CN109597682A - A kind of cloud computing workflow schedule method using heuristic coding strategy - Google Patents
A kind of cloud computing workflow schedule method using heuristic coding strategy Download PDFInfo
- Publication number
- CN109597682A CN109597682A CN201811416065.9A CN201811416065A CN109597682A CN 109597682 A CN109597682 A CN 109597682A CN 201811416065 A CN201811416065 A CN 201811416065A CN 109597682 A CN109597682 A CN 109597682A
- Authority
- CN
- China
- Prior art keywords
- individual
- heuristic
- population
- workflow schedule
- task
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of cloud computing workflow schedule methods using heuristic coding strategy, comprising: (1) the Population Size size set generates size individual；Corresponding workflow schedule scheme is generated using heuristic coding strategy, then calculates the adaptive value of corresponding scheduling scheme；(2) crossover operator is executed to needing to be intersected individual；(3) need the individual execution mutation operator that made a variation；(4) fitness function evaluation is carried out, with the worst individual of history optimum individual replacement new population；(5) it selects size optimal individual to enter in merging population to recycle next time；(6) the corresponding scheduling scheme of history optimum individual is exported if reaching termination condition；Otherwise step (2) are returned to.The present invention uses the genetic algorithm of novel heuristic coding strategy, by the adaptation value calculating method of search and the matching and formulation of seven kinds of heuristic informations, so that algorithm is efficiently searched and meet the cloud computing workflow schedule scheme that user constrains substantially and optimizes requirement.
Description
Technical field
The present invention relates to workflow schedule field more particularly to a kind of cloud computing workflows using heuristic coding strategy
Dispatching method.
Background technique
By making full use of the magnanimity computing resource of explosive growth, cloud computing mode has obtained hair at full speed in recent years
Exhibition.It by the way that large-scale calculating task is split as multiple subtasks, and is distributed in different service processes, so that cloud meter
Calculation can satisfy the large-scale calculations demand of high quality-of-service (QoS), and in economics, astronomy, the fields such as Neuscience are obtained
To being widely applied.And in the distributed system being made of a large amount of computing resources, the performance of each service, cost and reliable
Property is different, while user has different requirements to the duration of calculating task, expense and reliability again.How scientifically to dispatch
Subtask matches with computing resource, so that general assignment meets the demand for services of user, it is the critical issue of cloud computing service.
In order to solve the problems, such as distributed resource scheduling, general calculating task can be abstracted as Work flow model.Pass through
Single subtask is defined as node, the priority list between subtask is shown as oriented line, available to use directed acyclic
Scheme the Work flow model that (DAG) is indicated.In this model, only when the father node task of a upper priority is fully completed,
The task for the child node that can bring into operation.Computing resource is distributed according to dominance relation for each node, keeps the interests of user maximum
Change, becomes workflow schedule the very corn of a subject.In past research, the work of a series of performance based on service processes
Stream scheduling system is developed, and is referred to as BRS including the algorithm of Pegasus system, Kepler system etc., these systems
Algorithm.BRS algorithm is based primarily upon the highest computing capability reality on the earliest finish time of task, earliest start time and service
It now dispatches, high performance scheduling scheme can be efficiently found.However as universal and scale the expansion of cloud computing, economy
Become the factor that can not ignore with indexs such as reliabilities, this series of BRS algorithm can not be realized effectively to multiple indexs
Constraint and optimization.
For current cloud computing workflow schedule problem, main bugbear includes following two points: first, service quality is not
It is dependent only on the total construction period of service, further includes the reliability of the expense and service processes using computing resource.Scheduling needs
While meeting the minimum standard of above three index, need to provide peak performance/lowest overhead/highest further according to user
The scheduling scheme of reliability.Second, workflow schedule problem is np complete problem, and region of search is huge in large-scale scheduling problem
Greatly, it is difficult to find optimal solution.Heuritic approach provides effective solution method for the NP problem of search completely of this belt restraining.
In addition there are a series of dispatching methods based on genetic algorithm (GA), ant group algorithm (ACO) and simulated annealing (SA), adopt
Feasible solution is searched for similar coding mode.However the above algorithm is difficult in searching process since coding mode is limited
The feasible solution for being met multiple constraints simultaneously also lacks search efficiency to the optimization of single index.
Summary of the invention
It is an object of the invention to solve the problems, such as that the optimization efficiency problem of workflow schedule, the present invention provide a kind of use and open
The cloud computing workflow schedule method of hairdo coding strategy.The present invention is to the scheduling of workflow using a kind of novel coding plan
Slightly, the method for the set of traditional duty mapping to service is improved to the mapping method based on heuristic information, and passes through solution
Certainly the genetic algorithm of discrete optimization problems of device optimizes scheduling problem.Since these heuristic informations can meet difference respectively
Index request, therefore solve search the limitation of different indexs can be more easily satisfied, so that the search efficiency of algorithm is improved, in reality
Meet the dispatching requirement of cloud computing system in the application of border.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of cloud computing workflow schedule method using heuristic coding strategy, specific steps include:
(1) preset Population Size size is generated by the method for each gene random value to each individual
Size individual.Corresponding workflow schedule scheme is generated according to using heuristic coding strategy for each individual, further according to
Evaluation function calculates the adaptive value of scheduling scheme corresponding to each individual.
(2) for each individual of existing population, a random number q is generated1∈ [0,1], and by its with preset
Crossover probability CR compare, as the q of generation1When less than CR, crossover operator is executed to the individual, is obtained using the method for single point crossing
To new individual.
(3) to each individual of the new population after intersection, a random number q is generated2∈ [0,1], and by its with set in advance
Fixed mutation probability MR compares, as the q of generation2When less than MR, mutation operator is executed to the individual, the method to make a variation using single-point
It makes a variation to the individual.
(4) fitness function evaluation, and the individual replacement optimal with history are carried out to each of new population after variation individual
The worst individual of new population.
(5) new population and previous generation population are ranked up according to fitness function size, are selected in merging population optimal
Size individual enter recycle next time.
(6) terminate program, the corresponding scheduling scheme of output history optimum individual if reaching termination condition.Otherwise step is returned to
Suddenly (2).
Further, for each individual in population, gene coding are as follows:
(T1,T2,…,Tn)
Wherein, n indicates the total task number of workflow.Each gene representation of individual subtask to be matched, according to work
The priority ranking of task in stream.For each task Ti, region of search { RG, TG, CG, SC, ST, TC, GP } is expressed as seven kinds
The heuristic information of service-oriented process.(T is encoded for the gene of individual1,T2,…,Tn), according to from T1To TnSequence, it is right
The heuristic information value of each gene carries out the calculating of corresponding heuristic information, and to inspiring in respective service processes list
The highest process of the formula value of information is matched, and a workflow schedule solution S={ s is obtained1,s2,…,sn}。
Further, the adaptive value of all individuals of population, the calculation method of individual fitness are calculated are as follows:
Scheduling is solved, according to service processes siTri- parameter s of QoS of (1≤i≤n)i.t, si.c, si.r, tune is calculated
The qos parameter of degree solution S, is total construction period S.t, overhead S.c and overall reliability S.r respectively.
For three optimization problems of three different parameters optimization, the calculation of adaptive value are as follows:
1. total construction period optimization problem.For total construction period and overhead optimization problem, by all service processes not
The process for meeting reliability minimum requirements excludes, and task schedule is carried out in remaining process, so that the constant satisfaction of scheduling problem
Least reliability constraint in QoS constraint.Only need to consider the constraint of overall overhead at this time, therefore total construction period optimization problem is suitable
Answer the calculation of function are as follows:
Wherein, min_Cost and max_Cost respectively indicates the minimal-overhead and maximum cost of workflow schedule problem,
It can be by the way that all task schedules be obtained into the min/max service processes of expense respectively.LimitDuration is indicated
Restricted duration.The adaptive value calculated according to value function is adapted to, can satisfy as S.t≤LimitDuration,
The value of S.fitness_D will be greater than value as S.t > LimitDuration certainly, so as to being unsatisfactory for constraining
Individual assign punishment adaptive value, reach algorithm tend to the duration optimize while tend to meet constraint requirement.
2. overhead optimization problem.It is similar to total construction period optimization problem, by exclude to be worth servicing lower than lowest reliable into
Journey simplifies problem.The then calculation of the fitness function of overhead optimization problem are as follows:
Wherein, min_Time and max_Time respectively indicate workflow schedule problem the shortest duration and longest work
Phase again may be by respectively obtaining all task schedules into duration most short/longest service processes.LimitCost
Indicate limitation expense.
3. overall reliability optimization problem.To the optimization problem of overall reliability, need considering duration and expense about
While beam, the workflow schedule solution for possessing maximum reliability is found.In view of the punishment to two constraints are not met, totally may be used
By the fitness function of property optimization problem is defined as:
Wherein, min_Reliability and max_Reliability respectively indicates in all service processes reliability most
Low/corresponding reliability index of highest service processes.The fitness function of design adapts to when two constraints all meet simultaneously
Functional value is possible to be greater than 1, and becomes larger with the raising of overall reliability.Otherwise, the size of fitness function depends on
Meet the punishment of how many constraints and scheduling solution beyond constraint.
Further, the step (2) specifically: as the random number q of generation1It is wait hand over by individual mark when less than CR
The individual of fork.Crossover operator is executed to needing to be intersected individual, is matched two-by-two using the method for random fit.To every a pair
The individual of pairing generates a random integers q '1∈ [1, n] is used as crosspoint.The individual of pairing will be from q '1What is started is subsequent
All gene swapping, and obtain two two original individuals of new individual substitution.
Further, the step (3) specifically: to the individual of each variation, generate a random integers q '2∈
[1, n] it is used as catastrophe point, and for positioned at the gene T of catastrophe pointq′2The assignment, and determining T again by way of random valueq′2's
Value and original difference, successively obtain new variation individual and substitute original individual.
The present invention compared to the prior art, have it is below the utility model has the advantages that
1, the present invention uses genetic algorithm, optimizes to workflow schedule problem for different indexs.The calculation of invention
Method uses novel coding strategy, proposes seven kinds of heuristic informations and encodes to the mapping between task and service, big
Region of search can be effectively reduced in the workflow schedule problem of scale, while improving the probability for searching feasible solution, in certain journey
The efficiency of optimization is improved on degree.
2, it can be searched to higher efficiency the present invention is based on the coding mode of heuristic information and meet what all constraints were adjusted
Solution, and then under the premise of can satisfy three QoS index minimum requirements, find one of optimal mode.
3, the present invention can effectively reduce search using the coding mode of heuristic information in extensive scheduling problem
Domain, and then can efficiently realize real-time cloud computing resources scheduling.
Detailed description of the invention
Fig. 1 is the specific flow chart of the embodiment of the present invention.
Fig. 2 is the workflow diagrams in the embodiment of the present invention.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
Embodiment
In the present embodiment, workflow task model is as shown in Figure 1, all node T={ T therein1,T2,…,TnTable
Show task to be scheduled in workflow, and the oriented line between node represents the priority between task, i.e. high priority
Father node is directed toward the child node of next priority.In the scheduling problem of workflow, the execution of task must be in strict accordance with preferential
Grade, all tasks and if only if high priority are all completed, and the task of next priority can be executed.When scheduling process
When, the task of different priorities can be assigned to the same process.Conversely, the task of same priority cannot be assigned to
The same process.For each task T in task listi, have corresponding current schedulable service processes list, be expressed asWherein miIndicate TiSchedulable task quantity,It indicates corresponding to appoint
Be engaged in TiIn current schedulable service processes number be j process.
The list be in workflow schedule problem it is transformable, when with task TiOther tasks with equal priority
It is scheduled for PiIn some process when, the process is in PiIn will also be removed.The essence of workflow schedule is exactly for work
Each of stream node (task) Ti, a service is selected in schedulable process listA workflow is obtained with this
Scheduling solution S={ s1,s2,…,sn}。
For process P all in cloud computing system1∪P2∪…∪Pn, wherein each processAll
There are three correspond to task T for tooliQoS attribute, be respectively:
1. processSolution task TiTime used
2. processSolution task TiRequired expense
3. processSolution task TiThe reliability showed
Based on this, the QoS index of a workflow schedule solution S can be expressed as following three aspects:
(1) total construction period.The total construction period of cloud computing workflow execution is the addition of the workflow execution time of different priorities.
Wherein the execution time of some priority is then shown functionally in the maximum execution time of the task of the priority.Specific statement are as follows:
The wherein priority quantity of n ' expression workflow, njIndicate the quantity for being located at the task of jth priority,It indicates
Service processes, wherein i indicates corresponding task Ti, siIndicate the task T of schedulingiCorrespondence workflow schedule solution S={ s1,s2,…,
sn}。
(2) total cost.The overhead of workflow schedule is equal to the total value of the expense for all processes called, statement are as follows:
(3) reliability.Each ring in the reliability index requirements workflow of cloud computing will meet minimum reliability
It is required that.Under the premise of this, the reliability for dispatching S can be stated are as follows:
In workflow schedule problem, S is solved for a scheduling, three of them QoS index all needs to meet certain user
It is required that being to be no more than the duration to limit the time limit respectively, cost is no more than restriction amount and reliability is wanted not less than lowest reliable
It asks.These constraints show themselves in that
In practical applications, user may be required according to different requirements, to one of progress in three QoS indexes
Optimization, for example require cloud computing to reach highest reliability, minimum cost either highest operation efficiency, while there is still a need for full
The minimum standard constraint of these three indexs of foot.
Therefore, a kind of cloud computing workflow schedule method based on heuristic coding strategy is provided in the present embodiment, specifically
Flow chart is as shown in Fig. 2, specific steps include:
(1) preset Population Size size is generated by the method for each gene random value to each individual
Size individual.Corresponding workflow schedule scheme is generated according to using heuristic coding strategy for each individual, further according to
The evaluation function of algorithm calculates the adaptive value of scheduling scheme corresponding to each individual.
(2) for each individual of existing population, a random number q is generated1∈ [0,1], and by its with preset
Crossover probability CR compare, as the q of generation1When less than CR, crossover operator is executed to the individual, is obtained using the method for single point crossing
To new individual.
(3) to each individual of the new population after intersection, a random number q is generated2∈ [0,1], and by its with set in advance
Fixed mutation probability MR compares, as the q of generation2When less than MR, mutation operator is executed to the individual, the method to make a variation using single-point
It makes a variation to the individual.
(4) fitness function evaluation, and the individual replacement new population optimal with history are carried out for each of new population individual
Worst individual.
(5) new population and previous generation population are ranked up according to fitness function size, are selected in merging population optimal
Size individual enter recycle next time.
(6) terminate program, the corresponding scheduling scheme of output history optimum individual if reaching termination condition.Otherwise step is returned to
Suddenly (2).
Further, for each individual in population, gene coding are as follows:
(T1,T2,…,Tn)
Wherein, n indicates the total task number of workflow.Each gene representation of individual subtask to be matched, according to work
The priority ranking of task in stream.For each task Ti, region of search { RG, TG, CG, SC, ST, TC, GP } is expressed as seven kinds
The heuristic information of service-oriented process.(T is encoded for the gene of individual1,T2,…,Tn), according to from T1To TnSequence, it is right
The heuristic information value of each gene carries out the calculating of corresponding heuristic information, and to inspiring in respective service processes list
The highest process of the formula value of information is matched, and a workflow schedule solution S={ s is obtained1,s2,…,sn}。
Further, in heuristic coding strategy, RG, TG, CG, SD, SB, TC, OP respectively correspond seven kinds of different works
The heuristic information of service processes for each task.The heuristic information is all based on process
These three QoS attributes respectively evaluate duration, expense and reliability and its immixture come what is calculated, can help
Algorithm, which more efficiently obtains, meets the constrained scheduling solution of institute.This seven kinds of heuristic informations are defined as follows:
Reliability greed (RG) indicates to select the highest service processes of reliability in process list, that is, has
Wherein,Indicate service processesReliability greed value.
Duration greed (TG) indicates to select duration shortest service processes in process list, that is, has
Expense greed (CG) indicates to select the smallest service processes of expense in process list, that is, has
It is recommended that Overhead (SC) indicates that selection is just able to satisfy the service processes of total construction period limitation.Based on operation cost compared with
Small process, operation efficiency and reliability often lower premise are advantageously selected for being just met for the clothes of overhead limitation
Business process can theoretically make scheduling while meeting overhead constraint, obtain relatively higher efficiency and higher reliable
Property, it can be used for reliability optimization problem and Time Optimization problem.Based on the above premise, for i-th task, it is recommended that the phase
Limit SCiIt indicates to be conducive to be just met for the time limit that total cost limits, calculation is as follows:
By all tasks in addition to i-th task in the smallest process matching of respective cost, obtained overhead
It is subtracted by LimitCost, then obtains task TiTheoretical maximum cost budget max_costi:
Meanwhile task TiTheoretical minimal-overhead budget for cost minimum in its all process the corresponding cost of resource,
I.e.
The average value of theoretical maximum expense and theoretical minimal-overhead is known as TiAverage cost, then will average cost according to
The proportion weighted of LimitCost and minimum total cost, just obtain for task TiFor, it is recommended that cost are as follows:
Wherein, min_Cost indicates all tasks matching total process obtained from the smallest process of cost, i.e.,
In the heuristic operator of SC, higher inspiration will be obtained by executing time closer suggestion execution the process of time
Value, therefore the heuristic operator definitions are as follows:
It is recommended that the time limit (ST) is similar to SC heuristic information, ST heuristic information wishes that selection is just able to satisfy total construction period limit
The service processes of system.Since all durations that the total construction period of workflow is not directly equal to process are added, calculation can be with
SC is different, is expressed as follows:
For all tasks of priority before i-th task, operation duration shortest clothes in service list are substituted into without exception
Business process is calculated, when obtaining the chief engineer of these tasks under this dispatching method.It is defined as appointing when calculating chief engineer
Be engaged in TiTheoretical earliest start time ESTi.It is same to substitute into conversely, for i-th task and its all tasks of later priority
Operation duration shortest service processes in service list, what when obtained chief engineer was subtracted by time limit in working hour LimitDuration
Value, then be defined as task TiTheoretical end time LFT the latesti.In scheduling problem, task TiIt both can not be earlier than ESTi
Start to execute, it is also not possible to be later than LFTiTerminate to execute, therefore TiTheoretical maximum execution time max_excuted_time be
LFTi-ESTi.Meanwhile TiMost short the executions time min_excuted_time of theory depending on efficiency in its server list most
The high machine corresponding time, i.e.,It is similarly short by maximum execution time and most to execute being averaged for time
Value is known as average performance times, then by average performance times according to the proportion weighted of LimitDuration and minimum total construction period, just
It obtains for task TiFor, it is recommended that the execution time are as follows:
Wherein min_Duration indicates that each task presses duration the smallest process distribution obtained minimum total construction period.
In the heuristic operator of ST, higher inspiration value will be obtained by executing time closer suggestion execution the process of time, therefore should
Heuristic operator definitions are as follows:
Duration-expense (TC) heuristic information comprehensively considers duration and the expense factor of service processes, it is desirable to efficiency of selection
With the better service processes of cost resultant effect.In order to allow TG and CG heuristic information to be preferably combined together, first to TG
It is normalized with the value of CG, calculates the T that goes out on missionsiAll processesAverage duration avg_excuted_
timeiAnd average overhead avg_costi.Consider further that shorter to the duration, the smaller process of expense more meets user's needs, then
Duration-Overhead under comprehensively considering is
Comprehensive performance (GP) heuristic information comprehensively considers these three factors of reliability, time and expense, it is desirable to select comprehensive
Close the best service processes of effect.It is similar to TC, in order to which the value to TG, CG, RG normalizes, the T that goes out on missions is calculated firstiIt is all
ProcessAverage duration avg_excuted_timei, average overhead avg_costiAnd mean reliability
avg_reliabilityi.Consider further that shorter to the duration, expense is smaller, and the higher process of safety more meets user's needs, obtains
To the comprehensive performance value of information are as follows:
Further, the adaptive value of all individuals of population, the calculation method of individual fitness are calculated are as follows:
Scheduling is solved, according to service processes siTri- parameter s of QoS of (1≤i≤n)i.t, si.c, si.r, tune is calculated
The qos parameter of degree solution S, is total construction period S.t, overhead S.c and overall reliability S.r respectively.
For three optimization problems of three different parameters optimization, the calculation of adaptive value are as follows:
1. total construction period optimization problem.For total construction period and overhead optimization problem, by all service processes not
The process for meeting reliability minimum requirements excludes, and task schedule is carried out in remaining process, scheduling problem can be made constant
Meet the least reliability constraint in QoS constraint.Only need to consider the constraint of overall overhead, therefore total construction period optimization problem at this time
Fitness function calculation is defined as:
Wherein, min_Cost and max_Cost respectively indicates the minimal-overhead and maximum cost of workflow schedule problem,
It can be by the way that all task schedules be obtained into the min/max service processes of expense respectively.Value function is adapted to according to this
The adaptive value calculated can satisfy as S.t≤LimitDuration, and the value of S.fitness_D will be greater than certainly to be worked as
Value when S.t > LimitDuration reaches calculation so as to assign the adaptive value of punishment to the individual for being unsatisfactory for constraint
Method tends to the requirement for meeting constraint while tending to duration optimization.
2. overhead optimization problem.It is similar to total construction period optimization problem, by exclude to be worth servicing lower than lowest reliable into
Journey simplifies problem.The then fitness function of overhead optimization problem are as follows:
Wherein, min_Time and max_Time respectively indicate workflow schedule problem the shortest duration and longest work
Phase again may be by respectively obtaining all task schedules into duration most short/longest service processes.
3. overall reliability optimization problem.To the optimization problem of overall reliability, need considering duration and expense about
While beam, the workflow schedule solution for possessing maximum reliability is found.In view of the punishment to two constraints are not met, totally may be used
By the fitness function of property optimization problem is defined as:
Wherein, min_Reliability and max_Reliability respectively indicates in all service processes reliability most
Low/corresponding reliability index of highest service processes.The fitness function of design adapts to when two constraints all meet simultaneously
Functional value is possible to be greater than 1, and becomes larger with the raising of overall reliability.Otherwise, the size of fitness function depends on
Meet the punishment of how many constraints and scheduling solution beyond constraint.
Further, the step (2) specifically: as the random number q of generation1It is wait hand over by individual mark when less than CR
The individual of fork.Crossover operator is executed to needing to be intersected individual, is matched two-by-two using the method for random fit.To every a pair
The individual of pairing generates a random integers q '1∈ [1, n] is used as crosspoint.The individual of pairing will be from q '1What is started is subsequent
All gene swapping, and obtain two two original individuals of new individual substitution.
Further, the step (3) specifically: to the individual of each variation, generate a random integers q '2∈
[1, n] it is used as catastrophe point, and for positioned at the gene T of catastrophe pointq′2The assignment, and determining T again by way of random valueq′2's
Value and original difference, successively obtain new variation individual and substitute original individual.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (5)
1. a kind of cloud computing workflow schedule method using heuristic coding strategy, which is characterized in that specific steps include:
(1) preset Population Size size generates size by the method for each gene random value to each individual
Individual；Corresponding workflow schedule scheme is generated according to using heuristic coding strategy for each individual, further according to evaluation
Function calculates the adaptive value of scheduling scheme corresponding to each individual；
(2) for each individual of existing population, a random number q is generated1∈ [0,1], and it is intersected with preset
Probability CR compares, as the q of generation1When less than CR, crossover operator is executed to the individual, is obtained using the method for single point crossing new
Individual；
(3) to each individual of the new population after intersection, a random number q is generated2∈ [0,1], and by itself and preset change
Different probability MR compares, as the q of generation2When less than MR, mutation operator is executed to the individual, using the method for single-point variation to this
Body makes a variation；
(4) fitness function evaluation is carried out to each of new population after variation individual, and replaces novel species with the optimal individual of history
The worst individual of group；
(5) new population and previous generation population are ranked up according to fitness function size, are selected in merging population optimal
Size individual enters to be recycled next time；
(6) terminate program, the corresponding scheduling scheme of output history optimum individual if reaching termination condition；Otherwise step is returned to
(2)。
2. a kind of cloud computing workflow schedule method using heuristic coding strategy according to claim 1, feature
It is, for each individual in population, gene coding are as follows:
(T1, T2..., Tn)
Wherein, n indicates the total task number of workflow；Each gene representation of individual subtask to be matched, according in workflow
The priority ranking of task；For each task Ti, region of search { RG, TG, CG, SC, ST, TC, GP } be expressed as seven kinds towards
The heuristic information of service processes；(T is encoded for the gene of individual1, T2..., Tn), according to from T1To TnSequence, to each
The heuristic information value of gene carries out the calculating of corresponding heuristic information, and to heuristic letter in respective service processes list
The highest process of breath value is matched, and a workflow schedule solution S={ s is obtained1, s2..., sn}。
3. a kind of cloud computing workflow schedule method using heuristic coding strategy according to claim 1, feature
It is, the adaptive value of all individuals of population, circular is calculated in the step (1) are as follows:
Scheduling is solved, according to service processes siTri- parameter s of QoS of (1≤i≤n)i.t, si.c, si.r, scheduling solution S is calculated
Qos parameter, be total construction period S.t, overhead S.c and overall reliability S.r respectively；
For three optimization problems of three different parameters optimization, the calculation of adaptive value is respectively as follows:
1. the calculation of the fitness function of total construction period optimization problem are as follows:
Wherein, min_Cost and max_Cost respectively indicates the minimal-overhead and maximum cost of workflow schedule problem, passes through
All task schedules are obtained into the min/max service processes of expense respectively；LimitDuration indicates restricted period
Limit；
2. the calculation of the fitness function of overhead optimization problem are as follows:
Wherein, min_Time and max_Time respectively indicates the shortest duration and longest duration of workflow schedule problem,
By the way that all task schedules are obtained into duration most short/longest service processes respectively；LimitCost indicates that limitation is opened
Pin；
3. the fitness function calculation of overall reliability optimization problem are as follows:
Wherein, min_Reliability and max_Reliability respectively indicate reliability in all service processes it is minimum/most
The corresponding reliability index of high service processes.
4. a kind of cloud computing workflow schedule method using heuristic coding strategy according to claim 1, feature
It is, the step (2) specifically: as the random number q of generation1It is individual to be intersected by individual mark when less than CR；To institute
Need to be intersected individual execution crossover operator, be matched two-by-two using the method for random fit；It is raw to the individual of every a pair of of pairing
At a random integers q '1∈ [1, n] is used as crosspoint；The individual of pairing will be from q '1The subsequent all gene swapping started,
And obtain two two original individuals of new individual substitution.
5. a kind of cloud computing workflow schedule method using heuristic coding strategy according to claim 1, feature
It is, the step (3) specifically: to the individual of each variation, generate a random integers q '2∈ [1, n] is as mutation
Point, and for positioned at the gene of catastrophe pointThe assignment, and determination again by way of random valueValue and original
Difference successively obtains new variation individual and substitutes original individual.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811416065.9A CN109597682A (en) | 2018-11-26 | 2018-11-26 | A kind of cloud computing workflow schedule method using heuristic coding strategy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811416065.9A CN109597682A (en) | 2018-11-26 | 2018-11-26 | A kind of cloud computing workflow schedule method using heuristic coding strategy |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109597682A true CN109597682A (en) | 2019-04-09 |
Family
ID=65959793
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811416065.9A Pending CN109597682A (en) | 2018-11-26 | 2018-11-26 | A kind of cloud computing workflow schedule method using heuristic coding strategy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109597682A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110928671A (en) * | 2019-12-10 | 2020-03-27 | 浙江工业大学 | Cloud workflow scheduling optimization method based on hierarchy and load balancing genetic algorithm |
| CN113010319A (en) * | 2021-03-31 | 2021-06-22 | 华南理工大学 | Dynamic workflow scheduling optimization method based on hybrid heuristic rule and genetic algorithm |
| CN114925935A (en) * | 2022-06-21 | 2022-08-19 | 福州大学 | Multi-workflow scheduling method for time delay constraint in cloud edge environment |
| CN115033373A (en) * | 2022-03-08 | 2022-09-09 | 西安电子科技大学 | Method for scheduling and unloading logic dependency tasks in mobile edge computing network |
| US20230161630A1 (en) * | 2020-03-20 | 2023-05-25 | Synapse Innovation Inc. | Method and system for resource allocation |
| CN116401055A (en) * | 2023-04-07 | 2023-07-07 | 天津大学 | Resource efficiency optimization-oriented server non-perception computing workflow arrangement method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279818A (en) * | 2013-04-25 | 2013-09-04 | 中山大学 | Method for cloud workflow scheduling based on heuristic genetic algorithm |
| US20170213128A1 (en) * | 2016-01-27 | 2017-07-27 | Bonsai AI, Inc. | Artificial intelligence engine hosted on an online platform |
| CN108170530A (en) * | 2017-12-26 | 2018-06-15 | 北京工业大学 | A kind of Hadoop Load Balancing Task Scheduling methods based on mixing meta-heuristic algorithm |
-
2018
- 2018-11-26 CN CN201811416065.9A patent/CN109597682A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279818A (en) * | 2013-04-25 | 2013-09-04 | 中山大学 | Method for cloud workflow scheduling based on heuristic genetic algorithm |
| US20170213128A1 (en) * | 2016-01-27 | 2017-07-27 | Bonsai AI, Inc. | Artificial intelligence engine hosted on an online platform |
| CN108170530A (en) * | 2017-12-26 | 2018-06-15 | 北京工业大学 | A kind of Hadoop Load Balancing Task Scheduling methods based on mixing meta-heuristic algorithm |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110928671A (en) * | 2019-12-10 | 2020-03-27 | 浙江工业大学 | Cloud workflow scheduling optimization method based on hierarchy and load balancing genetic algorithm |
| US20230161630A1 (en) * | 2020-03-20 | 2023-05-25 | Synapse Innovation Inc. | Method and system for resource allocation |
| US11989590B2 (en) * | 2020-03-20 | 2024-05-21 | Synapse Innovation Inc. | Method and system for resource allocation |
| CN113010319A (en) * | 2021-03-31 | 2021-06-22 | 华南理工大学 | Dynamic workflow scheduling optimization method based on hybrid heuristic rule and genetic algorithm |
| CN115033373A (en) * | 2022-03-08 | 2022-09-09 | 西安电子科技大学 | Method for scheduling and unloading logic dependency tasks in mobile edge computing network |
| CN114925935A (en) * | 2022-06-21 | 2022-08-19 | 福州大学 | Multi-workflow scheduling method for time delay constraint in cloud edge environment |
| CN116401055A (en) * | 2023-04-07 | 2023-07-07 | 天津大学 | Resource efficiency optimization-oriented server non-perception computing workflow arrangement method |
| CN116401055B (en) * | 2023-04-07 | 2023-10-03 | 天津大学 | Resource efficiency optimization-oriented server non-perception computing workflow arrangement method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109597682A (en) | A kind of cloud computing workflow schedule method using heuristic coding strategy | |
| Zhang et al. | Dynamic cloud task scheduling based on a two-stage strategy | |
| CN101237469B (en) | Method for optimizing multi-QoS grid workflow based on ant group algorithm | |
| Xiao et al. | Workflow scheduling in distributed systems under fuzzy environment | |
| CN104679595B (en) | A kind of application oriented IaaS layers of dynamic resource allocation method | |
| CN106357739B (en) | It is a kind of for batch cloud service request two stages combination and dispatching method | |
| Shi et al. | Location-aware and budget-constrained service deployment for composite applications in multi-cloud environment | |
| Yao et al. | New scheduling algorithms for improving performance and resource utilization in hadoop YARN clusters | |
| Adabi et al. | Bi-level fuzzy based advanced reservation of Cloud workflow applications on distributed Grid resources | |
| CN103279818A (en) | Method for cloud workflow scheduling based on heuristic genetic algorithm | |
| CN115239173A (en) | Scheduling plan generation method and device, electronic equipment and storage medium | |
| CN105809302A (en) | Production line scheduling method and system based on CEP inference engine | |
| CN107943561B (en) | A kind of scientific workflow method for scheduling task towards cloud computing platform | |
| Méndez-Hernández et al. | A multi-objective reinforcement learning algorithm for jssp | |
| Derakhshan et al. | Optimization of tasks in cloud computing based on MAX-MIN, MIN-MIN and priority | |
| Nazeri et al. | Energy aware resource provisioning for multi-criteria scheduling in cloud computing | |
| Li et al. | A QoS-based scheduling algorithm for instance-intensive workflows in cloud environment | |
| CN110138988A (en) | A kind of method and device of smart allocation visitor | |
| CN106598716B (en) | A kind of method for scheduling task based on multiprocessor | |
| CN110084507A (en) | The scientific workflow method for optimizing scheduling of perception is classified under cloud computing environment | |
| Ye et al. | Dynamic scheduling stochastic multiworkflows with deadline constraints in clouds | |
| Wan et al. | Mobile resource aware scheduling for mobile edge environment | |
| Chen et al. | Delay-optimal distributed edge computation offloading with correlated computation and communication workloads | |
| Gupta et al. | Deadline-credit aware heuristic for dynamic resource provisioning in virtualised cloud environment | |
| Megow | Coping with incomplete information in scheduling—stochastic and online models |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190409 |
|
| RJ01 | Rejection of invention patent application after publication |