CN106951059A - Based on DVS and the cloud data center power-economizing method for improving ant group algorithm - Google Patents
Based on DVS and the cloud data center power-economizing method for improving ant group algorithm Download PDFInfo
- Publication number
- CN106951059A CN106951059A CN201710191575.XA CN201710191575A CN106951059A CN 106951059 A CN106951059 A CN 106951059A CN 201710191575 A CN201710191575 A CN 201710191575A CN 106951059 A CN106951059 A CN 106951059A
- Authority
- CN
- China
- Prior art keywords
- virtual machine
- task
- dvs
- formula
- ant
- 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
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000005265 energy consumption Methods 0.000 claims abstract description 37
- 230000006872 improvement Effects 0.000 claims abstract description 22
- 238000007726 management method Methods 0.000 claims abstract description 16
- 230000001105 regulatory effect Effects 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 18
- 239000003016 pheromone Substances 0.000 claims description 9
- 230000001276 controlling effect Effects 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 241000257303 Hymenoptera Species 0.000 claims description 3
- 238000004134 energy conservation Methods 0.000 claims description 3
- PSGAAPLEWMOORI-PEINSRQWSA-N medroxyprogesterone acetate Chemical compound C([C@@]12C)CC(=O)C=C1[C@@H](C)C[C@@H]1[C@@H]2CC[C@]2(C)[C@@](OC(C)=O)(C(C)=O)CC[C@H]21 PSGAAPLEWMOORI-PEINSRQWSA-N 0.000 claims description 3
- 239000011800 void material Substances 0.000 claims description 3
- 230000009977 dual effect Effects 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 9
- 238000005457 optimization Methods 0.000 abstract description 7
- 238000004364 calculation method Methods 0.000 abstract description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005059 dormancy Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/26—Power supply means, e.g. regulation thereof
- G06F1/32—Means for saving power
- G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
- G06F1/3234—Power saving characterised by the action undertaken
- G06F1/329—Power saving characterised by the action undertaken by task scheduling
-
- 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
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
-
- 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
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
- G06F2009/4557—Distribution of virtual machine instances; Migration and load balancing
-
- 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)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Power Sources (AREA)
Abstract
The invention provides a kind of cloud data center power-economizing method based on DVS with improvement ant group algorithm, it is related to field of cloud calculation to be combined DVS technologies with optimization ant group algorithm task scheduling technique, in terms of physical regulating and task scheduling algorithm two, integrated power-saving management is carried out to cloud data center, low energy consumption target is reached while proof load balance, best performance., can be according to loading condition dynamic regulation server voltage using DVS integrated management devices;Using Load Balancing Manager, alternative virtual machine can be selected before task scheduling, for treating that scheduler task is selected, the purpose of load balancing can be so reached;In task dispatcher, using improved ant colony dispatching algorithm, performance, Energy Consumption Factors are incorporated into dispatching algorithm, task scheduling and execution time are not only shortened, and reduce energy consumption.
Description
Technical field
The present invention relates to cloud data center Energy Saving Algorithm technical field, more particularly to a kind of DVS that is based on is with improving ant colony calculation
The cloud data center power-economizing method of method.
Background technology
Cloud computing is a kind of emerging computation model developed by grid computing, is the product that multiple technologies are blended.
Cloud computing service business is directed to providing the user efficient, easily a variety of calculating services, such as IaaS, PaaS, SaaS.Cloud computing
Pattern is very easy to user, it is bought and configures the computing device of complexity, need to only apply according to the need for oneself
Computing resource simultaneously pays corresponding expense, while decreasing spending.The cloud computing service chamber of commerce is according to service condition dynamic
Cloud computing system is extended or downsizing on ground.
With the arrival in big data epoch, calculating demand constantly expands, and the scale of cloud data center is also increasing, and its is huge
Energy consumption cost and carbon emission amount turn into puzzlement cloud supplier a problem, show according to the study, U.S. Data center in 2014
The electric power of about 70,000,000,000 kilowatt hours of consumption, occupies the 2% of american energy consumption, how to be saved while ensureing and calculating performance altogether
Energy consumption cost turns into the hot issue of a concern.
For the high energy consumption issues of data center, numerous studies work has been done by domestic and international related scientific research mechanism, public organization.
At present, the energy optimization scheme of cloud data center mainly includes 3 classes:Intel Virtualization Technology, dynamic voltage scaling technology, closing/stop
Dormancy technology.For example, Berral et al. utilizes machine learning techniques, virtual machine is dynamically integrated, calculating task is being met
Minimum (Zomaya AY, Lee the Y C.8.Toward Energy-Aware Scheduling of energy ezpenditure on the premise of SLA
Using Machine Learning[M]Energy-Efficient Distributed Computing Systems.John
Wiley&Sons,Inc.2012:215-244.);Xiao Peng et al. proposes a kind of scheduling virtual machine based on energy consumption scale model
Algorithm, is assessed each recent energy consumption of virtual machine and is scheduled using nearest least energy consumption ratio preference strategy, so as to reach load
It is balanced with energy consumption that (Xiao Peng, Liu Dongbo bend scheduling virtual machine algorithm [J] based on energy consumption scale model in happiness dragon cloud computings
Electronic letters, vol, 2015,43 (2):305-311.);Xia et al. proposes one at random based on DVS in the application of cloud data center
Analytical framework, the framework can help to design and optimize energy consumption perception high-performance cloud system (XiaYN, Zhou M C, Luo X,
et al.A Stochastic Approach to Analysis of Energy-Aware DVS-Enabled Cloud
Datacenters[J].IEEE Transactions on Systems Man&Cybernetics Systems,2014,45
(1):1-1.), but author do not provide specifically can good Managed Solution;Rational task scheduling can also reach to a certain extent
To the optimization of performance and energy consumption, Tan Yiming et al. to cloud system by modeling with analyzing, it is proposed that one kind passes through task scheduling side
Cloud computing can be greatly lowered on the premise of execution performance is ensured in the energy optimization management method ME3PC algorithms of formula, the algorithm
Energy consumption expense (Tan Yiming, the Zeng Guosun, the optimum management method of Wang Wei's Random Tasks energy consumption in cloud computing platform of system
[J] Journal of Software, 2012,23 (2):266-278.), but this method depends on the distribution probability that task is reached, it is special in processing
Defect is there may be during different task flow.The scheduling of cloud task belongs to np complete problem, can be solved using heuristic mutation operations method,
Ant group algorithm is a kind of simulated evolutionary algorithm, and research shows that the algorithm has advantageous property, can be used to solve cloud task scheduling
Problem.Wei Yun et al. is scheduled using a kind of improvement ant group algorithm based on most short task time delay, public taking into account scheduling
On the premise of levelling and efficiency, maximizing shortening task time delay, (Wei's Yun, Chen Yuan members is based on the cloud meter for improving ant group algorithm
Calculate Task Scheduling Model [J] computer engineering, 2015,41 (2):12-16.);Look into pacify the people et al. and be directed to the spy of cloud computing environment
Point, is improved to population and ant group algorithm, proposes a kind of task scheduling algorithm of both fusions, can reduce the total complete of task
(look into and pacify the people, cloud computing task scheduling algorithm [J] the computer technologies of Tan Wenan fusion populations and ant colony are with sending out into the time
Exhibition, 2016,26 (8):24-29.).But these algorithms only account for the optimization to performance, do not account for energy consumption problem.
In summary, existing cloud data center energy-saving task schedule method also imperfection, energy-saving scheme is excessively single.Have
In consideration of it, being necessary to propose a kind of cloud data center power-economizing method based on DVS with improvement ant group algorithm.
The content of the invention
In order to solve the deficiency of existing cloud data center energy-saving scheme, equalization performance and energy consumption, make cloud data center comprehensive
Maximum revenue, the present invention proposes a kind of cloud data center power-economizing method based on DVS with improvement ant group algorithm, introduces DVS
Integrated management device, is combined with ant colony dispatching method, and energy-conservation regulation and control are carried out in terms of physical host and scheduling scheme two, performance is reached
It is optimal with energy consumption.
The present invention specifically uses following technical scheme:
Based on DVS with improve ant group algorithm cloud data center power-economizing method, DVS integrated managements device obtain physical host and
Virtual machine state information, dynamic regulation physical host voltage reaches energy-conservation purpose, and virtual machine state information is reached is based on
The energy saving task scheduler module of ant group algorithm, is that each task distributes suitable virtual machine, physical host, virtual machine Double regulating
Control, reaches that performance is optimal with energy consumption, it is characterised in that specifically include following steps:
Step 1:DVS integrated management devices obtain each physical host and its virtual machine running state information, and calculating main frame is born
Carry, main frame is regulated and controled according to presetting threshold xi;
Step 2:According to virtual machine and waiting list state, Load Balancing Manager selects available m virtual machine to add
Alternate list;
Step 3:M task before task queue is obtained, the financial value of calculating task on a virtual machine is (during prediction tasks carrying
Between and energy consumption of virtual machine), obtain the prediction gain matrix of tasks carrying;
Step 4:Energy saving task scheduler module obtains the virtual machine state data of Load Balancing Manager, is each available void
Plan machine computing node assigns initial information element;
Step 5:Initialization, ant is randomly placed on available virtual machine;
Step 6:Calculate ant k and move on to next node probability, according to probability selection next node;
Step 7:Judge whether ant k completes epicycle search, if so, being updated to local information element, held if it is not, returning
Row step 6;
Step 8:Judge whether all ants complete this search, if so, carrying out global letter to this search optimal path
Breath element is updated, and step 6 is performed if it is not, returning;
Step 9:Judge whether to meet ant termination condition, if so, this group task optimal distributing scheme is exported, by task point
It is assigned to correspondence virtual machine to perform, step 5 is performed if it is not, returning;
Step 10:Judge whether also have task to be allocated in task queue, step 1 is performed if so, returning, if nothing, terminate
This subtask is distributed.
As a further improvement on the present invention, in the step 1, DVS integrated management devices are continuously evaluated every physical host
Task load, main frame i load Load (i) is determined according to following formula:
In formula, RunVM (i) represents the virtual machine quantity being currently running on main frame i, and MaxVM (i) represents to hold on main frame i
The maximum virtual machine quantity of load, sets η value as 0.5, and when load on host computers is more than η, DVS control units are carried by regulating and controlling voltage
CPU speed steps are risen, DVS control units reduce CPU speed steps by regulating and controlling voltage when load on host computers is less than ξ.
As a further improvement on the present invention, in the step 3, the revenue function I that task i is performed on virtual machine j
(xi,j) determined according to following formula:
I(xi,j)=B (xi,j)-P(xi,j)
In formula, B (xi,j) represent performance revenue function, P (xi,j) represent energy consumption overhead functions.
Performance benefits function B (xi,j) according to the following formula calculate obtain:
In formula, λ is a regulation parameter, Time (xi,j) represent that predictions of the task i on virtual machine j performs the time, according to
Following formula is calculated and obtained:
Time(xi,j)=γ (LoadVMj+Ni)
In formula, γ is empirical parameter, is obtained by training, LoadVMjRepresent virtual machine j current load situations, NiRepresent pre-
Increased load is surveyed, its value determines that span is N according to task sizei=0.1~0.4.
Energy consumption overhead functions P (xi,j) according to the following formula calculate obtain:
P(xj)=λ ' (acpuucpu,j+amenNLLC,j+aiobio,j)
In formula, λ ' is a regulation parameter, acpu、amen、aioThe specific constant of model is represented, can be obtained by training,
ucpu,jRepresent virtual machine j processor utilization, NLLC,jRepresent virtual machine j last layer on all cores in certain period of time
Cache missing times, bio,jRepresent virtual machine j read-write total bytes.
As a further improvement on the present invention, in the step 4, pheromones τs of the virtual machine j in initial timej(0) according to
Following formula is calculated:
τj(0)=bcpu(numj×Pj)+bmemRj+bnetWj
In formula, bcpu、bmem、bnetCPU information element, memory information element, the plain weight coefficient of bandwidth information, num are represented respectivelyj
Represent the quantity of virtual machine j processors, PjThe processing speed of each processors of virtual machine j is represented, unit is MIPS, RjRepresent empty
Plan machine j internal memory surpluses, unit is M, WjVirtual machine j amount of bandwidth is represented, unit is Mbps.
As a further improvement on the present invention, in the step 6, ant k moves on to the computing formula of the probability of next node
For:
In formula, Pj k(t) t ant k selection virtual machines j probability is represented, α represents money as information heuristic greedy method
Source information element significance level, β represents the relative importance of control desired value, τ as expected heuristic valuej(t) t is represented
Moment virtual machine j pheromone concentration, ηjThe expected degree that expression task i is performed on virtual machine j, ηj=I (xi,j)。
As a further improvement on the present invention, in the step 7, it is to the formula that local information element is updated:
τj(t+1)=(1- ρ) τj(t)+Δτj
In formula, ρ is volatilization factor, meets 0≤ρ<1,1- ρ represents that pheromones remain degree, Δ τjComputing formula is:
Δτj=D1·RES(Ik,l)
In formula, Ik,lRepresent the task allocative decision that kth ant searches in the l times iteration, RES (Ik,l) value according to
This allocative decision revenue function value ∑ I (xi,j) determine, D1It is constant.
As a further improvement on the present invention, in the step 8, global information element is updated, Δ τjComputing formula is:
Δτj=D2·MAX(RES(Ik,l))
In formula, MAX (RES (Ik,l)) represent the optimal distributing scheme that k Ant Search is arrived in the l times iteration, D2It is normal
Amount.
The technical solution adopted by the present invention compared with prior art, with following technique effect:
The present invention devises a kind of new cloud data center comprehensive energy-saving method, and DVS technologies are appointed with optimization ant group algorithm
Business dispatching technique is combined, in terms of physical regulating and task scheduling algorithm two, and integrated power-saving management is carried out to cloud data center,
Low energy consumption target is reached while proof load balance, best performance., can be according to loading condition using DVS integrated management devices
Dynamic regulation server voltage;Using Load Balancing Manager, alternative virtual machine can be selected before task scheduling, for waiting to adjust
Task choosing is spent, the purpose of load balancing can be so reached;, will using improved ant colony dispatching algorithm in task dispatcher
Performance, Energy Consumption Factors incorporate dispatching algorithm, not only shorten task scheduling and execution time, and reduce energy consumption.
Brief description of the drawings
Fig. 1 is the cloud data center control flow model of the inventive method.
Fig. 2 is based on DVS and the cloud data center power-economizing method flow chart for improving ant group algorithm.
Embodiment
Technical scheme is described in detail below in conjunction with the accompanying drawings:
Cloud data center controlling stream, its basic model are as shown in figure 1, task enters waiting list, DVS integrated pipes after reaching
Reason device contains DVS controllers and data center's monitor supervision platform two parts, and wherein DVS controllers are according to load on host computers dynamic regulation
Voltage, data center's monitor supervision platform is responsible for monitoring each main frame and virtual machine state and its data is reached into Load Balancing Manager,
Available virtual machine resource is determined with reference to waiting task amount to calculate, task dispatcher obtains available virtual machine resource and waiting list
Middle mission bit stream, obtains performance-energy consumption optimal case using ant group algorithm is improved, finally task is allocated.
The specific method flow of the present invention, as shown in Fig. 2 specifically including following steps:
Step 1:DVS integrated management devices obtain each physical host and its virtual machine running state information, and calculating main frame is born
Carry, main frame is regulated and controled according to presetting threshold xi.Main frame i load Load (i) is determined according to following formula:
In formula, RunVM (i) represents the virtual machine quantity being currently running on main frame i, and MaxVM (i) represents to hold on main frame i
The maximum virtual machine quantity of load, sets ξ value as 0.5, and when load on host computers is more than ξ, DVS control units are carried by regulating and controlling voltage
CPU speed steps are risen, DVS control units reduce CPU speed steps by regulating and controlling voltage when load on host computers is less than ξ.
Step 2:According to virtual machine and waiting list state, Load Balancing Manager selects available m virtual machine to add
Alternate list.
Step 3:M task before task queue is obtained, the financial value of calculating task on a virtual machine is (during prediction tasks carrying
Between and energy consumption of virtual machine), obtain the prediction gain matrix of tasks carrying.The revenue function I that task i is performed on virtual machine j
(xi,j) determined according to following formula:
I(xi,j)=B (xi,j)-P(xi,j)
In formula, B (xi,j) represent performance revenue function, P (xi,j) represent energy consumption overhead functions.
Performance benefits function B (xi,j) according to the following formula calculate obtain:
In formula, λ is a regulation parameter, Time (xi,j) represent that predictions of the task i on virtual machine j performs the time, according to
Following formula is calculated and obtained:
Time(xi,j)=γ (LoadVMj+Ni)
In formula, γ is empirical parameter, is obtained by training, LoadVMjRepresent virtual machine j current load situations, NiRepresent pre-
Increased load is surveyed, its value determines that span is N according to task sizei=0.1~0.4.
Energy consumption overhead functions P (xi,j) according to the following formula calculate obtain:
P(xj)=λ ' (acpuucpu,j+amenNLLC,j+aiobio,j)
In formula, λ ' is a regulation parameter, acpu、amen、aioThe specific constant of model is represented, can be obtained by training,
ucpu,jRepresent virtual machine j processor utilization, NLLC,jRepresent virtual machine j last layer on all cores in certain period of time
Cache missing times, bio,jRepresent virtual machine j read-write total bytes.
Predict that gain matrix form is as shown in table 1.
Table 1 predicts gain matrix
Step 4:Energy saving task scheduler module obtains the virtual machine state data of Load Balancing Manager, is each available void
Plan machine computing node assigns initial information element.Pheromones τs of the virtual machine j in initial timej(0) calculate according to the following formula:
τj(0)=bcpu(numj×Pj)+bmemRj+bnetWj
In formula, bcpu、bmem、bnetCPU information element, memory information element, the plain weight coefficient of bandwidth information, num are represented respectivelyj
Represent the quantity of virtual machine j processors, PjThe processing speed of each processors of virtual machine j is represented, unit is MIPS, RjRepresent empty
Plan machine j internal memory surpluses, unit is M, WjVirtual machine j amount of bandwidth is represented, unit is Mbps.
Step 5:Ant is initialized, ant is randomly placed on available virtual machine.
Step 6:According to gain matrix, calculate ant k and move on to next node probability, according to probability selection next node.Ant
The computing formula that ant k moves on to the probability of next node is:
In formula, Pj k(t) t ant k selection virtual machines j probability is represented, α represents money as information heuristic greedy method
Source information element significance level, β represents the relative importance of control desired value, τ as expected heuristic valuej(t) t is represented
Moment virtual machine j pheromone concentration, ηjThe expected degree that expression task i is performed on virtual machine j, ηj=I (xi,j)。
Step 7:Judge whether ant k completes epicycle search, if so, being updated to local information element, held if it is not, returning
Row step 6.It is to the formula that local information element is updated:
τj(t+1)=(1- ρ) τj(t)+Δτj
In formula, ρ is volatilization factor, meets 0≤ρ<1,1- ρ represents that pheromones remain degree, Δ τjComputing formula is:
Δτj=D1·RES(Ik,l)
In formula, Ik,lRepresent the task allocative decision that kth ant searches in the l times iteration, RES (Ik,l) value according to
This allocative decision revenue function value ∑ I (xi,j) determine, D1It is constant.
Step 8:Judge whether all ants complete this search, if so, carrying out global letter to this search optimal path
Breath element is updated, and step 6 is performed if it is not, returning.Global information element is updated, Δ τjComputing formula is:
Δτj=D2·MAX(RES(Ik,l))
In formula, MAX (RES (Ik,l)) represent the optimal distributing scheme that k Ant Search is arrived in the l times iteration, D2It is normal
Amount.
Step 9:Judge whether to meet ant termination condition, if so, this group task optimal distributing scheme is exported, by task point
It is assigned to correspondence virtual machine to perform, step 5 is performed if it is not, returning;
Step 10:Judge whether also have task to be allocated in task queue, step 1 is performed if so, returning, if nothing, terminate
This subtask is distributed.
The present invention devises a kind of new cloud data center comprehensive energy-saving method, and DVS technologies are appointed with optimization ant group algorithm
Business dispatching technique is combined, in terms of physical regulating and task scheduling algorithm two, and integrated power-saving management is carried out to cloud data center,
Low energy consumption target is reached while proof load balance, best performance., can be according to loading condition using DVS integrated management devices
Dynamic regulation server voltage;Using Load Balancing Manager, alternative virtual machine can be selected before task scheduling, for waiting to adjust
Task choosing is spent, the purpose of load balancing can be so reached;, will using improved ant colony dispatching algorithm in task dispatcher
Performance, Energy Consumption Factors incorporate dispatching algorithm, not only shorten task scheduling and execution time, and reduce energy consumption.
It will be appreciated by those skilled in the art that to the cloud number based on DVS with improvement ant group algorithm disclosed in foregoing invention
According to center energy-saving method, various improvement can also be made on the basis of present invention is not departed from.Therefore, protection of the invention
Scope should be determined by the content of appending claims.
Claims (11)
1. based on DVS and the cloud data center power-economizing method for improving ant group algorithm, DVS integrated managements device obtains physical host and void
Plan machine status information, dynamic regulation physical host voltage reaches energy-conservation purpose, and virtual machine state information is reached based on ant
The energy saving task scheduler module of group's algorithm, is that each task distributes suitable virtual machine, the dual regulation and control of physical host, virtual machine,
Reach that performance is optimal with energy consumption, it is characterised in that specifically include following steps:
Step 1:DVS integrated management devices obtain each physical host and its virtual machine running state information, calculating main frame load, root
Main frame is regulated and controled according to presetting threshold xi;
Step 2:According to virtual machine and waiting list state, it is alternative that Load Balancing Manager selects available m virtual machine to add
List;
Step 3:Obtain task queue before m task, calculating task on a virtual machine financial value (prediction task execution time and
Energy consumption of virtual machine), obtain the prediction gain matrix of tasks carrying;
Step 4:Energy saving task scheduler module obtains the virtual machine state data of DVS managers, is each available virtual machine computing
Node assigns initial information element;
Step 5:Ant is initialized, ant is randomly placed on available virtual machine;
Step 6:According to gain matrix, calculate ant k and move on to next node probability, according to probability selection next node;
Step 7:Judge whether ant k completes epicycle search, if so, being updated to local information element, step is performed if it is not, returning
Rapid 6;
Step 8:Judge whether all ants complete this search, if so, carrying out global information element to this search optimal path
Update, step 6 is performed if it is not, returning;
Step 9:Judge whether to meet ant termination condition, if so, export this group task optimal distributing scheme, by task distribute to
Correspondence virtual machine is performed, and step 5 is performed if it is not, returning;
Step 10:Judge whether also have task to be allocated in task queue, perform step 1 if so, returning, if nothing, terminate this
Task is distributed.
2. the cloud data center power-economizing method according to claim 1 based on DVS with improvement ant group algorithm, its feature exists
In in the step 1, DVS integrated management devices are continuously evaluated every physical host task load, main frame i load Load (i) roots
Determined according to following formula:
In formula, RunVM (i) represents the virtual machine quantity being currently running on main frame i, and MaxVM (i) represents what can be carried on main frame i
Maximum virtual machine quantity, sets ξ value as 0.5, and when load on host computers is more than ξ, DVS control units are by regulating and controlling voltage increase CPU
Speed step, when load on host computers is less than ξ, DVS control units reduce CPU speed steps by regulating and controlling voltage.
3. the cloud data center power-economizing method according to claim 1 based on DVS with improvement ant group algorithm, its feature exists
In, in the step 3, the revenue function I (x that task i is performed on virtual machine ji,j) determined according to following formula:
I(xi,j)=B (xi,j)-P(xi,j)
In formula, B (xi,j) represent performance revenue function, P (xi,j) represent energy consumption overhead functions.
4. the cloud data center power-economizing method according to claim 3 based on DVS with improvement ant group algorithm, its feature exists
In performance benefits function B (xi,j) according to the following formula calculate obtain:
In formula, λ is a regulation parameter, Time (xi,j) represent that predictions of the task i on virtual machine j performs the time.
5. the cloud data center power-economizing method according to claim 4 based on DVS with improvement ant group algorithm, its feature exists
In predictions of the task i on virtual machine j performs time Time (xi,j) according to the following formula calculate obtain:
Time(xi,j)=γ (LoadVMj+Ni)
In formula, γ is empirical parameter, is obtained by training, LoadVMjRepresent virtual machine j current load situations, NiRepresent that prediction increases
Plus load, its value according to task size determine.
6. the cloud data center power-economizing method according to claim 5 based on DVS with improvement ant group algorithm, its feature exists
In prediction increase load NiAccording to task size value, span is Ni=0.1~0.4.
7. the cloud data center power-economizing method according to claim 3 based on DVS with improvement ant group algorithm, its feature exists
In energy consumption overhead functions P (xi,j) according to the following formula calculate obtain:
P(xj)=λ ' (acpuucpu,j+amenNLLC,j+aiobio,j)
In formula, λ ' is a regulation parameter, acpu、amen、aioThe specific constant of model is represented, can be obtained by training, ucpu,j
Represent virtual machine j processor utilization, NLLC,jRepresent virtual machine j last layer of Cache on all cores in certain period of time
Missing times, bio,jRepresent virtual machine j read-write total bytes.
8. the cloud data center power-economizing method according to claim 1 based on DVS with improvement ant group algorithm, its feature exists
In, in the step 4, pheromones τs of the virtual machine j in initial timej(0) calculate according to the following formula:
τj(0)=bcpu(numj×Pj)+bmemRj+bnetWj
In formula, bcpu、bmem、bnetCPU information element, memory information element, the plain weight coefficient of bandwidth information, num are represented respectivelyjRepresent empty
The quantity of plan machine j processors, PjThe processing speed of each processors of virtual machine j is represented, unit is MIPS, RjRepresent in virtual machine j
Surplus is deposited, unit is M, WjVirtual machine j amount of bandwidth is represented, unit is Mbps.
9. the cloud data center power-economizing method according to claim 1 based on DVS with improvement ant group algorithm, its feature exists
In in the step 6, the computing formula that ant k moves on to the probability of next node is:
In formula,T ant k selection virtual machines j probability is represented, α represents resource information as information heuristic greedy method
Plain significance level, β represents the relative importance of control desired value, τ as expected heuristic valuej(t) represent that t is empty
Plan machine j pheromone concentration, ηjThe expected degree that expression task i is performed on virtual machine j, ηj=I (xi,j)。
10. the cloud data center power-economizing method according to claim 1 based on DVS with improvement ant group algorithm, its feature exists
In in the step 7, being to the formula that local information element is updated:
τj(t+1)=(1- ρ) τj(t)+Δτj
In formula, ρ is volatilization factor, meets 0≤ρ<1,1- ρ represents that pheromones remain degree, Δ τjComputing formula is:
Δτj=D1·RES(Ik,l)
In formula, Ik,lRepresent the task allocative decision that kth ant searches in the l times iteration, RES (Ik,l) value is according to this
Allocative decision revenue function value ∑ I (xi,j) determine, D1It is constant.
11. the cloud data center power-economizing method according to claim 1 based on DVS with improvement ant group algorithm, its feature exists
In in the step 8, to global information element renewal, Δ τjComputing formula is:
Δτj=D2·MAX(RES(Ik,l))
In formula, MAX (RES (Ik,l)) represent the optimal distributing scheme that k Ant Search is arrived in the l times iteration, D2It is constant.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710191575.XA CN106951059A (en) | 2017-03-28 | 2017-03-28 | Based on DVS and the cloud data center power-economizing method for improving ant group algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710191575.XA CN106951059A (en) | 2017-03-28 | 2017-03-28 | Based on DVS and the cloud data center power-economizing method for improving ant group algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106951059A true CN106951059A (en) | 2017-07-14 |
Family
ID=59473984
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710191575.XA Pending CN106951059A (en) | 2017-03-28 | 2017-03-28 | Based on DVS and the cloud data center power-economizing method for improving ant group algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106951059A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107729070A (en) * | 2017-10-16 | 2018-02-23 | 燕山大学 | Scheduling virtual machine system and method based on dual rate and work and sleep |
CN108845886A (en) * | 2018-07-27 | 2018-11-20 | 中南大学 | Cloud computing energy consumption optimization method and system based on phase space |
CN109101316A (en) * | 2018-07-11 | 2018-12-28 | 南京邮电大学 | A kind of relevant cloud computing platform resource regulating method of efficiency |
CN109617939A (en) * | 2018-10-15 | 2019-04-12 | 西安理工大学 | A kind of WebIDE Cloud Server resource allocation methods of task based access control pre-scheduling |
CN110012044A (en) * | 2018-01-04 | 2019-07-12 | 财团法人工业技术研究院 | Dynamic duty transfer method and server |
CN110764912A (en) * | 2019-10-25 | 2020-02-07 | 东北大学 | Self-adaptive task scheduler and method |
CN110989824A (en) * | 2019-11-15 | 2020-04-10 | 北京浪潮数据技术有限公司 | Energy consumption management method, device, equipment and medium for cloud platform |
CN111245717A (en) * | 2018-11-28 | 2020-06-05 | 中国移动通信集团浙江有限公司 | Cloud service route distribution method and device |
CN111324444A (en) * | 2020-03-24 | 2020-06-23 | 广东工业大学 | Cloud computing task scheduling method and device |
CN111431961A (en) * | 2020-02-20 | 2020-07-17 | 重庆邮电大学 | Energy-saving task allocation method for cloud data center |
CN111722596A (en) * | 2019-03-20 | 2020-09-29 | 北京京东尚科信息技术有限公司 | Multistage scheduling method and device |
CN112148482A (en) * | 2020-09-11 | 2020-12-29 | 电子科技大学 | Edge computing task scheduling method based on combination with load balancing |
CN112769910A (en) * | 2020-12-29 | 2021-05-07 | 杭州电子科技大学 | Fog calculation task unloading method based on dynamic voltage regulation technology |
CN113986518A (en) * | 2021-12-28 | 2022-01-28 | 华南理工大学 | Real-time cloud task multi-objective optimization scheduling method based on ant colony algorithm |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102932422A (en) * | 2012-09-29 | 2013-02-13 | 南京邮电大学 | Cloud environment task scheduling method based on improved ant colony algorithm |
CN103970609A (en) * | 2014-04-24 | 2014-08-06 | 南京信息工程大学 | Cloud data center task scheduling method based on improved ant colony algorithm |
CN104317646A (en) * | 2014-10-23 | 2015-01-28 | 西安电子科技大学 | Cloud data central virtual machine scheduling method based on OpenFlow frame |
CN104915557A (en) * | 2015-06-04 | 2015-09-16 | 中山大学 | Cloud task allocation method based on double-objective ant colony algorithm |
-
2017
- 2017-03-28 CN CN201710191575.XA patent/CN106951059A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102932422A (en) * | 2012-09-29 | 2013-02-13 | 南京邮电大学 | Cloud environment task scheduling method based on improved ant colony algorithm |
CN103970609A (en) * | 2014-04-24 | 2014-08-06 | 南京信息工程大学 | Cloud data center task scheduling method based on improved ant colony algorithm |
CN104317646A (en) * | 2014-10-23 | 2015-01-28 | 西安电子科技大学 | Cloud data central virtual machine scheduling method based on OpenFlow frame |
CN104915557A (en) * | 2015-06-04 | 2015-09-16 | 中山大学 | Cloud task allocation method based on double-objective ant colony algorithm |
Non-Patent Citations (2)
Title |
---|
尔雅莉等: "云中基于蚁群算法改进的负载均衡策略", 《计算机工程与设计》 * |
齐鲁欣: ""基于CPU任务分解的DVS技术的研究与实现"", 《万方数据库》 * |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107729070A (en) * | 2017-10-16 | 2018-02-23 | 燕山大学 | Scheduling virtual machine system and method based on dual rate and work and sleep |
CN110012044A (en) * | 2018-01-04 | 2019-07-12 | 财团法人工业技术研究院 | Dynamic duty transfer method and server |
CN110012044B (en) * | 2018-01-04 | 2022-01-14 | 财团法人工业技术研究院 | Dynamic work transfer method and server |
CN109101316A (en) * | 2018-07-11 | 2018-12-28 | 南京邮电大学 | A kind of relevant cloud computing platform resource regulating method of efficiency |
CN109101316B (en) * | 2018-07-11 | 2022-10-04 | 南京邮电大学 | Energy efficiency related cloud computing platform resource scheduling method |
CN108845886A (en) * | 2018-07-27 | 2018-11-20 | 中南大学 | Cloud computing energy consumption optimization method and system based on phase space |
CN108845886B (en) * | 2018-07-27 | 2022-03-08 | 中南大学 | Cloud computing energy consumption optimization method and system based on phase space |
CN109617939A (en) * | 2018-10-15 | 2019-04-12 | 西安理工大学 | A kind of WebIDE Cloud Server resource allocation methods of task based access control pre-scheduling |
CN109617939B (en) * | 2018-10-15 | 2021-10-26 | 西安理工大学 | WebIDE cloud server resource allocation method based on task pre-scheduling |
CN111245717A (en) * | 2018-11-28 | 2020-06-05 | 中国移动通信集团浙江有限公司 | Cloud service route distribution method and device |
CN111722596A (en) * | 2019-03-20 | 2020-09-29 | 北京京东尚科信息技术有限公司 | Multistage scheduling method and device |
CN110764912B (en) * | 2019-10-25 | 2022-09-09 | 东北大学 | Self-adaptive task scheduler and method |
CN110764912A (en) * | 2019-10-25 | 2020-02-07 | 东北大学 | Self-adaptive task scheduler and method |
CN110989824A (en) * | 2019-11-15 | 2020-04-10 | 北京浪潮数据技术有限公司 | Energy consumption management method, device, equipment and medium for cloud platform |
CN110989824B (en) * | 2019-11-15 | 2021-08-10 | 北京浪潮数据技术有限公司 | Energy consumption management method, device, equipment and medium for cloud platform |
CN111431961B (en) * | 2020-02-20 | 2022-05-03 | 重庆邮电大学 | Energy-saving task allocation method for cloud data center |
CN111431961A (en) * | 2020-02-20 | 2020-07-17 | 重庆邮电大学 | Energy-saving task allocation method for cloud data center |
CN111324444A (en) * | 2020-03-24 | 2020-06-23 | 广东工业大学 | Cloud computing task scheduling method and device |
CN111324444B (en) * | 2020-03-24 | 2023-05-09 | 广东工业大学 | Cloud computing task scheduling method and device |
CN112148482A (en) * | 2020-09-11 | 2020-12-29 | 电子科技大学 | Edge computing task scheduling method based on combination with load balancing |
CN112148482B (en) * | 2020-09-11 | 2023-08-22 | 电子科技大学 | Edge computing task scheduling method based on combination of load balancing |
CN112769910A (en) * | 2020-12-29 | 2021-05-07 | 杭州电子科技大学 | Fog calculation task unloading method based on dynamic voltage regulation technology |
CN113986518A (en) * | 2021-12-28 | 2022-01-28 | 华南理工大学 | Real-time cloud task multi-objective optimization scheduling method based on ant colony algorithm |
CN113986518B (en) * | 2021-12-28 | 2022-05-24 | 华南理工大学 | Real-time cloud task multi-objective optimization scheduling method based on ant colony algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106951059A (en) | Based on DVS and the cloud data center power-economizing method for improving ant group algorithm | |
Yi et al. | Task optimization and scheduling of distributed cyber–physical system based on improved ant colony algorithm | |
Zhang et al. | Energy-aware virtual machine allocation for cloud with resource reservation | |
Sathya Sofia et al. | Multi-objective task scheduling to minimize energy consumption and makespan of cloud computing using NSGA-II | |
Gao et al. | Quality of service aware power management for virtualized data centers | |
Wen et al. | Energy and cost aware scheduling with batch processing for instance-intensive IoT workflows in clouds | |
Sampaio et al. | Towards high-available and energy-efficient virtual computing environments in the cloud | |
Sahoo et al. | A learning automata-based scheduling for deadline sensitive task in the cloud | |
CN102662750A (en) | Virtual machine resource optimal control method and control system based on elastic virtual machine pool | |
CN107168770B (en) | Low-energy-consumption cloud data center workflow scheduling and resource supply method | |
Xu et al. | Resource pre-allocation algorithms for low-energy task scheduling of cloud computing | |
Liang et al. | Memory-aware resource management algorithm for low-energy cloud data centers | |
CN109491760A (en) | A kind of high-effect data center's Cloud Server resource autonomous management method and system | |
Fernández-Cerero et al. | Energy policies for data-center monolithic schedulers | |
Li et al. | Cost-aware automatic scaling and workload-aware replica management for edge-cloud environment | |
Feng et al. | Towards heat-recirculation-aware virtual machine placement in data centers | |
Hussin et al. | Efficient energy management using adaptive reinforcement learning-based scheduling in large-scale distributed systems | |
Xu et al. | Energy-Aware Resource Management in Cloud Computing Considering Load Balance. | |
Zhang et al. | GreenDRL: managing green datacenters using deep reinforcement learning | |
Chen et al. | An energy‐efficient method of resource allocation based on request prediction in multiple cloud data centers | |
Chen et al. | Power and thermal-aware virtual machine scheduling optimization in cloud data center | |
Ghoreyshi | Energy-efficient resource management of cloud datacenters under fault tolerance constraints | |
Jakóbik et al. | Energy efficient scheduling methods for computational grids and clouds | |
Leite et al. | Power‐aware server consolidation for federated clouds | |
Tong et al. | Energy and performance-efficient dynamic consolidate VMs using deep-Q neural network |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170714 |