CN105138404B - Towards power consumption delay and the multi-core array method for scheduling task of thermal balance - Google Patents
Towards power consumption delay and the multi-core array method for scheduling task of thermal balance Download PDFInfo
- Publication number
- CN105138404B CN105138404B CN201510650686.3A CN201510650686A CN105138404B CN 105138404 B CN105138404 B CN 105138404B CN 201510650686 A CN201510650686 A CN 201510650686A CN 105138404 B CN105138404 B CN 105138404B
- Authority
- CN
- China
- Prior art keywords
- colony
- scheduling scheme
- scheduling
- power consumption
- thermal balance
- 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
-
- 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
- Mobile Radio Communication Systems (AREA)
- General Factory Administration (AREA)
Abstract
It is towards power consumption delay and the multi-core array method for scheduling task of thermal balance, the task scheduling approach the invention discloses a kind of:By setting a colony Pt, colony PtInclude N number of scheduling scheme;Initial population P is formed using scheduling scheme caused by List scheduling algorithm and the scheduling scheme randomly generated0, and it is based on initial population P0, to colony PtLimited number of time iteration is carried out, when iterations reaches setting number, then exports colony PtComprising scheduling scheme, wherein, N is positive integer, and t is iterations.The present invention is by improving initial population and crowded strategy, and penalty is added to fitness function, power consumption, delay and the compromise of thermal balance three are selected with flexibly controlling, ensure to make power consumption while thermal balance is optimized and the cost paid that is delayed is in controlled range, and then obtain more high-quality towards power consumption delay and the multi-core array task scheduling approach of thermal balance.
Description
Technical field
It is more particularly to a kind of towards power consumption delay and the multi-core array of thermal balance the invention belongs to multinuclear processing technology field
Method for scheduling task.
Background technology
As signal of communication handles the growing of complexity, future communication systems will be supported complicated with wired resource
Data exchange and processing, existing monokaryon processing platform are difficult to meet the performance requirements such as delay power consumption.At large-scale parallel
The development of reason technology, multi-core parallel concurrent processing will substitute traditional monokaryon serial processing mode.
The first step that task scheduling is applied as signal processing system on multi-core platform, determine that each task is being handled
Position on core.The process of task scheduling as shown in figure 1, this process determine power consumption of the upper layer application on multi-core platform,
The properties such as delay, throughput, therefore, task scheduling technique are one of technologies of most critical in multinuclear processing.
The optimal task schedule of task scheduling process is a np problem.Existing task scheduling algorithm mainly considers low work(
Consumption and two targets of low delay.Because the chip area of super large-scale integration is small, packaging density is high, and chip is higher and higher
Power density causes chip temperature easily to raise, while chip overheating can cause chip operation unstable, communication and signal transacting
The delay increase of system, and local temperature is too high can reduce the reliability of system, or even cause node failure.Therefore, radiate
Problem should also be as the target considered as multiple nucleus system.
During task scheduling, while consider that the optimization design that delay, power consumption and heat are distributed is a multiple target
Optimization problem.Based on propositions such as Deb based on non-dominated sorted genetic algorithm (K.Deb, A.Pratap, S.Agarwal, et
al.A fast and elitist multiobjective genetic algorithm:NSGA-II[J]
.Evolutionary Computation,IEEE Transactions on,2002,6(2):182-197.) it is current most normal
Multi-objective Evolutionary Algorithm.
But when the number of optimization aim increases, NSGA-II algorithms can not take into account well the diversity of solution with it is more
The performance compromise of target.Simultaneously because NSGA-II algorithms solve the problems, such as distributivity estimation using crowding distance, and the mechanism
The distribution relation of individual and adjacent body before screening can only be estimated, it is impossible to reaction and the distribution of the adjacent body after screening,
So as to cause to lose the good individual of some distributivities, solution is set to be absorbed in local optimum.
Therefore, it is necessary to which the performance compromise of a kind of diversity for taking into account solution and multiple target is delayed and thermal balance towards power consumption
Multi-core array method for scheduling task.
The content of the invention
In order to solve the above technical problems, the invention provides a kind of towards power consumption delay and the multi-core array task of thermal balance
Dispatching method, the method for scheduling task are:
An if colony Pt, the colony PtInclude N number of scheduling scheme;Using scheduling scheme caused by List scheduling algorithm and
The scheduling scheme randomly generated forms initial population P0, and it is based on the initial population P0, to the colony PtLimited number of time is carried out to change
In generation, when iterations reaches setting number, then export the colony PtComprising scheduling scheme, wherein, N is positive integer, and t is repeatedly
Generation number;
Wherein, to the colony PtThe step of being iterated includes:
S1:The colony PtThe colony Q for including N number of new scheduling scheme is formed after evolutiont, and by the colony PtWith
The colony QtMerge into colony Rt;
S2:To the colony RtNon-dominated ranking is carried out, and produces all non-dominant collection F=(F1,F2,····
Fi), wherein, i is positive integer;
S3:Successively from the non-dominant collection F=(F1,F2,····Fi) in, screening scheduling scheme to colony Pt+1In,
Until the colony Pt+1Comprising N number of scheduling scheme, and make Pt=Pt+1, t=t+1;
S4:Judge iterations;The iterations is equal to setting number, then exports the colony PtComprising dispatching party
Case, the iterations are less than setting number, then continue next iteration.
According to a kind of preferred embodiment, to the colony RtCarrying out the method for non-dominated ranking includes:
S21:Based on the colony Rt, calculate the colony RtIn each scheduling scheme respectively in power consumption model, delay mould
Assessed value in type and thermal balance model;
S22:According to assessed value of the scheduling scheme in different models, with the fitness corresponding with each model
Function calculates fitness value of the scheduling scheme respectively in different models;
S23:According to fitness value of the scheduling scheme in different models, to the colony RtInterior scheduling scheme enters
Row non-dominated ranking.
According to a kind of preferred embodiment, the fitness function of power consumption model, delay model and thermal balance model is distinguished
For object function corresponding to respective model and a penaltySum, wherein,For the colony RtIn dispatching party
Case;
The penaltyFor power consumption penaltyLatency penalties functionPunished with thermal balance
Penalty functionSum, make that there is correlation between the fitness of different models.
According to a kind of preferred embodiment, in the latency penalties functionMaximum tolerance delay overhead is set
T, in the power consumption penaltyMiddle setting maximum tolerance power dissipation overhead E, in the thermal balance penaltyMiddle setting maximum tolerance thermal balance expense H;Wherein,
In the scheduling schemeCorresponding average delayTotal power consumptionWith thermal balance valueAmong at least one exceed its corresponding maximum tolerance expense, the scheduling schemeFitness value increase.
According to a kind of preferred embodiment, the maximum tolerance delay overhead T, the maximum tolerance power dissipation overhead E and
The maximum tolerance thermal balance expense H is respectively:T=(1+k) Tlist_time, E=(1+k) Elist_powerAnd H=max
(Hlist_time,Hlist_power);
Wherein, k is expressed as the constraint strength to delay and power consumption, Tlist_timeAnd Elist_powerRespectively prolonged with minimizing
When delay overhead corresponding to scheduling scheme caused by List scheduling algorithm and table during minimizing communication overhead as target when being target
Power dissipation overhead corresponding to scheduling scheme caused by dispatching algorithm, Hlist_timeAnd Hlist_powerRespectively with delay and communication overhead
For target when thermal balance expense corresponding to list scheduling scheme.
According to a kind of preferred embodiment, successively from the non-dominant collection F=(F1,F2,····Fi) in filter out
N number of scheduling scheme is to colony Pt+1Method include:
S31:I=1, colony Pt+1Assign empty set;
S32:Calculate i-th of non-dominant collection FiMaximum retain number Ni;
S33:Calculate the colony Pt+1In the quantity of scheduling scheme that currently includes be N0;
If N0With NiAnd less than N, then from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to the colony Pt+1,
I=i+1 is made, and jumps to S32;
If N0With NiAnd not less than N, then make Ni=N-N0, and from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme
To the colony Pt+1;
S34:Make Pt=Pt+1。
According to a kind of preferred embodiment, i-th of non-dominant collection F is calculatediMaximum retain number NiMethod be:
Wherein, r be one positioned at section [0,1) random number, n is the colony RtWhat is obtained after non-dominated ranking is non-
Dominate the number of plies.
According to a kind of preferred embodiment, from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to the colony
Pt+1Method be:
S331:Set a colony Paccept, the colony PacceptComprising equal in power consumption model, delay model and heat respectively
There is the scheduling scheme of Boundary Solutions in weighing apparatus model;
S332:Calculate the non-dominant collection FiIn each scheduling scheme respectively with the colony PacceptComprising scheduling
The distance of scheme, and record each scheduling scheme apart from minimum value;
S333:In record in minimum value, scheduling scheme corresponding to selection maximum is added to the colony Pt+1With
The colony Paccept, and from the non-dominant collection FiIt is middle to delete scheduling scheme corresponding to the maximum range value;
S334:Update the non-dominant collection FiWith the colony Paccept。
According to a kind of preferred embodiment, the distance between distance of the scheduling scheme is entered using euclideam norm
Row calculates.
According to a kind of preferred embodiment, the solution of the List scheduling algorithm is to be adjusted using table obtained from HLEFT algorithms
Spend initial solution.
The beneficial effects of the present invention are:It is crowded tactful and initial by improving on the basis of current NSGA-II algorithms
Population, the individual of non-dominant layer is more uniformly evolved to the next generation, to increase population diversity, prevent too early local receipts
Hold back, increase searches the possibility of more high-quality solution, and adds penalty to fitness function, flexibly to control to power consumption, prolong
When and thermal balance three compromise selection, it is ensured that make power consumption while thermal balance optimize and the cost paid of being delayed be controllable
Scope.
Brief description of the drawings
Fig. 1 is the flow chart of the iteration of the present invention;
Fig. 2 is the flow chart of the maximum distance method of the present invention.
Embodiment
Present invention solves the technical problem that it is on the premise of considering delay, power consumption and thermal balance at the same time, to task scheduling
Scheme optimizes design, has obtained more high-quality scheduling scheme.Therefore, it is necessary to build the delay of multi-core array, power consumption respectively
With thermal balance mathematical modeling.
Specifically, the assessed value calculating formula of the averaging network delay of wireless multi-core array is as follows:
Wherein, | E | represent message number;t_transi,jThe task i of representative sends data to j transmission time, t_
transi,jMathematic(al) representation it is as follows:
Wherein, hopsi,jRepresent the hop count of packet process, ci,jThe data traffic between task i and j is represented,
DR is the data transmission rate of wireless transmission antenna, tarb,kAnd tcont,kIt is arbitration of the packet when kth during being wirelessly transferred is jumped
Taken with channel competition.
Wirelessly the power consumption assessment value expression of multi-core array system is:
Etotal=Ecompute_total+Etrans_total
Wherein, Ecompute_totalAll processing units in wireless multi-core array are represented to produce when each task is handled
Power consumption, its calculating formula is as follows:
Wherein,Power consumption corresponding to processing unit k is represented, is in be incremented by relation with processing unit percentage load.
Etrans_totalPower consumption caused by data transmission procedure, a bit data in wireless multi-core array is represented to pass in a network
Defeated one power consumption for jumping needs is e0, Etrans_totalMathematic(al) representation it is as follows:
Due to the power consumption on resource node and its temperature positive correlation, therefore, using resource node i total power consumption EiTo characterize
Its temperature level, its calculating formula are:
Ei=Ecomp_i+Ecomm_i
Wherein, Ecomp_iIt is the calculating power consumption of node i;Ecomm_iIt is power consumption caused by node i processing communication data, it is calculated
Formula is:
Wherein, Kj,kIt is from task j to the set of all routing nodes on task k routed path.Thermal balance assessed value
Expression formula is as follows:
ut=var sum (s) | s ∈ St(M)}
Wherein, by the total power consumption E on each resource nodei, form matrix a M, St(M) all t divided by matrix M
The submatrix s compositions of × t sizes, parameter t characterize thermal conductivity.Sum (s) represents (the i.e. resource node of all elements in submatrix s
Power consumption) sum.Thermal balance assessed value is used for assessing the uniformity coefficient of the heat distribution of wireless multi-core array, and value is smaller, represents platform
Heat distribution it is more balanced.
Delay, power consumption and thermal balance mathematical modeling based on above-mentioned multi-core array, the task scheduling side of unlimited multi-core array
The target of case optimization is to minimize delay, power consumption and thermal balance assessed value.That is, solving scheduling scheme map (V), and make
Scheduling scheme map (V), meets following expression:
Wherein, task distribution principle, load restraint and bandwidth constraint are expressed as:
Wherein, xi,jTask v is represented for 1iIt is projected in processing unit tj, piRepresent task viOperation time, hlFor 1 generation
Watch chain road l is task v in multi-core array resource mapiTransmit data to vjRouted path in a certain bar.
It is described in detail below in conjunction with the accompanying drawings.
The flow chart of iteration of the invention with reference to shown in Fig. 1;Wherein, if a colony Pt, in colony PtIt is interior comprising N number of
Scheduling scheme.Also, form initial population P using scheduling scheme caused by List scheduling algorithm and the scheduling scheme randomly generated0,
And it is based on initial population P0, to colony PtLimited number of time iteration is carried out, when iterations reaches setting number, then exports colony PtBag
The scheduling scheme contained, wherein, N is positive integer, and t is iterations.
Specifically, the solution of List scheduling algorithm is using HLEFT (Highest Level First with Estimated
Times) list scheduling initial solution obtained from algorithm.
Wherein, to colony PtThe step of being iterated includes:
S1:Colony PtThe colony Q for including N number of new scheduling scheme is formed after evolutiont, and by colony PtWith colony QtClose
And it is colony Rt。
Specifically, using tournament algorithm come the colony P that evolvest, crossover probability and mutation probability need to be set, makes colony PtIn
Scheduling scheme N number of new scheduling scheme is generated after intersection and variation, so as to form colony Qt, then by colony PtInterior N number of tune
Degree scheme and colony QtN number of scheduling scheme merge form colony Rt。
S2:To colony RtNon-dominated ranking is carried out, and produces all non-dominant collection F=(F1,F2,····Fi), its
In, i is positive integer.
Specifically, S21:Based on colony Rt, with reference to the power consumption model of above-mentioned structure, delay model and thermal balance model and
The assessed value calculating formula of corresponding model, calculates colony RtIn each scheduling scheme respectively in power consumption model, delay model and heat
Assessed value in equilibrium model.
S22:According to assessed value of each scheduling scheme in different models, with the fitness corresponding with each model
Function calculates fitness value of each scheduling scheme respectively in different models.
Wherein, the assessed value according to each scheduling scheme in different models, and fitted with corresponding with each model
Response function calculate each scheduling scheme respectively fitness value in different models when, present invention improves over each model
Corresponding fitness function.
Specifically, the fitness function of power consumption model, delay model and thermal balance model is respectively corresponding to respective model
Object function and a penaltySum, wherein,It is expressed as colony RtIn scheduling scheme.
Wherein, the expression formula of power consumption model, delay model and fitness function corresponding to thermal balance model is respectively:
Wherein, power consumption model, delay model and object function corresponding to thermal balance model are as follows:
PenaltyFor power consumption penaltyLatency penalties functionLetter is punished with thermal balance
NumberSum, i.e. penaltyExpression formula be:
By the way that penalty corresponding to each model is superimposed, make that there is phase between the fitness value of different models
Guan Xing, while performance of the scheduling scheme on delay, power consumption and thermal balance is influenced each other and containing, so as to obtain one
There is the scheduling scheme of compromise performance for delay, power consumption and thermal balance.
Specifically, power consumption penaltyLatency penalties functionWith thermal balance penaltyExpression formula be respectively:
Wherein, in latency penalties functionMaximum tolerance delay overhead T is set, in power consumption penaltyMiddle setting maximum tolerance power dissipation overhead E, in thermal balance penaltyMiddle setting maximum tolerance thermal balance is opened
Sell H;Wherein,
In scheduling schemeCorresponding average delayTotal power consumptionWith thermal balance valueIt
In at least one exceed its corresponding maximum tolerance expense, scheduling schemeFitness value increase.
Moreover, maximum tolerance delay overhead T, maximum tolerance power dissipation overhead E and maximum tolerance thermal balance expense H are respectively:T
=(1+k) Tlist_time, E=(1+k) Elist_powerWith H=max (Hlist_time,Hlist_power)。
Wherein, k is expressed as the constraint strength to delay and power consumption, Tlist_timeAnd Elist_powerRespectively prolonged with minimizing
When delay overhead corresponding to scheduling scheme caused by List scheduling algorithm and table during minimizing communication overhead as target when being target
Power dissipation overhead corresponding to scheduling scheme caused by dispatching algorithm, Hlist_timeAnd Hlist_powerRespectively with delay and communication overhead
For target when thermal balance expense corresponding to list scheduling scheme.
S23:The finally fitness value according to each scheduling scheme in different models, to colony RtInterior scheduling scheme enters
Row non-dominated ranking.
Specifically, it is above-mentioned to colony RtInterior scheduling scheme carries out non-dominated ranking and proposes non-dominant row using Deb etc.
Sequence genetic algorithm (K.Deb, A.Pratap, S.Agarwal, et al.A fast and elitist multiobjective
genetic algorithm:NSGA-II[J].Evolutionary Computation,IEEE Transactions on,
2002,6(2):182-197.)。
Simultaneously as fitness value of the scheduling scheme in each model is smaller, then the scheduling scheme is more high-quality, and its is preferential
Level it is higher, that is, when scheduling scheme each model fitness value increase, it can be caused to be in dominance relation
Inferior position, and in the lower level dominated in layer, thus it is more beneficial for the more high-quality scheduling scheme of screening and enters of future generation plant
Group.
S3:Successively from non-dominant collection F=(F1,F2,····Fi) in, screening scheduling scheme to colony Pt+1In, until
Colony Pt+1Comprising N number of scheduling scheme, and make Pt=Pt+1, t=t+1.
Specifically, from non-dominant collection F=(F1,F2,····Fi) in filter out the method for colony of future generation and be:
S31:I=1, colony Pt+1Assign empty set.
S32:Calculate i-th of non-dominant collection FiMaximum retain number Ni;Wherein, i initial values are 1, calculate i-th of non-branch
With collection FiMaximum retain number NiMethod be:
Wherein, r be one positioned at section [0,1) random number, n is colony RtWhat is obtained after non-dominated ranking is non-dominant
The number of plies.
S33:Calculate the colony Pt+1In the quantity of scheduling scheme that currently includes be N0;Wherein,
If N0With NiAnd less than N, then from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to the colony Pt+1,
I=i+1 is made, and jumps to S32.
If N0With NiAnd not less than N, then make Ni=N-N0, and from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme
To the colony Pt+1。
Specifically, from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to colony Pt+1.From i-th of non-dominant collection Fi
In filter out and maximum retain number NiMethod can direct random screening, or using the screening of maximum distance method.
S34:Make Pt=Pt+1.That is, generate new generation population, then need to judge iteration time, that is, determine be after
Continuous iteration, or output scheduling scenario outcomes.
S4:Judge iterations t;Iterations t is equal to setting number tmax, then colony P is exportedtComprising scheduling scheme,
Iterations t is less than setting number tmax, then next iteration is continued.
The present invention adds penalty on the basis of current NSGA-II algorithms, to fitness function, with flexibly control pair
Power consumption, delay and the compromise of thermal balance three selection, it is ensured that make power consumption while thermal balance is optimized and be delayed the generation paid
Valency is in controlled range.
The flow chart of maximum distance method of the invention with reference to shown in Fig. 2;Wherein, the present invention using maximum distance method from the
I non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to colony Pt+1Method be:
S331:Set a colony Paccept, colony PacceptComprising respectively in power consumption model, delay model and thermal balance mould
There is the scheduling scheme of Boundary Solutions in type.That is, in colony PacceptScheduling scheme in power consumption model, delay model and heat
In three corresponding target function values in equilibrium model, minimum of one is the maximum or minimum of the object function of corresponding model
Value.
S332:Calculate non-dominant collection FiIn each scheduling scheme respectively with colony PacceptComprising each scheduling scheme
The distance between, and record each scheduling scheme and colony PacceptComprising scheduling scheme distance minimum value.
S333:In the minimum range of record, scheduling scheme corresponding to maximum therein is chosen to colony Pt+1, simultaneously
The scheduling scheme is added to colony Paccept, and from non-dominant collection FiMiddle deletion scheduling scheme.
S334:Update non-dominant collection FiWith colony Paccept.Then S34 is entered.
Used in the present invention in maximum distance method, the distance between scheduling scheme is calculated using euclideam norm.
It can either avoid in the case where target increases, error increases, and further through crowded strategy is improved, makes the individual of non-dominant layer more
The next generation is equably evolved to, to increase population diversity, prevents too early local convergence, what increase searched more high-quality solution can
Can property.
It should be noted that above-mentioned specific embodiment is exemplary, those skilled in the art can disclose in the present invention
Various solutions are found out under the inspiration of content, and these solutions also belong to disclosure of the invention scope and fall into this hair
Within bright protection domain.It will be understood by those skilled in the art that description of the invention and its accompanying drawing are illustrative and are not
Form limitations on claims.Protection scope of the present invention is limited by claim and its equivalent.
Claims (10)
- It is 1. a kind of towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that the task scheduling Method is:An if colony Pt, the colony PtInclude N number of scheduling scheme;Using scheduling scheme caused by List scheduling algorithm and with Scheduling scheme caused by machine forms initial population P0, and it is based on the initial population P0, to the colony PtLimited number of time is carried out to change In generation, when iterations reaches setting number, then export the colony PtComprising scheduling scheme, wherein, N is positive integer, and t is repeatedly Generation number;Wherein, to the colony PtThe step of being iterated includes:S1:The colony PtThe colony Q for including N number of new scheduling scheme is formed after evolutiont, and by the colony PtWith it is described Colony QtMerge into colony Rt;S2:To the colony RtNon-dominated ranking is carried out, and produces all non-dominant collection F=(F1,F2,····Fi), its In, i is positive integer;S3:Successively from the non-dominant collection F=(F1,F2,····Fi) in, screening scheduling scheme to colony Pt+1In, until The colony Pt+1Comprising N number of scheduling scheme, and make Pt=Pt+1, t=t+1;S4:Judge iterations;The iterations is equal to setting number, then exports the colony PtComprising scheduling scheme, institute State iterations and be less than setting number, then continue next iteration.
- 2. as claimed in claim 1 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that To the colony RtCarrying out the method for non-dominated ranking includes:S21:Based on the colony Rt, calculate the colony RtIn each scheduling scheme respectively power consumption model, delay model and Assessed value in thermal balance model;S22:According to assessed value of the scheduling scheme in different models, with the fitness function corresponding with each model Calculate fitness value of the scheduling scheme respectively in different models;S23:According to fitness value of the scheduling scheme in different models, to the colony RtInterior scheduling scheme carries out non- Dominated Sorting.
- 3. as claimed in claim 2 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that The fitness function of power consumption model, delay model and thermal balance model is respectively that object function corresponding to respective model is punished with one Penalty functionSum, wherein,For the colony RtIn scheduling scheme;The penaltyFor power consumption penaltyLatency penalties functionLetter is punished with thermal balance NumberSum, make that there is correlation between the fitness of different models.
- 4. as claimed in claim 3 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that In the latency penalties functionMaximum tolerance delay overhead T is set, in the power consumption penaltyIn set Maximum tolerance power dissipation overhead E is put, in the thermal balance penaltyMiddle setting maximum tolerance thermal balance expense H;Its In,In the scheduling schemeCorresponding average delayTotal power consumptionWith thermal balance valueIt In at least one exceed its corresponding maximum tolerance expense, the scheduling schemeFitness value increase.
- 5. as claimed in claim 4 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that The maximum tolerance delay overhead T, the maximum tolerance power dissipation overhead E and the maximum tolerance thermal balance expense H are respectively:T =(1+k) Tlis_t tim, E=(1+k) Elist_powerWith H=max (Hlist_time,Hlist_power);Wherein, k is expressed as the constraint strength to delay and power consumption, Tlist_timeAnd Elist_powerRespectively it is delayed with minimizing as mesh Delay overhead corresponding to scheduling scheme caused by timestamp List scheduling algorithm and list scheduling is calculated during minimizing communication overhead as target Power dissipation overhead corresponding to scheduling scheme caused by method, Hlist_timeAnd Hlist_powerRespectively using delay and communication overhead as target When thermal balance expense corresponding to list scheduling scheme.
- 6. as claimed in claim 1 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that Successively from the non-dominant collection F=(F1,F2,…Fi) in filter out N number of scheduling scheme to colony Pt+1Method include:S31:I=1, colony Pt+1Assign empty set;S32:Calculate i-th of non-dominant collection FiMaximum retain number Ni;S33:Calculate the colony Pt+1In the quantity of scheduling scheme that currently includes be N0;If N0With NiAnd less than N, then from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to the colony Pt+1, make i=i + 1, and jump to S32;If N0With NiAnd not less than N, then make Ni=N-N0, and from i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to institute State colony Pt+1;S34:Make Pt=Pt+1。
- 7. as claimed in claim 6 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that Calculate i-th of non-dominant collection FiMaximum retain number NiMethod be:<mrow> <msub> <mi>N</mi> <mi>i</mi> </msub> <mo>=</mo> <mi>N</mi> <mfrac> <mrow> <mn>1</mn> <mo>-</mo> <mi>r</mi> </mrow> <mrow> <mn>1</mn> <mo>-</mo> <msup> <mi>r</mi> <mi>n</mi> </msup> </mrow> </mfrac> <msup> <mi>r</mi> <mrow> <mi>i</mi> <mo>-</mo> <mn>1</mn> </mrow> </msup> </mrow>Wherein, r be one positioned at section [0,1) random number, n is the colony RtWhat is obtained after non-dominated ranking is non-dominant The number of plies.
- 8. as claimed in claim 6 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that From i-th of non-dominant collection FiMiddle selection NiIndividual scheduling scheme is to the colony Pt+1Method be:S331:Set a colony Paccept, the colony PacceptComprising respectively in power consumption model, delay model and thermal balance mould There is the scheduling scheme of Boundary Solutions in type;S332:Calculate the non-dominant collection FiIn each scheduling scheme respectively with the colony PacceptComprising scheduling scheme Distance, and record each scheduling scheme apart from minimum value;S333:In record in minimum value, scheduling scheme corresponding to selection maximum is added to the colony Pt+1With it is described Colony Paccept, and from the non-dominant collection FiIt is middle to delete scheduling scheme corresponding to the maximum range value;S334:Update the non-dominant collection FiWith the colony Paccept。
- 9. as claimed in claim 8 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, it is characterised in that The distance between distance of the scheduling scheme is calculated using euclideam norm.
- 10. exist as claimed in claim 1 towards power consumption delay and the multi-core array method for scheduling task of thermal balance, its feature In the solution of the List scheduling algorithm is using list scheduling initial solution obtained from HLEFT algorithms.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510650686.3A CN105138404B (en) | 2015-10-09 | 2015-10-09 | Towards power consumption delay and the multi-core array method for scheduling task of thermal balance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510650686.3A CN105138404B (en) | 2015-10-09 | 2015-10-09 | Towards power consumption delay and the multi-core array method for scheduling task of thermal balance |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105138404A CN105138404A (en) | 2015-12-09 |
CN105138404B true CN105138404B (en) | 2018-04-03 |
Family
ID=54723760
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510650686.3A Active CN105138404B (en) | 2015-10-09 | 2015-10-09 | Towards power consumption delay and the multi-core array method for scheduling task of thermal balance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105138404B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105959960B (en) * | 2016-06-14 | 2019-10-18 | 中国人民解放军国防科技大学 | Frequency spectrum decision method based on evolutionary multi-objective optimization matching |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102508708A (en) * | 2011-11-30 | 2012-06-20 | 湖南大学 | Heterogeneous multi-core energy-saving task schedule method based on improved genetic algorithm |
CN103345657A (en) * | 2013-04-02 | 2013-10-09 | 江苏大学 | Task scheduling method based on heredity and ant colony in cloud computing environment |
CN103576829A (en) * | 2012-08-01 | 2014-02-12 | 复旦大学 | Hybrid genetic algorithm based dynamic cloud-computing virtual machine scheduling method |
CN103870317A (en) * | 2012-12-10 | 2014-06-18 | 中兴通讯股份有限公司 | Task scheduling method and system in cloud computing |
CN104516785A (en) * | 2014-12-19 | 2015-04-15 | 上海电机学院 | Cloud computing resource scheduling system and method |
-
2015
- 2015-10-09 CN CN201510650686.3A patent/CN105138404B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102508708A (en) * | 2011-11-30 | 2012-06-20 | 湖南大学 | Heterogeneous multi-core energy-saving task schedule method based on improved genetic algorithm |
CN103576829A (en) * | 2012-08-01 | 2014-02-12 | 复旦大学 | Hybrid genetic algorithm based dynamic cloud-computing virtual machine scheduling method |
CN103870317A (en) * | 2012-12-10 | 2014-06-18 | 中兴通讯股份有限公司 | Task scheduling method and system in cloud computing |
CN103345657A (en) * | 2013-04-02 | 2013-10-09 | 江苏大学 | Task scheduling method based on heredity and ant colony in cloud computing environment |
CN104516785A (en) * | 2014-12-19 | 2015-04-15 | 上海电机学院 | Cloud computing resource scheduling system and method |
Also Published As
Publication number | Publication date |
---|---|
CN105138404A (en) | 2015-12-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111445111B (en) | Electric power Internet of things task allocation method based on edge cooperation | |
CN112469100B (en) | Hierarchical routing algorithm based on rechargeable multi-base-station wireless heterogeneous sensor network | |
CN104185242B (en) | A kind of wireless Mesh netword channel distribution for reducing interference and route combined optimization system and method | |
CN101959244B (en) | Method for controlling hierarchical type route suitable for wireless sensor network | |
CN104937901B (en) | For providing the method for the traffic engineering of routing and storage in the network of content oriented | |
CN105246097B (en) | A kind of wireless sense network optimization method for survival time with mobile Sink node | |
CN104166630B (en) | Optimization caching laying method based on prediction in a kind of content oriented central site network | |
Ouyang et al. | Importance-different charging scheduling based on matroid theory for wireless rechargeable sensor networks | |
Chen et al. | Albrl: Automatic load-balancing architecture based on reinforcement learning in software-defined networking | |
CN111416356A (en) | Transmission and distribution network linkage optimization method based on alternating direction multiplier method and optimal power flow | |
Jin et al. | Latency and energy-consumption optimized task allocation in wireless sensor networks | |
CN105323166A (en) | Cloud computing-oriented routing method based on network efficiency priority | |
Boukerche et al. | A novel joint optimization method based on mobile data collection for wireless rechargeable sensor networks | |
CN105138404B (en) | Towards power consumption delay and the multi-core array method for scheduling task of thermal balance | |
Tukisi et al. | Multi-hop PSO based routing protocol for Wireless Sensor Networks with Energy Harvesting | |
Peng et al. | An improved energy-aware routing protocol using multiobjective particular swarm optimization algorithm | |
CN109152103A (en) | A kind of pair of wireless sensor network carries out network-building method and device | |
CN103995749B (en) | The calculation task allocating method and system of cell cloud system | |
Srivastava et al. | Load-balanced cluster head selection enhancing network lifetime in WSN using hybrid approach for IoT applications | |
Shen et al. | An improved genetic algorithm for 0-1 knapsack problems | |
Wang et al. | Charging path optimization for wireless rechargeable sensor network | |
CN108183867A (en) | Information centre's network node buffer replacing method | |
CN108093056A (en) | Information centre's wireless network virtualization nodes buffer replacing method | |
CN106717084A (en) | System and method for overlapping rate region zoning | |
CN103619047A (en) | Opportunistic routing method in multiple-concurrent-flow wireless mesh network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |