, is otherwise directly entered S46.
T452. crosspoint is determined according to physical machine evaluation of estimate.After determining crossover probability in crossing operation, need to select to hand over
Crunode.With in gene physical machine wasting of resources degree and SLA violate rate as according to judging which section gene will be intersected.It is non-selected
It is close that the power consumption of physical machine as the reason of judgment basis is that SLA violates rate and power consumption and has with cpu busy percentage
Relationship and variation tendency are close, and the size of separate unit physical machine power consumption cannot reflect the excellent of deployment scheme in the physical machine
Bad degree.The wasting of resources degree of physical machine is calculated according to S32 and S33 step and SLA violates rate, and the weighted sum of the two is
The evaluation of estimate of physical machine in individual, the physical node small to evaluation of estimate ascending sort selective value carry out crossover operation.
T453 implements crossing operation.Assuming that two parent chromosome X, Y carries out crossing operation, by the intersection in chromosome x
In point insertion chromosome Y, if the case where chromosome Y duplicates computing unit in multiple physical machines at this time, it will weigh
The case where physical machine of multiple computing unit is deleted, and delete operation may cause certain computing units unassigned physical machine,
It then needs to be re-encoded into physical machine for unappropriated computing unit.Similarly, chromosome is inserted into the crosspoint in chromosome Y
In X.Two new child chromosomes will be generated after crossing operation.
S46. mutation operator, mutation operator is to randomly choose one or more bases in chromosome in basic genetic algorithmic
Because making a variation, to avoid locally optimal solution is fallen into, population diversity is kept.Performance of the selection of mutation probability to genetic algorithm
Most important with convergence, mutation probability is too small, and the individual configurations being just not likely to produce, mutation probability is excessive, genetic algorithm just at
The step of pure random search algorithm, mutation operator are as follows:
T461. adaptive polo placement mutation probability.When ideal adaptation angle value is greater than average individual fitness value, select smaller
Mutation probability, retain outstanding gene as far as possible, conversely, selecting biggish mutation probability.Adaptive polo placement mutation probability PmIt is as follows:
Wherein, fmaxIndicate maximum fitness value in population, favgIndicate the average fitness value of all individuals in population,
fmThe fitness value of variation individual, k are wanted in expression3, k4Take the value between (0,1).It is adjusted in calculating process.According to step
The population at individual that S45 is obtained, successively the individual in selected population, calculates its variation according to the calculation formula of self-adaptive mutation
Probability generates random number r, judges whether mutation probability Pm> r, if then carrying out mutation operation to the individual chosen, otherwise directly
Tap into row S47.
T462. according to physical machine evaluation of estimate definitive variation point, the physical machine evaluation of estimate being calculated according to T452, selection is commented
The physical machine of Maximum Value is as change point.
T463. implement mutation operator.Variation mode is random variation, i.e., it is single to change original calculating in determining physical machine
One or more changes in member, are likely to occur duplicate computing unit in entire individual after variation, at this time will be duplicate
Computing unit is deleted, if occurring the computing unit that do not dispose at this time, it is re-encoded into physics according to optimal adaptation algorithm
On machine.
S47. offspring individual is generated, by generating new progeny population after above-mentioned steps, if the number of iterations is at this time
Reached maximum times, then in current population the smallest individual of fitness value be algorithm optimal solution, algorithm terminates；If not reaching
To maximum number of iterations, then skips to step S44 circulation and execute.
S5. the chromosome coding of optimal solution individual is converted into corresponding computing unit deployment scheme, transform mode is with Fig. 3
For computing unit deployment scheme, which corresponds to the matrix of 7 rows 3 column in the algorithm, as follows:
Indicate that computing unit i is deployed on physical machine j for 1 position, the matrix after output conversion, i.e. output computing unit
Deployment scheme.
Claims (6)
1. the dispatching method of computing unit, feature include:
S1. it inputs physical machine resource: determining resource category and its capacity that the quantity of physical machine and each physical machine are possessed；
S2. the computing unit resource of user demand is inputted；
S3. formulation description is carried out to optimization aim by data model；
S4. computing unit is scheduled by group genetic algorithm: determines gene coding mode first, then pass through random side
Method generates multiple groups computing unit deployment scheme, and each computing unit deployment scheme is corresponded in group genetic algorithm one by one
Body, the gene being converted into the mapping relations between computing unit and physical machine in the way of block encoding in group genetic algorithm
Coding, all group of individuals corresponding with computing unit deployment scheme are at the initial population in group genetic algorithm；Pass through calculating again
The fitness value of each individual in the initial population is genetic to follow-on by the select probability selection of each individual
Body carries out crossing operation and mutation operator to corresponding physical machine according to the crossover probability and mutation probability of each individual, later
Progeny population is generated, if current the number of iterations has reached the maximum times of algorithm setting, is adapted in current progeny population
The smallest individual of angle value is the optimal solution of group genetic algorithm, enters step S5；If current the number of iterations is not up to algorithm and sets
The maximum times set then are back to the step circulation for selecting to be genetic to follow-on individual and execute, until being grouped
The optimal solution of genetic algorithm, subsequently into step S5；
S5. computing unit deployment scheme corresponding with optimal solution is exported.
2. the dispatching method of computing unit as described in claim 1, it is characterized in that: computing unit resource packet described in step S2
Include the number of requests of computing unit and the resource category and resource capacity of each computing unit.
3. the dispatching method of computing unit as described in claim 1, it is characterized in that: in step S4, by fuzzy logic theory with
The mode for maximizing min-satisfaction degree calculates the fitness value of each individual in the initial population.
4. the dispatching method of computing unit as described in claim 1, it is characterized in that: in step S4, calculated by roulette algorithm
The select probability of each individual.
5. the dispatching method of computing unit as described in claim 1, it is characterized in that: in step S4, determined with adaptive algorithm a
The crossover probability and mutation probability of body calculate each physical machine evaluation of estimate in individual, arrange institute's evaluation values ascending order, selection evaluation
It is worth maximum physical machine and carries out crossing operation and mutation operator.
6. the dispatching method of computing unit as described in one of claim 1 to 5, it is characterized in that: optimization mesh described in step S3
Mark includes the minimum power consumption of physical machine, minimizes resource utilization and minimize SLA and violate rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610875838.4A CN106484512B (en) | 2016-10-08 | 2016-10-08 | The dispatching method of computing unit |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610875838.4A CN106484512B (en) | 2016-10-08 | 2016-10-08 | The dispatching method of computing unit |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106484512A CN106484512A (en) | 2017-03-08 |
| CN106484512B true CN106484512B (en) | 2019-07-09 |
Family
ID=58269148
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610875838.4A Active CN106484512B (en) | 2016-10-08 | 2016-10-08 | The dispatching method of computing unit |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106484512B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108279968B (en) * | 2017-12-29 | 2021-05-11 | 中国联合网络通信集团有限公司 | Virtual machine resource scheduling method and device |
| CN108287666B (en) * | 2018-01-16 | 2021-01-26 | 中国人民公安大学 | Data storage method and device for cloud storage environment |
| CN108521446B (en) * | 2018-03-20 | 2021-02-23 | 深圳大学 | Scheduling method, device and equipment of cloud computing resources and storage medium |
| CN109857526A (en) * | 2018-12-27 | 2019-06-07 | 曙光信息产业(北京)有限公司 | A kind of scheduling system towards mixing computation frame |
| CN110058924B (en) * | 2019-04-23 | 2023-08-04 | 东华大学 | Multi-objective optimized container scheduling method |
| CN110308993B (en) * | 2019-06-27 | 2022-12-13 | 大连理工大学 | Cloud computing resource allocation method based on improved genetic algorithm |
| CN113132445B (en) * | 2020-01-10 | 2023-04-28 | 阿里巴巴集团控股有限公司 | Resource scheduling method, equipment, network system and storage medium |
| CN111274030B (en) * | 2020-01-16 | 2022-09-23 | 中国人民解放军国防科技大学 | Efficient multiprocessor system-on-chip design space mining method oriented to application features |
| CN113300982B (en) * | 2020-06-08 | 2022-08-23 | 阿里巴巴集团控股有限公司 | Resource allocation method, device, system and storage medium |
| CN116502473B (en) * | 2023-06-27 | 2024-01-12 | 中科航迈数控软件(深圳)有限公司 | Wire harness electromagnetic compatibility optimization method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102055196A (en) * | 2010-12-20 | 2011-05-11 | 南京软核科技有限公司 | 10 kv-distribution network reactive power compensation optimization method in power system |
| CN103412792A (en) * | 2013-07-18 | 2013-11-27 | 成都国科海博计算机系统有限公司 | Dynamic task scheduling method and device under cloud computing platform environment |
| CN103744714A (en) * | 2011-12-31 | 2014-04-23 | 华茂云天科技(北京)有限公司 | Virtual machine management platform based on cloud computing |
| CN104657215A (en) * | 2013-11-19 | 2015-05-27 | 南京鼎盟科技有限公司 | Virtualization energy-saving system in Cloud computing |
| CN105302632A (en) * | 2015-11-19 | 2016-02-03 | 国家电网公司 | Cloud computing working load dynamic integration method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9429921B2 (en) * | 2011-09-16 | 2016-08-30 | Siemens Aktiengesellschaft | Method and system for energy control management |
-
2016
- 2016-10-08 CN CN201610875838.4A patent/CN106484512B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102055196A (en) * | 2010-12-20 | 2011-05-11 | 南京软核科技有限公司 | 10 kv-distribution network reactive power compensation optimization method in power system |
| CN103744714A (en) * | 2011-12-31 | 2014-04-23 | 华茂云天科技(北京)有限公司 | Virtual machine management platform based on cloud computing |
| CN103412792A (en) * | 2013-07-18 | 2013-11-27 | 成都国科海博计算机系统有限公司 | Dynamic task scheduling method and device under cloud computing platform environment |
| CN104657215A (en) * | 2013-11-19 | 2015-05-27 | 南京鼎盟科技有限公司 | Virtualization energy-saving system in Cloud computing |
| CN105302632A (en) * | 2015-11-19 | 2016-02-03 | 国家电网公司 | Cloud computing working load dynamic integration method |
Non-Patent Citations (2)
| Title |
|---|
| 云环境下调度问题的研究与实现;秦烁;《中国优秀硕士学位论文全文数据库 信息科技辑》;20160215(第02期);I139-10 * |
| 基于遗传算法的分组算法;李峰 等;《长春工业大学学报（自然科学版）》;20130430;第34卷(第2期);214-217 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106484512A (en) | 2017-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106484512B (en) | The dispatching method of computing unit | |
| Reddy et al. | Multi‐objective particle swarm optimization for generating optimal trade‐offs in reservoir operation | |
| CN106845642B (en) | A kind of adaptive multi-target evolution method of belt restraining cloud workflow schedule | |
| CN107070704A (en) | A kind of Trusted Web services combined optimization method based on QoS | |
| CN105323166B (en) | A kind of method for routing preferential based on network energy efficiency towards cloud computing | |
| Phan et al. | Evolutionary multiobjective optimization for green clouds | |
| CN102063339A (en) | Resource load balancing method and equipment based on cloud computing system | |
| CN109447264B (en) | Virtual machine placement genetic optimization method based on VHAM-R model in cloud computing environment | |
| CN110033076B (en) | Workflow data layout method for cost optimization in mixed cloud environment | |
| Zhang et al. | MCTE: Minimizes task completion time and execution cost to optimize scheduling performance for smart grid cloud | |
| CN117539726B (en) | Energy efficiency optimization method and system for green intelligent computing center | |
| CN109582985A (en) | A kind of NoC mapping method of improved genetic Annealing | |
| CN109919485A (en) | A kind of Multipurpose Optimal Method of the Water Resources Allocation based on NSGA-II | |
| Su et al. | The optimizing resource allocation and task scheduling based on cloud computing and Ant Colony Optimization Algorithm | |
| CN111754091A (en) | Power consumer demand side regulation and control system | |
| CN110163546A (en) | A method of optimal replenishment quantity is determined based on big data | |
| Ouyang et al. | Amended harmony search algorithm with perturbation strategy for large-scale system reliability problems | |
| Zhang | Reliable virtual network mapping algorithm with network characteristics and associations | |
| CN107180286A (en) | Manufacturing service supply chain optimization method and system based on modified pollen algorithm | |
| CN115277445B (en) | Micro-service system scheduling optimization method based on QoS (quality of service) target | |
| Yu | Research on optimization strategy of task scheduling software based on genetic algorithm in cloud computing environment | |
| CN113285832B (en) | NSGA-II-based power multi-mode network resource optimization allocation method | |
| Tang | Load balancing optimization in cloud computing based on task scheduling | |
| CN107451943A (en) | The site selecting method of Urban renewal | |
| Sun et al. | Hybrid evolutionary algorithm for integrated supply chain network design with assembly line balancing |
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 |