CN109800066A - A kind of data center's energy-saving scheduling method and system - Google Patents
A kind of data center's energy-saving scheduling method and system Download PDFInfo
- Publication number
- CN109800066A CN109800066A CN201811524080.5A CN201811524080A CN109800066A CN 109800066 A CN109800066 A CN 109800066A CN 201811524080 A CN201811524080 A CN 201811524080A CN 109800066 A CN109800066 A CN 109800066A
- Authority
- CN
- China
- Prior art keywords
- data center
- server
- parameter
- power consumption
- energy
- 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.)
- Granted
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
- Air Conditioning Control Device (AREA)
Abstract
The present invention provides a kind of data center's energy-saving scheduling method and system, by obtaining the resource utilization of Servers-all in current data center, resource requirement and environment nowadays parameter in task queue to scheduler task, predict that parameter is arranged in data center's total power consumption after scheduler task dispose on any one server and air conditioner in machine room using preset prediction model；It is determined for compliance with data center's energy-saving distribution scheme of preset condition according to prediction result, and data center's total energy consumption is scheduled according to the program.The present invention passes through the method combined dispatching server system of machine learning and the energy consumption of computer-room air conditioning system, it solves the problems, such as to exist in the related technology to have achieved the effect that cross-layer uniformly optimizes data center's total energy consumption for the optimization of single level, using inaccurate energy consumption model, low based on energy-saving efficiency caused by CFD emulation dispatch.
Description
Technical field
The present invention relates to data center field, specifically a kind of data realized data center's cross-layer and unify energy optimization
Center energy-saving dispatching method and system.
Background technique
Data center with the fast development of cloud computing technology, as the physical platform of cloud computing, in global range
Unprecedented development is obtained.And the data center's number increased rapidly also brings huge energy consumption expense to operator.
For example, data center's power consumption total amount in the U.S. in 2014 has already taken up the 1.8% of the whole America whole year total power consumption, and the number
Value also is keeping increasing year by year.In order to reduce the consumption of data center's electric energy, academia and industry propose a series of energy conservations
Scheme.Since server system and computer-room air conditioning system occupy the power consumption of data center about 80% or more in total, greatly
Part research is carried out both for the two systems.However, being directed to the energy conservation object of server system and being directed to air-conditioning system
Energy conservation object between there are trade-off relationships.For example, the energy-saving scheme for being directed to server system can polymerize IT and load to a small number of clothes
It is engaged on device, reaches energy saving purpose to close more servers as far as possible.However, due to the rotation speed of the fan and energy consumption of air-conditioning system
Superlinearity functional relation, for air-conditioning system energy-saving scheme can be shared equally between Servers-all as far as possible IT load.Therefore,
The one-sided energy consumption for saving any one system can not achieve the purpose that save data center's overall energy consumption.
Currently, being broadly divided into three classes for data center server and the unified energy-saving scheme of air-conditioning system cross-layer.First
Class assume that two kinds of systems IT load between there are specific mathematical function relationships, it is assumed herein that on the basis of design scheduling calculate
Method carries out overall energy conservation.However, due to there is complicated interaction between the various parameters of influence consumption of data center
And feedback control loop, it is to be grossly inaccurate to consumption of data center modeling with the method for Traditional project formula, therefore such scheme
It is ineffective in practice.Second class is that temperature sensor is arranged on the server, and the temperature number come is sent back according to sensor
It is configured according to cooling system parameter, such method is unable to prediction data center overall energy consumption, it is difficult to energy-saving distribution be instructed to calculate
Method optimization.Last one kind is the heating power distribution map that the method based on numerical analysis model simulates data center, such method meter
Calculate that expense is excessive, and can not accurate response data center heat dissipation complexity, therefore, it is impossible to realize by IT load into
Row Real-Time Scheduling and system parameter setting achieve the purpose that minimize data center's total energy consumption.
Summary of the invention
The present invention provides a kind of data center's energy-saving scheduling method and systems, pass through the method combined dispatching of machine learning
The energy consumption of server system and computer-room air conditioning system solves in the related technology and exists for the optimization of single level, using inaccurate
True energy consumption model, based on the low problem of energy-saving efficiency caused by CFD emulation dispatch.
According to an aspect of the invention, there is provided a kind of data center's energy-saving scheduling method, which includes containing
There are server system, computer-room air conditioning system, task queue and the external environment condition parameter monitoring system of at least one server, the party
Method the following steps are included:
Obtain the Current resource utilization rate of Servers-all in data center, the resource in task queue to scheduler task needs
It asks and environment nowadays parameter；
According to obtain Servers-all Current resource utilization rate and environment nowadays parameter, by preset one by
For neural network based on the prediction model that machine learning method generates come prediction result, which includes in office to scheduler task
Anticipate corresponding data center's total power consumption and corresponding air conditioner in machine room setting parameter, the data center after disposing on a server is total
Power consumption includes server system total power consumption and computer-room air conditioning system total power consumption；
According to above-mentioned prediction result, it is determined for compliance with data center's energy-saving distribution scheme of preset condition, that is, traverses all energy
Enough servers accommodated to scheduler task, select the server that deployment meets preset condition after scheduler task, and labeled as to
Dispatch server according to prediction model predicts that the corresponding data center after dispatch server will be deployed in after scheduler task total
Parameter is arranged in power consumption and corresponding air conditioner in machine room；
Data center's total energy consumption is scheduled according to above-mentioned energy-saving distribution scheme.
Optionally, prediction model is with the resource utilization of Servers-all in data center and corresponding external environment condition parameter
The input as neural network of history or experimental data, to meet one group of air conditioner in machine room of server cooling requirement, parameter is set
History or experimental data with data center total power consumption are obtained as output by means of the Nonlinear Processing ability training of neural network
It arrives.Wherein, the total resources utilization rate reason as input of not all servers using the resource utilization of Servers-all,
It is that the opposite location distribution of server and computer-room air conditioning system will largely influence data center's total energy consumption.
Optionally, preset condition includes: so that data center's total power consumption is minimum.
Optionally, being scheduled to data center's total energy consumption includes: that will be deployed to scheduler task on dispatch server,
The working condition of server is adjusted to realize to the scheduling of the power consumption of server system, adjustment air conditioner in machine room setting parameter is to realize pair
The power consumption of computer-room air conditioning system is dispatched.
Optionally, the power consumption of computer-room air conditioning system is dispatched further include: detect whether current air-conditioning setting reaches server
Cooling requirement, if not up to, fine tuning air-conditioning parameter is to reaching server cooling requirement.
According to another aspect of the present invention, a kind of data center's energy-conserving scheduling system is provided, which includes
Server system, computer-room air conditioning system, task queue and external environment condition parameter monitoring system containing at least one server, should
System includes:
Acquisition device, including server resource utilization rate acquisition device, to scheduler task resource requirement acquisition device, outside
Environmental parameter acquisition device, computer-room air conditioning system parameter obtaining device and data center's total power consumption acquisition device, are each responsible for
Obtain the resource utilization of Servers-all in data center, the resource requirement in task queue to scheduler task, external environment
Parameter, air conditioner in machine room setting parameter and data center's total power consumption；
Prediction meanss are responsible for the data obtained by above-mentioned acquisition device, according to the method training nerve net of machine learning
Network, generate prediction model, to export disposed on any one server after scheduler task after corresponding data center's total work
Parameter is arranged in consumption and corresponding air conditioner in machine room；
Energy-saving distribution schemes generation device is responsible for the prediction model generated according to above-mentioned prediction meanss and above-mentioned acquisition dress
Set the Current resource utilization rate of the server of acquisition, resource requirement and external environment condition parameter in task queue to scheduler task
Generate the data center's energy-saving distribution scheme for meeting preset condition；
Device is set, and the energy-saving distribution plan implementation for being responsible for generating in above-mentioned energy-saving distribution schemes generation device is into data
The heart.
Optionally, data center's total power consumption acquisition device includes server energy consumption acquisition device and computer-room air conditioning system power consumption
Acquisition device is each responsible for obtaining server system total power consumption and Air Conditioning Facilities total system power consumption.
Optionally, prediction meanss include a training device, are responsible for the utilization of resources with Servers-all in data center
Input of the history or experimental data of rate and corresponding external environment condition parameter as neural network, is arranged with corresponding air conditioner in machine room and is joined
Several and data center's total power consumption history or experimental data are as output, by means of the Nonlinear Processing ability training of neural network
Obtain prediction model.
Optionally, energy-saving distribution schemes generation device includes:
Server determining device is responsible for all servers that can be accommodated to scheduler task of traversal, selects and dispose wait dispatch
Meet the server of preset condition after task, and labeled as to dispatch server；
Air-conditioning determining device, it is right after dispatch server to predict to be deployed in after scheduler task according to prediction model to be responsible for
Parameter is arranged in the data center's total power consumption answered and corresponding air conditioner in machine room.
Optionally, meeting preset condition includes: so that data center's total power consumption is minimum.
Optionally, setting device includes:
Device is disposed to scheduler task, is responsible for be deployed to scheduler task on dispatch server；
Device is arranged in server contention states, is responsible for the working condition of adjustment server to realize the function to server system
Consumption scheduling；
Computer-room air conditioning system parameter setting apparatus is responsible for adjustment air conditioner in machine room setting parameter to realize to computer-room air conditioning system
Power consumption scheduling.
Optionally, computer-room air conditioning system parameter setting apparatus further include:
Detection device, is responsible for whether the current air-conditioning setting of detection reaches server cooling requirement；
Micromatic setting, if detection device testing result be it is below standard, micromatic setting be responsible for finely tune air-conditioning parameter to reach clothes
Business device cooling requirement.
Through the invention, using obtain the resource utilization of Servers-all in current data center, in task queue to
The resource requirement and environment nowadays parameter of scheduler task；According to server resource utilization rate and the current environment ginseng obtained
Number, according to the preset prediction model generated based on machine learning method, prediction is to scheduler task on any one server
Parameter is arranged in corresponding data center's total power consumption and air conditioner in machine room after deployment；Preset condition is determined for compliance with according to prediction result
Data center's energy-saving distribution scheme, and data center's total energy consumption is scheduled according to the program, it solves in the related technology
It is asked in the presence of for the optimization of single level, using inaccurate energy consumption model, based on energy-saving efficiency caused by CFD emulation dispatch is low
Topic, has achieved the effect that cross-layer uniformly optimizes data center's total energy consumption.
Detailed description of the invention
Fig. 1 is the flow chart of data center's energy-saving scheduling method according to an embodiment of the present invention；
Fig. 2 is the structural block diagram of data center's energy-conserving scheduling system according to an embodiment of the present invention；
Fig. 3 is the structural block diagram of data center's energy-conserving scheduling system acquisition device 201 according to an embodiment of the present invention；
Fig. 4 is the preferred structure frame of data center's energy-conserving scheduling system prediction meanss 202 according to an embodiment of the present invention
Figure；
Fig. 5 is data center's energy-conserving scheduling system energy-saving distribution schemes generation device 203 according to an embodiment of the present invention
Preferred structure frame diagram；
Fig. 6 is the preferred structure frame of data center's energy-conserving scheduling system setting device 204 according to an embodiment of the present invention
Figure.
Specific embodiment
To be further understood to technical solution of the present invention, it is described in detail hereinafter with reference to attached drawing and in conjunction with the embodiments.
A kind of data center's energy-saving scheduling method is provided in the present embodiment, and data center includes at least one or several
Platform server, computer-room air conditioning system, task queue and external environment condition parameter monitoring system, Fig. 1 are according to an embodiment of the present invention
The flow chart of data center's energy-saving scheduling method, the process include the following steps:
Step 101, obtain the resource utilization of Servers-all in current data center, in task queue to scheduler task
Resource requirement and environment nowadays parameter.As known to persons skilled in the art, resource refers mainly to CPU, can also be with
Include memory, network bandwidth, hard disk, IO.Since the pressure and temperature of external environment is to the energy consumption shadow of data center air-conditioner system
Sound is larger, and therefore, external environment condition parameter mainly includes pressure and temperature；
Step 102, according to the server resource utilization rate and current environment parameter obtained, according to preset prediction model,
Prediction corresponding data center's total power consumption and corresponding air conditioner in machine room after scheduler task is disposed on any one server
Parameter is set, wherein prediction model is generated based on machine learning method；
Step 103, according to above-mentioned prediction result, it is determined for compliance with data center's energy-saving distribution scheme of preset condition, and
Data center's total energy consumption is scheduled according to above-mentioned energy-saving scheme.
Through the above steps, by the energy consumption of combined dispatching server system and computer-room air conditioning system, related skill is solved
Data center's whole energy inefficiency is asked caused by energy consumption in art only for the single level of data center optimizes
Topic, in addition, solving the energy consumption model in the related technology based on inaccuracy using the method based on machine learning and carrying out task tune
The invalid and task scheduling strategy based on CFD emulation of Energy Saving Strategy caused by spending is unable to satisfy in data since computing cost is excessive
The problem of heart Real-Time Scheduling demand, further improves the effect of consumption of data center saving.
Preset prediction model can be generated using following methods: obtaining in data center own at a certain time interval
Input of the history or experimental data of the resource utilization of server and corresponding external environment condition parameter as neural network, to meet
One group of air conditioner in machine room setting parameter of server cooling requirement and the history or experimental data of data center's total power consumption are as output
Training obtains.Neural network has powerful Nonlinear Processing ability, is very suitable to complicated non-linear in process data center
Relationship is suitable for data center in addition, the time for needing to spend when being predicted using the model that neural metwork training comes out is short
Real-time online scheduling.As known to persons skilled in the art, air conditioner in machine room parameter mainly includes that (wind speed is determined for temperature and wind speed
Surely the air quantity sent out).The setting of air-conditioning parameter needs to meet the condition of server refrigeration in data center, for example, U.S. heating,
The condition that refrigeration and air-conditioning man's teachers learn (ASHRAE) formulated server air inlet in 2008 is 18-27 DEG C.Therefore, it is keeping
In the case that server resource utilization rate and external environment condition parameter are constant, traversal air-conditioning setting parameter, which can be obtained, a series of meets clothes
The air-conditioning parameter setting of business device cryogenic conditions selects the smallest one group of air-conditioning setting parameter of wherein corresponding data center total power consumption i.e.
It can be used as the output of neural network.
Energy conservation object for server system and for there are trade-off relationships between the energy conservation object of air-conditioning system.Example
Such as, IT can be polymerize for the energy-saving scheme of server system to load on a small number of servers, to close more services as far as possible
Device reaches energy saving purpose.However, due to the rotation speed of the fan of air-conditioning system and the superlinearity functional relation of energy consumption, for air-conditioning system
Energy-saving scheme can be shared equally between Servers-all as far as possible IT load.Therefore, the one-sided energy consumption for saving any one system
It can not achieve the purpose that save data center's overall energy consumption.In order to save the energy consumption of data center on the whole, data in
Heart total power consumption includes two parts of server system total power consumption and computer-room air conditioning system total power consumption.
The data center's energy-saving distribution scheme for being determined for compliance with preset condition can traverse all energy using following processing first
Enough servers accommodated to scheduler task meet the server of preset condition after selecting deployment task, and take labeled as wait dispatch
Business device.It is the remaining resource of the server not less than the resource to scheduler task that server, which can be accommodated to the condition of scheduler task,
Demand.Assuming that there is N platform server, calculating will after scheduler task is deployed in server i (1≤i≤N) server money
Source utilization rate situation, and be input to together with the resource utilization of other Servers-alls and environment nowadays parameter preset
In prediction model, the output of (can also save) model is recorded, the server for meeting preset condition in N number of output is selected, marked
For I；Then, it predicts the task deployment according to preset prediction model in data center corresponding after dispatch server total work
Consumption and corresponding air conditioner in machine room be arranged parameter, if above-mentioned determination to dispatch server during saved task
It is deployed in corresponding output in server I, then can directly transfer record obtains corresponding data center's total power consumption and right
The air conditioner in machine room setting parameter answered.
Meeting preset condition includes: so that data center's total power consumption is minimum.
Data center's total energy consumption is scheduled including three aspects according to above-mentioned energy-saving scheme: by task deployment to above-mentioned
It is determining on dispatch server；Since the task in data center dynamically reaches and leaves at any time, in deployment new task
When it is possible that part server utilization rate reduce even part server because handle its carry all tasks and
The phenomenon that being in idle condition, at this point, the server of idle state will waste a large amount of energy consumptions, therefore, it is necessary to adjust server sheet
The working condition of body is to realize the power consumption scheduling to server system, for example, by low-power consumption is switched in idle server
Suspend mode (being even switched off), reduce the operating voltage and frequency of the low server of utilization rate；Adjust computer-room air conditioning system ginseng
Number setting group is to realize the power consumption scheduling to computer-room air conditioning system.Due to the variation of server load level and working condition, need
Air Conditioning Facilities system is adjusted accordingly to adapt to new server refrigeration demand, this is because if server system cooling capacity
Demand is reduced, and air-conditioning system needs corresponding reduction semen donors to reduce consumption of data center, and if server system cooling capacity needs
Increase is asked, air-conditioning system needs to increase accordingly semen donors to meet server refrigeration demand.
Due to prediction model prediction computer-room air conditioning system parameter setting there are it is certain a possibility that be not able to satisfy current clothes
Therefore the refrigeration demand of business device also needs to detect current air-conditioning parameter group after adjusting computer-room air conditioning system parameter setting group and sets
It sets and whether has reached server cooling requirement, as known to a person skilled in the art, detection machine cabinet top upper inlet can be passed through
Whether temperature meets regulation (for example, between 18-27 DEG C) to determine whether up to standard；If not up to, can be by finely tuning air-conditioning
Parameter is to reaching server cooling requirement, for example, illustrate air-conditioning semen donors deficiency at this time if temperature is higher than 27 DEG C, it can be gradually
Air-conditioner temperature and wind speed are improved until up to standard；If temperature is lower than 18 DEG C, illustrate that air-conditioning semen donors are excessive at this time, can gradually reduce
Air-conditioner temperature and wind speed are until up to standard.
Provide a kind of data center's energy-conserving scheduling system in the present embodiment, the system for realizing above-described embodiment,
Therefore, the explanation carried out will not be described in great detail herein.Following all devices can be by hardware or software or hardware
With being implemented in combination with for software.Fig. 2 is the structural block diagram of data center's energy-conserving scheduling system according to an embodiment of the present invention, such as Fig. 2
Shown, which includes acquisition device 201, prediction meanss 202, energy-saving distribution schemes generation device 203 and setting device
204, in which:
Acquisition device 201, be responsible for obtaining the resource utilization of Servers-all in data center, in task queue wait dispatch
Resource requirement, external environment condition parameter, air conditioner in machine room parameter and the data center's total power consumption of task, Fig. 3 are acquisition device 201
Structural block diagram, as shown in figure 3, acquisition device 201 includes server resource utilization rate acquisition device 205, computer-room air conditioning system ginseng
Count acquisition device 206, to scheduler task resource requirement acquisition device 207, external environment condition parameter acquisition device 208, data center
Total power consumption acquisition device 209；
Prediction meanss 202 are responsible for resource utilization, the task of Servers-all in the data center obtained by acquisition device
To the resource requirement of scheduler task, external environment condition parameter, air conditioner in machine room parameter setting and data center's total power consumption in queue, press
According to the method for machine learning, generate can be by appointing in the resource utilization of Servers-all in data center, task queue wait dispatch
The resource requirement of business and external environment condition parameter predict the prediction model of corresponding air conditioner in machine room parameter, data center's total power consumption；
Energy-saving distribution schemes generation device 203, the prediction model and acquisition device for being responsible for generating according to prediction meanss obtain
Resource requirement and external environment condition parameter in the resource utilization of the current server taken, task queue to scheduler task generate
Meet data center's energy-saving distribution scheme of preset condition；
Device 204 is set, the data center's energy-saving distribution plan implementation for generating energy-saving distribution schemes generation device is responsible for
To data center.
Preferably, data center's total power consumption acquisition device 209 includes server system power consumption acquisition device and air conditioner in machine room
System power dissipation acquisition device.
Fig. 4 is the preferred structure frame of prediction meanss 202 in data center's energy-conserving scheduling system according to embodiments of the present invention
Figure, as shown in figure 4, prediction meanss 202 can by built-in training device 210 generate prediction model, training device 210 be responsible for
The history or experimental data of the resource utilization of Servers-all and corresponding external environment condition parameter are as nerve net in data center
The input of network, the history or experimental data that parameter and data center's total power consumption is arranged using corresponding air conditioner in machine room are borrowed as output
Help the powerful Nonlinear Processing ability training of neural network and obtains prediction model.As known to a person skilled in the art, mind
It can be by software realization through network.
Fig. 5 is energy-saving distribution schemes generation device 203 in data center's energy-conserving scheduling system according to embodiments of the present invention
Preferred structure frame diagram, as shown in figure 5, the device 203 includes server determining device 211 and air-conditioning determining device 212, difference
It is described as follows:
Server determining device 211 is responsible for all servers that can be accommodated to scheduler task of traversal, selects deployment task
Meet the server of preset condition afterwards, and labeled as to dispatch server；
Air-conditioning determining device 212 is responsible for predicting the task deployment according to preset prediction model to dispatch server
Parameter is arranged in corresponding data center's total power consumption and corresponding air conditioner in machine room afterwards.
Fig. 6 is the preferred structure frame that device 204 is arranged in data center's energy-conserving scheduling system according to embodiments of the present invention
Figure, as shown in fig. 6, the device 204 include to scheduler task deployment device 213, server contention states setting device 214 and
Computer-room air conditioning system parameter setting apparatus 215, is respectively described below:
Device 213 is disposed to scheduler task, is responsible for task deployment on to dispatch server；
Device 214 is arranged in server contention states, is responsible for the working condition of adjustment server to realize to server system
Power consumption scheduling；
Computer-room air conditioning system parameter setting apparatus 215 is responsible for adjustment computer-room air conditioning system parameter setting group to realize to machine
The power consumption of room air-conditioning system is dispatched.
Preferably, computer-room air conditioning system parameter setting apparatus 215 further include:
Detection device, is responsible for whether the current air-conditioning setting of detection reaches server cooling requirement；
Micromatic setting, if detection device testing result be it is below standard, micromatic setting be responsible for finely tune air-conditioning parameter to reach clothes
Business device cooling requirement.
In recent years, data center constantly develops towards standardization, large-scale direction.Many large size cloud computation data centers
Interior proper alignment a large amount of standardized cabinet is several standardized service devices in cabinet.Based on the relevant technologies, at present in data
The mode that the cabinet of the heart generallys use hot and cold channel spacing arranges, i.e., cold air out of aperture on two rows of rack room floors up
It blows, is freezed by cabinet sucking for cabinet server, hot gas is discharged from cabinet rear, is inhaled by refrigeration equipment slave roof portion
It walks.Server is in the temperature ability trouble free service specified no more than manufacturer, and therefore, the parameter setting of air conditioner in machine room has to
Servers-all temperature can be guaranteed not above its safe range.Due to the influence of the factors such as air-flow, load distribution, environment,
There are larger differences for the temperature of different server entrance cold air in data center, and a large amount of energy consumptions of data center is caused to be wasted.Shadow
It rings the factor of consumption of data center much and relationship complicated difficult with numerical relation accurately to be portrayed, and the rule of cloud computation data center
Mould usually all causes greatly some energy-saving schemes to be unable to satisfy Real-Time Scheduling requirement because computing cost is excessive very much, these factors are all
Hinder the promotion of data center's energy-saving effect.
In view of the above-mentioned problems, providing a kind of load dispatching method based on energy consumption perception, this method in the present embodiment
The distribution that can be loaded meets the energy conservation object of data center as far as possible, to reduce data center's overall energy consumption.This method packet
Include following steps:
Step 301, by data center's monitor supervision platform obtain current data center in Servers-all resource utilization, to
Scheduler task resource requirement and environment nowadays parameter；
Step 302, according to the server resource utilization rate and current environment parameter obtained, mould is predicted according to preset cabinet
Type, prediction corresponding data center's total power consumption and corresponding air conditioner in machine room after any one interior of equipment cabinet administration after scheduler task
Parameter is set, wherein prediction model is generated based on machine learning method, and corresponding data center total power consumption is minimum after selecting deployment
Cabinet be used as cabinet to be dispatched；
Step 303, it obtains wait dispatch server resource utilization rate all in cabinet, according to preset server prediction mould
Type is predicted after the total power consumption of scheduler task corresponding cabinet after any one server internal administration, wherein prediction model is base
It is generated in machine learning method, corresponds to the smallest server of cabinet total power consumption as to dispatch server after selecting deployment；
Step 304, it will be deployed to scheduler task in dispatch server.
Prediction model in above-mentioned steps 302 and step 303 is generated by machine learning method, since neural network is to multiple
Miscellaneous relationship has powerful processing capacity, is very suitable to the power saving with data center, thus can using neural network come
Carry out model training.As known to a person skilled in the art, neural network is divided into three parts: input layer, hidden layer and defeated
Layer out can use software realization.Cabinet prediction model neural network based can be generated by following method in step 302:
Obtain data center in the resource utilization of all cabinets and the history of external environment condition parameter (mainly temperature and pressure) or
Input of person's experimental data as neural network obtains corresponding data center's total power consumption and meets server cooling requirement
The history or experimental data of air conditioner in machine room setting parameter are trained as the output of neural network, and above-mentioned data can be set
Certain sampling time is acquired.Server prediction model neural network based can be by following mode in step 302
It generates: with certain sampling time interval, obtaining the history or experiment of the resource utilization of Servers-all in given cabinet
Input of the data as neural network, obtain corresponding cabinet power consumption history or experimental data as neural network export into
Row training.Above-mentioned power consumption data can carry out actual measurement by power meter or is collected using simulation software.
In computer room in the case where all cabinets uniform homogeneity, the model of step 303 training can be adapted for all cabinets,
Therefore, it can be saved compared with the method for disposing direct prediction data center total power consumption by load with neural network building a large amount of
Training amount.Obviously, above-described embodiment is also applied for the scene of the inhomogenous homogeneity of cabinet in computer room.Through the foregoing embodiment may be used
To realize the load dispatch based on energy consumption, portrayed not moreover, the above method can solve energy consumption present in existing power-saving technology
The not high problem of data center's totality energy-saving efficiency caused by accurate, in addition, passing through decoupling, the nerve of step 302 and step 303
Network inputs output parameter is substantially reduced, and the time exported by input data by trained neural computing is also big
Width reduces, and can be used to implement energy saving load dispatch in real time.
The above embodiments are merely illustrative of the technical solutions of the present invention rather than is limited, the ordinary skill of this field
Personnel can be with modification or equivalent replacement of the technical solution of the present invention are made, without departing from the spirit and scope of the present invention, this
The protection scope of invention should be subject to claims.
Claims (10)
1. a kind of data center's energy-saving scheduling method, the data center include the server system containing at least one server,
Computer-room air conditioning system, task queue and external environment condition parameter monitoring system, which is characterized in that method includes the following steps:
Obtain the Current resource utilization rate of Servers-all in data center, in task queue to the resource requirement of scheduler task with
And environment nowadays parameter；
According to the Current resource utilization rate and environment nowadays parameter of the Servers-all obtained, by preset one by nerve
For network based on the prediction model that machine learning method generates come prediction result, which includes to scheduler task any one
Parameter, data center's total power consumption is arranged in corresponding data center's total power consumption and corresponding air conditioner in machine room after disposing on platform server
Including server system total power consumption and computer-room air conditioning system total power consumption；
According to above-mentioned prediction result, it is determined for compliance with data center's energy-saving distribution scheme of preset condition, that is, traversing all can hold
The server received to scheduler task selects the server that deployment meets preset condition after scheduler task, and labeled as wait dispatch
Server according to prediction model predicts that corresponding data center's total power consumption after dispatch server will be deployed in after scheduler task
Parameter is set with corresponding air conditioner in machine room；
Data center's total energy consumption is scheduled according to above-mentioned energy-saving distribution scheme.
2. the method as described in claim 1, which is characterized in that prediction model is with the resource of Servers-all in data center
Input of the history or experimental data of utilization rate and corresponding external environment condition parameter as neural network, to meet the cooling need of server
The history or experimental data of the one group of air conditioner in machine room setting parameter and data center's total power consumption asked are as output, by means of nerve net
The Nonlinear Processing ability training of network obtains.
3. the method as described in claim 1, which is characterized in that preset condition includes so that data center's total power consumption is minimum.
4. the method as described in claim 1, which is characterized in that data center's total energy consumption is scheduled include: will be wait dispatch
Task deployment adjusts the working condition of server to realize the power consumption scheduling to server system, adjusts on to dispatch server
Whole air conditioner in machine room setting parameter is to realize the power consumption scheduling to computer-room air conditioning system；
Power consumption scheduling to computer-room air conditioning system further include: detect whether current air-conditioning setting reaches server cooling requirement, if
Not up to, fine tuning air-conditioning parameter is to reaching server cooling requirement.
5. a kind of data center's energy-conserving scheduling system, the data center include the server system containing at least one server,
Computer-room air conditioning system, task queue and external environment condition parameter monitoring system, which is characterized in that the system includes:
Acquisition device, including server resource utilization rate acquisition device, to scheduler task resource requirement acquisition device, external environment
Parameter obtaining device, computer-room air conditioning system parameter obtaining device and data center's total power consumption acquisition device, are each responsible for obtaining
The resource utilization of Servers-all in data center, in task queue to the resource requirement of scheduler task, external environment condition parameter,
Parameter and data center's total power consumption is arranged in air conditioner in machine room；
Prediction meanss are responsible for the data obtained by above-mentioned acquisition device, raw according to the method training neural network of machine learning
At prediction model, to export disposed on any one server after scheduler task after corresponding data center's total power consumption and right
The air conditioner in machine room setting parameter answered；
Energy-saving distribution schemes generation device, the prediction model and above-mentioned acquisition device for being responsible for generating according to above-mentioned prediction meanss obtain
Resource requirement and external environment condition parameter in the Current resource utilization rate of the server taken, task queue to scheduler task generate
Meet data center's energy-saving distribution scheme of preset condition；
Device is set, is responsible for the energy-saving distribution plan implementation for generating above-mentioned energy-saving distribution schemes generation device to data center.
6. system as claimed in claim 5, which is characterized in that data center's total power consumption acquisition device includes that server energy consumption obtains
Device and computer-room air conditioning system power consumption acquisition device are taken, is each responsible for obtaining server system total power consumption and Air Conditioning Facilities system is total
Power consumption.
7. system as claimed in claim 5, which is characterized in that prediction meanss include a training device, are responsible in data
The history or experimental data of the resource utilization of intracardiac Servers-all and corresponding external environment condition parameter are as the defeated of neural network
Enter, the history or experimental data that parameter and data center's total power consumption is arranged using corresponding air conditioner in machine room are as output, by means of mind
Nonlinear Processing ability training through network obtains prediction model.
8. system as claimed in claim 5, which is characterized in that energy-saving distribution schemes generation device includes:
Server determining device is responsible for all servers that can be accommodated to scheduler task of traversal, selects deployment to scheduler task
Meet the server of preset condition afterwards, and labeled as to dispatch server；
Air-conditioning determining device is responsible for predicting according to prediction model being deployed in this after dispatch server after scheduler task corresponding
Data center's total power consumption and corresponding air conditioner in machine room be arranged parameter.
9. system as claimed in claim 5, which is characterized in that device, which is arranged, includes:
Device is disposed to scheduler task, is responsible for be deployed to scheduler task on dispatch server；
Device is arranged in server contention states, is responsible for the working condition of adjustment server to realize the power consumption tune to server system
Degree；
Computer-room air conditioning system parameter setting apparatus is responsible for adjustment air conditioner in machine room setting parameter to realize the function to computer-room air conditioning system
Consumption scheduling.
10. system as claimed in claim 5, which is characterized in that computer-room air conditioning system parameter setting apparatus further include:
Detection device, is responsible for whether the current air-conditioning setting of detection reaches server cooling requirement；
Micromatic setting, if detection device testing result be it is below standard, micromatic setting is responsible for finely tuning air-conditioning parameter to reaching server
Cooling requirement.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811524080.5A CN109800066B (en) | 2018-12-13 | 2018-12-13 | Energy-saving scheduling method and system for data center |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811524080.5A CN109800066B (en) | 2018-12-13 | 2018-12-13 | Energy-saving scheduling method and system for data center |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109800066A true CN109800066A (en) | 2019-05-24 |
| CN109800066B CN109800066B (en) | 2020-11-03 |
Family
ID=66556684
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811524080.5A Active CN109800066B (en) | 2018-12-13 | 2018-12-13 | Energy-saving scheduling method and system for data center |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109800066B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110376896A (en) * | 2019-07-30 | 2019-10-25 | 浙江大学 | It is a kind of that refrigerating method is optimized based on deep learning and the single heat source air-conditioning of fuzzy control |
| CN110781125A (en) * | 2019-09-12 | 2020-02-11 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Management method and system for complete cabinet super-fusion server |
| CN111174375A (en) * | 2019-12-11 | 2020-05-19 | 西安交通大学 | Data center energy consumption minimization-oriented job scheduling and machine room air conditioner regulation and control method |
| CN112070353A (en) * | 2020-08-04 | 2020-12-11 | 中国科学院信息工程研究所 | Method and system for accurately detecting energy efficiency of data center |
| CN112308734A (en) * | 2020-10-27 | 2021-02-02 | 中国科学院信息工程研究所 | Method for metering and allocating non-IT energy consumption of IT equipment and electronic device |
| CN112612305A (en) * | 2020-12-04 | 2021-04-06 | 格力电器（武汉）有限公司 | Temperature adjusting method, device, equipment, storage medium and air conditioning system |
| CN112888268A (en) * | 2021-02-04 | 2021-06-01 | 中国工商银行股份有限公司 | Energy-saving control method, device and equipment for data center machine room and storage medium |
| CN113515150A (en) * | 2021-04-12 | 2021-10-19 | 天津大学 | Variable-temperature control method for realizing instant photovoltaic power generation consumption of data center |
| CN113826078A (en) * | 2019-09-11 | 2021-12-21 | 阿里巴巴集团控股有限公司 | Resource scheduling and information prediction method, device, system and storage medium |
| CN113835460A (en) * | 2021-11-23 | 2021-12-24 | 中铁建设集团有限公司 | Data computer lab intelligence environmental control system based on hydrodynamics analog digital twin |
| CN114063545A (en) * | 2022-01-14 | 2022-02-18 | 宁波亮控信息科技有限公司 | Data center energy consumption control system and method fusing edge calculation and controller |
| CN114595851A (en) * | 2022-02-20 | 2022-06-07 | 特斯联科技集团有限公司 | Air conditioner room power consumption analysis device using neural network |
| CN115113705A (en) * | 2021-03-18 | 2022-09-27 | 百度（美国）有限责任公司 | Data center control hierarchy for neural network integration |
| WO2022199608A1 (en) * | 2021-03-25 | 2022-09-29 | 华为技术有限公司 | Energy saving method, apparatus and system applied to machine room |
| CN115907138A (en) * | 2022-11-18 | 2023-04-04 | 安华数据(东莞)有限公司 | Method, system and medium for predicting PUE value of data center |
| CN116193819A (en) * | 2023-01-19 | 2023-05-30 | 中国长江三峡集团有限公司 | Energy-saving control method, system and device for data center machine room and electronic equipment |
| CN116527416A (en) * | 2023-07-03 | 2023-08-01 | 深圳市立湾科技有限公司 | Intelligent AI energy-saving control system and method applied to data center |
| CN117421131A (en) * | 2023-12-18 | 2024-01-19 | 武汉泽塔云科技股份有限公司 | Intelligent scheduling method and system for monitoring power consumption load of server |
| CN117472167A (en) * | 2023-12-28 | 2024-01-30 | 苏州元脑智能科技有限公司 | Method and device for adjusting energy consumption of server, computer equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102445980A (en) * | 2011-09-19 | 2012-05-09 | 浪潮电子信息产业股份有限公司 | Energy-saving control system based on back-propagation (BP) neural network |
| CN102620378A (en) * | 2011-01-27 | 2012-08-01 | 国际商业机器公司 | Method and system for data center energy saving controlling |
| US8457938B2 (en) * | 2007-12-05 | 2013-06-04 | International Business Machines Corporation | Apparatus and method for simulating one or more operational characteristics of an electronics rack |
| CN104317654A (en) * | 2014-10-09 | 2015-01-28 | 南京大学镇江高新技术研究院 | Data center task scheduling method based on dynamic temperature prediction model |
| CN104423531A (en) * | 2013-09-05 | 2015-03-18 | 中兴通讯股份有限公司 | Data center energy consumption scheduling method and data center energy consumption scheduling device |
-
2018
- 2018-12-13 CN CN201811524080.5A patent/CN109800066B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8457938B2 (en) * | 2007-12-05 | 2013-06-04 | International Business Machines Corporation | Apparatus and method for simulating one or more operational characteristics of an electronics rack |
| CN102620378A (en) * | 2011-01-27 | 2012-08-01 | 国际商业机器公司 | Method and system for data center energy saving controlling |
| CN102445980A (en) * | 2011-09-19 | 2012-05-09 | 浪潮电子信息产业股份有限公司 | Energy-saving control system based on back-propagation (BP) neural network |
| CN104423531A (en) * | 2013-09-05 | 2015-03-18 | 中兴通讯股份有限公司 | Data center energy consumption scheduling method and data center energy consumption scheduling device |
| CN104317654A (en) * | 2014-10-09 | 2015-01-28 | 南京大学镇江高新技术研究院 | Data center task scheduling method based on dynamic temperature prediction model |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110376896A (en) * | 2019-07-30 | 2019-10-25 | 浙江大学 | It is a kind of that refrigerating method is optimized based on deep learning and the single heat source air-conditioning of fuzzy control |
| CN113826078A (en) * | 2019-09-11 | 2021-12-21 | 阿里巴巴集团控股有限公司 | Resource scheduling and information prediction method, device, system and storage medium |
| CN110781125A (en) * | 2019-09-12 | 2020-02-11 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Management method and system for complete cabinet super-fusion server |
| CN111174375A (en) * | 2019-12-11 | 2020-05-19 | 西安交通大学 | Data center energy consumption minimization-oriented job scheduling and machine room air conditioner regulation and control method |
| CN111174375B (en) * | 2019-12-11 | 2021-02-02 | 西安交通大学 | Data center energy consumption minimization-oriented job scheduling and machine room air conditioner regulation and control method |
| CN112070353A (en) * | 2020-08-04 | 2020-12-11 | 中国科学院信息工程研究所 | Method and system for accurately detecting energy efficiency of data center |
| CN112070353B (en) * | 2020-08-04 | 2023-09-29 | 中国科学院信息工程研究所 | Method and system for accurately detecting energy efficiency of data center |
| CN112308734A (en) * | 2020-10-27 | 2021-02-02 | 中国科学院信息工程研究所 | Method for metering and allocating non-IT energy consumption of IT equipment and electronic device |
| CN112308734B (en) * | 2020-10-27 | 2024-01-05 | 中国科学院信息工程研究所 | IT equipment non-IT energy consumption metering and cost sharing method and electronic device |
| CN112612305B (en) * | 2020-12-04 | 2022-04-08 | 格力电器（武汉）有限公司 | Temperature adjusting method, device, equipment, storage medium and air conditioning system |
| CN112612305A (en) * | 2020-12-04 | 2021-04-06 | 格力电器（武汉）有限公司 | Temperature adjusting method, device, equipment, storage medium and air conditioning system |
| CN112888268A (en) * | 2021-02-04 | 2021-06-01 | 中国工商银行股份有限公司 | Energy-saving control method, device and equipment for data center machine room and storage medium |
| CN115113705A (en) * | 2021-03-18 | 2022-09-27 | 百度（美国）有限责任公司 | Data center control hierarchy for neural network integration |
| WO2022199608A1 (en) * | 2021-03-25 | 2022-09-29 | 华为技术有限公司 | Energy saving method, apparatus and system applied to machine room |
| CN113515150A (en) * | 2021-04-12 | 2021-10-19 | 天津大学 | Variable-temperature control method for realizing instant photovoltaic power generation consumption of data center |
| CN113515150B (en) * | 2021-04-12 | 2022-05-03 | 天津大学 | Variable-temperature control method for realizing instant photovoltaic power generation consumption of data center |
| CN113835460A (en) * | 2021-11-23 | 2021-12-24 | 中铁建设集团有限公司 | Data computer lab intelligence environmental control system based on hydrodynamics analog digital twin |
| CN113835460B (en) * | 2021-11-23 | 2022-05-03 | 中铁建设集团有限公司 | Data computer lab intelligence environmental control system based on hydrodynamics analog digital twin |
| CN114063545B (en) * | 2022-01-14 | 2022-06-07 | 宁波亮控信息科技有限公司 | Data center energy consumption control system and method fusing edge calculation and controller |
| CN114063545A (en) * | 2022-01-14 | 2022-02-18 | 宁波亮控信息科技有限公司 | Data center energy consumption control system and method fusing edge calculation and controller |
| CN114595851A (en) * | 2022-02-20 | 2022-06-07 | 特斯联科技集团有限公司 | Air conditioner room power consumption analysis device using neural network |
| CN114595851B (en) * | 2022-02-20 | 2022-09-30 | 特斯联科技集团有限公司 | Air conditioner room power consumption analysis device using neural network |
| CN115907138B (en) * | 2022-11-18 | 2023-06-30 | 安华数据(东莞)有限公司 | Method, system and medium for predicting PUE value of data center |
| CN115907138A (en) * | 2022-11-18 | 2023-04-04 | 安华数据(东莞)有限公司 | Method, system and medium for predicting PUE value of data center |
| CN116193819A (en) * | 2023-01-19 | 2023-05-30 | 中国长江三峡集团有限公司 | Energy-saving control method, system and device for data center machine room and electronic equipment |
| CN116193819B (en) * | 2023-01-19 | 2024-02-02 | 中国长江三峡集团有限公司 | Energy-saving control method, system and device for data center machine room and electronic equipment |
| CN116527416A (en) * | 2023-07-03 | 2023-08-01 | 深圳市立湾科技有限公司 | Intelligent AI energy-saving control system and method applied to data center |
| CN116527416B (en) * | 2023-07-03 | 2023-09-01 | 深圳市立湾科技有限公司 | Intelligent AI energy-saving control system and method applied to data center |
| CN117421131A (en) * | 2023-12-18 | 2024-01-19 | 武汉泽塔云科技股份有限公司 | Intelligent scheduling method and system for monitoring power consumption load of server |
| CN117421131B (en) * | 2023-12-18 | 2024-03-26 | 武汉泽塔云科技股份有限公司 | Intelligent scheduling method and system for monitoring power consumption load of server |
| CN117472167A (en) * | 2023-12-28 | 2024-01-30 | 苏州元脑智能科技有限公司 | Method and device for adjusting energy consumption of server, computer equipment and storage medium |
| CN117472167B (en) * | 2023-12-28 | 2024-03-22 | 苏州元脑智能科技有限公司 | Method and device for adjusting energy consumption of server, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109800066B (en) | 2020-11-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109800066A (en) | A kind of data center's energy-saving scheduling method and system | |
| CN107223195B (en) | Variable air quantity for HVAC system models | |
| CN104698843B (en) | A kind of data center's energy-saving control method based on Model Predictive Control | |
| Tang et al. | Sensor-based fast thermal evaluation model for energy efficient high-performance datacenters | |
| Bash et al. | Dynamic thermal management of air cooled data centers | |
| US8155793B2 (en) | System and method for controlling air conditioning facilities, and system and method for power management of computer room | |
| CN102620378B (en) | Method and system for data center energy saving controlling | |
| CN104061664B (en) | The monitoring air-conditioner of communications equipment room, method and device | |
| CN109189190A (en) | A kind of data center's thermal management method based on temperature prediction | |
| US10180261B1 (en) | Model based cooling control system | |
| Berezovskaya et al. | Modular model of a data centre as a tool for improving its energy efficiency | |
| CN109116812A (en) | Intelligent power distribution cabinet, energy conserving system and method based on SparkStreaming | |
| CN107143981B (en) | A kind of controlling system of central air conditioner and method | |
| Fang et al. | QoS-driven power management of data centers via model predictive control | |
| CN102224474A (en) | System and method for assessing and managing data center airflow and energy usage | |
| CN107276069A (en) | Approximate the polymerization modeling method and system of a kind of area power grid resident temperature control load | |
| CN104423531A (en) | Data center energy consumption scheduling method and data center energy consumption scheduling device | |
| CN108510100A (en) | A kind of energy optimization management implementation method of the intelligent building based on correlation rule | |
| CN108006897A (en) | Group control method and device | |
| CN107655140B (en) | A kind of air conditioning control method and device | |
| Mohsenian et al. | A novel integrated fuzzy control system toward automated local airflow management in data centers | |
| Zhao et al. | Energy-saving and management of telecom operators’ remote computer rooms using IoT technology | |
| CN114970358A (en) | Data center energy efficiency optimization method and system based on reinforcement learning | |
| Conficoni et al. | Hpc cooling: A flexible modeling tool for effective design and management | |
| CN105423492A (en) | Machine room monitoring system and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |