CN109840180A - Server runs power consumption management method, device and computer readable storage medium - Google Patents
Server runs power consumption management method, device and computer readable storage medium Download PDFInfo
- Publication number
- CN109840180A CN109840180A CN201811550110.XA CN201811550110A CN109840180A CN 109840180 A CN109840180 A CN 109840180A CN 201811550110 A CN201811550110 A CN 201811550110A CN 109840180 A CN109840180 A CN 109840180A
- Authority
- CN
- China
- Prior art keywords
- server
- cabinet
- power loss
- actual power
- power consumption
- 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
Abstract
This programme is related to pedestal O&M, provides a kind of server operation power consumption management method, device and storage medium, and method includes: that the server of data center is classified according to same brand, same configuration;Acquire the actual power loss value of same class server, establish actual power loss load relationship table, the actual power loss of same class server calculates in the following ways, valuation power consumption=cabinet actual power loss/equipment cabinet server number, wherein, what is installed on cabinet is all same class server, is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season, in conjunction with cabinet rated disspation, cabinet configuration server scheme is formed.The present invention configures cabinet by monitoring the actual power loss of cabinet server with the actual power loss of server is foundation, and cabinet utilization rate can be improved, reduce cabinet space waste.Cabinet power supply utilization rate is improved, PUE value is reduced.
Description
Technical field
The present invention relates to pedestal O&Ms, specifically, being related to a kind of server operation power consumption management method, device and calculating
Machine readable storage medium storing program for executing.
Background technique
In data center, server brand and model is more, and server dispatches from the factory marked load power consumption as interval range, however
In order to improve the utilization rate of limited cabinet resource, data center needs the quantity of rationally setting cabinet server, and with
The increase of portfolio and type of business, it is also necessary to increase or decrease server, in cabinet with the needs of adaptation service.If needed
To increase server in cabinet, with riseing for server utilization rate, it is understood that there may be it is more than negative that power consumption load, which rises bring,
Carry risk.Therefore, data center is all to use more careful configuration mode at present, too cautious to cause cabinet power supply inabundant
It utilizes, wastes cabinet resource, assess power consumption by the famous brand that manufacturer server provides, cabinet space is caused to waste, power supply uses
Low efficiency, PUE are excessively high.
Summary of the invention
In order to solve the above technical problems, the present invention provides a kind of server operation power consumption management method, is applied to electronics and fills
Set, comprising: by the server of data center according to same brand, same configuration classify, and by same brand, same configuration
Server as same class server;The actual power loss value for acquiring same class server, the cpu load according to acquisition time
It being sampled, establishes actual power loss load relationship table, wherein the actual power loss of same class server calculates in the following ways,
Valuation power consumption=cabinet actual power loss/equipment cabinet server number, wherein what is installed on cabinet is all same class server, also,
The also monthly average actual power loss of calculation server, monthly average actual power loss=cabinet monthly average power consumption/equipment cabinet server number,
In, what is installed on cabinet is all same class server, also, this season for going back calculation server is averaged actual power loss, and this season is average
Actual power loss=cabinet season average power consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server;
It is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season, in conjunction with the specified function of cabinet
Consumption forms cabinet configuration server scheme.
Preferably, the actual power loss of each server of real-time monitoring is gone back, and seeks actual power loss and the place of each server
The difference of the average value of the actual power loss of Servers-all in cabinet, it is highest to power consumption if difference is higher than difference limit value
The CPU down conversion process of server, until the difference of the actual power loss of server is in difference limits, if difference is lower than poor
It is worth limit value, then the CPU raising frequency of the highest server of power consumption is handled, so that each server mean allocation load, power consumption reaches
It is balanced.
Preferably, it is averaged practical function in the valuation power consumption or monthly average actual power loss or this season according to same class server
Consumption after forming cabinet configuration server scheme, also assists preferred cabinet to match in conjunction with cabinet rated disspation using machine learning method
Server scheme is set, the cabinet configuration server scheme of input is received using neural network model and handles the machine of the input
Cabinet configuration server scheme according to scoring situation to select optimal to the cabinet configuration server schemes generation corresponding scores
Cabinet configuration server scheme.
Preferably, the step of optimizing cabinet configuration server scheme using machine learning method includes: building cabinet configuration
Disaggregated model;The training dataset for being trained to cabinet configuration disaggregated model is obtained, training data concentration includes pair
Answer the cabinet configuration server scheme of different business requirement description, and the score value judged according to customized standards of grading;It adopts
Disaggregated model is configured with training dataset training cabinet, passes through the different cabinet configuration server schemes for concentrating training data
It inputs cabinet and configures disaggregated model, and the output that cabinet configures disaggregated model is classified by classifier, then pass through damage
Function is lost to control the nicety of grading of cabinet configuration disaggregated model, to improve the nicety of grading of cabinet configuration disaggregated model;Benefit
The cabinet configuration disaggregated model completed with training is cabinet configuration server for business demand.
Preferably, building cabinet configuration disaggregated model is the following steps are included: setting cabinet configuration training pattern parameter, described
It is deep neural network model that cabinet, which configures training pattern, cabinet configuration training pattern include input layer, two-way GRU,
Softmax layers and full articulamentum;Multiple cabinets configuration training program is inputted into the cabinet and configures training pattern, to the machine
Cabinet configuration training pattern is trained, and updates two-way GRU parameter in the cabinet configuration training pattern;According to updated cabinet
The two-way GRU parameter of the two-way GRU parameter initialization cabinet configuration disaggregated model of training pattern is configured, while configuring cabinet configuration
The parameter in addition to two-way GRU parameter of disaggregated model;Cabinet configuration training program input cabinet is configured into training pattern, is supervised
Dual training is superintended and directed, obtains cabinet configuration disaggregated model, and dual training is supervised into newly-increased cabinet configuration server scheme input
Cabinet afterwards configures disaggregated model, carries out unsupervised virtual dual training, updates the parameter of cabinet configuration disaggregated model, obtains machine
Cabinet configures disaggregated model.
Preferably, the calculation formula of GRU is as follows:
zt=σ (ftUz+s(t-1)Wz)
rt=σ (ftUr+s(t-1)Wr)
ht=tanh (ftUh+(s(t-1)*rt)Wh)
st=(1-zt)*ft+zt*s(t-1)
Wherein, ztIt is to update door, how many candidate hidden layer h are added in controltInformation;
rtIt is resetting door, for calculating candidate hidden layer ht, the how many previous moment hidden layer s of control reservation(t-1)Information;
htIt is candidate hidden layer;
U, W is weight matrix;
ftIt is the input data of t moment;
s(t-1)It is the activation value of t-1 moment hidden layer neuron;
σ indicates sigmoid activation primitive;
Tanh is activation primitive;
stIt is the activation value of t moment hidden layer neuron.
Preferably, it is averaged practical function in the valuation power consumption or monthly average actual power loss or this season according to same class server
Consumption during forming cabinet configuration server scheme, is greater than given threshold using actual power loss in conjunction with cabinet rated disspation
The one-to-one mode of server that server and actual power loss are less than given threshold is come for cabinet configuration server, wherein institute
State the average value for the actual power loss that given threshold is inhomogeneous server.
The present invention also provides a kind of electronic device, which includes: memory and processor, is deposited in the memory
Server operation power managed program is contained, the server operation power managed program is realized such as when being executed by the processor
Lower step: by the server of data center according to same brand, same configuration classify, and by same brand, same configuration
Server as same class server;The actual power loss value for acquiring same class server, the cpu load according to acquisition time
It being sampled, establishes actual power loss load relationship table, wherein the actual power loss of same class server calculates in the following ways,
Valuation power consumption=cabinet actual power loss/equipment cabinet server number, wherein what is installed on cabinet is all same class server, also,
The also monthly average actual power loss of calculation server, monthly average actual power loss=cabinet monthly average power consumption/equipment cabinet server number,
In, what is installed on cabinet is all same class server, also, this season for going back calculation server is averaged actual power loss, and this season is average
Actual power loss=cabinet season average power consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server;
It is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season, in conjunction with the specified function of cabinet
Consumption forms cabinet configuration server scheme.
Preferably, the actual power loss of each server of real-time monitoring is gone back, and seeks actual power loss and the place of each server
The difference of the average value of the actual power loss of Servers-all in cabinet, it is highest to power consumption if difference is higher than difference limit value
The CPU down conversion process of server, until the difference of the actual power loss of server is in difference limits, if difference is lower than poor
It is worth limit value, then the CPU raising frequency of the highest server of power consumption is handled, so that each server mean allocation load, power consumption reaches
It is balanced.
The present invention also provides a kind of computer readable storage mediums, which is characterized in that the computer readable storage medium
Be stored with computer program, the computer program includes program instruction, when described program instruction is executed by processor, realize with
The upper server runs power consumption management method.
The present invention is configured by monitoring the actual power loss of cabinet server with the actual power loss of server is foundation
Cabinet utilization rate can be improved in cabinet, reduces cabinet space waste.Cabinet power supply utilization rate is improved, PUE value is reduced.
Detailed description of the invention
By the way that embodiment is described in conjunction with following accompanying drawings, features described above of the invention and technological merit will become
More understands and be readily appreciated that.
Fig. 1 is the flow diagram of the server operation power consumption management method of the embodiment of the present invention;
Fig. 2 is the hardware structure schematic diagram of the electronic device of the embodiment of the present invention;
Fig. 3 is the module structure drafting of the server operation power managed program of the embodiment of the present invention.
Specific embodiment
Below with reference to the accompanying drawings come server of the present invention operation power consumption management method, device and computer are described can
Read the embodiment of storage medium.Those skilled in the art will recognize, without departing from the spirit and scope of the present invention
In the case where, described embodiment can be modified with a variety of different modes or combinations thereof.Therefore, attached drawing and description
It is regarded as illustrative in nature, is not intended to limit the scope of the claims.In addition, in the present specification, attached drawing is not
It is drawn to scale, and identical appended drawing reference indicates identical part.
Fig. 1 is the flow diagram that server provided in an embodiment of the present invention runs power consumption management method.This method includes
Following steps:
Step S10, by the server of data center according to same brand, same configuration classify, and by same brand,
The server of same configuration is as same class server.
The main purpose of data center is operation application to handle the data of the tissue of business and running.It respectively can be with using quotient
The server of data center is rented to complete its business function to be realized.It mainly include that HP is serviced by taking certain data center as an example
Device, main type are DL 380Gen9 power consumption 160W~290W;Leonvo is the second multi-brand server, and main type is
RD650 power consumption 170W~245W;Dell server quantity is slightly less than Lenovo, and main type is R720 power consumption 220W and R730 function
200W~260W is consumed, R730XD power consumption is in 275W or so, R920 power consumption 450W, R930 power consumption 500W or so;Tide server master
Wanting type is NF5270M3 power consumption 210W or so and NF5270M power consumption 130W or so.Store 8060 dual controller power consumption of FAS
3.6KW~3.7KW or so.
Step S30 acquires the actual power loss value of same class server, and the cpu load according to acquisition time is sampled,
Establish actual power loss load relationship table.
It is all with the nominal of manufacturer server mark since current each data center is when to cabinet configuration server
Power consumption is estimated on the basis of power, estimation power consumption is usually the 25% of nominal power.However the power consumption range of each server is usually
It is that type of business relevant to type of business, different is formed by load difference again.Therefore, the function of same class server is acquired
Consumption value, OS (CPU) load according to acquisition time is sampled, and foundation and power consumption load relationship table, as shown in Table 1.
Table one
Wherein, the actual power loss of same class server calculates in the following ways,
Valuation power consumption=cabinet actual power loss/equipment cabinet server number, wherein what is installed on cabinet is all same class service
Device,
Also, the monthly average actual power loss of calculation server is gone back, monthly average actual power loss=cabinet monthly average power consumption/cabinet
Server number, wherein what is installed on cabinet is all same class server,
Also, this season for going back calculation server is averaged actual power loss, and this season is averaged the function that is averaged in actual power loss=cabinet season
Consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server;
Step S50 is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season,
In conjunction with cabinet rated disspation, cabinet configuration server scheme is formed.
Illustratively the present embodiment is come with specific example below.Data center needs the server of fair amount, configuration
Distribute to it is different apply quotient, so that application quotient can carry out the business of itself using the server of distribution, for example, Jingdone district is leased
The multiple servers of data center are supported the business in its Jingdone district store.In the present embodiment, it is contemplated that same class server
The actual power loss of corresponding different business, can play the efficiency of server to a greater extent.According to type of business, cabinet is set
In server brand, configuration and quantity.For example, the nominal power of the R430 server of Dell is 1100W, power consumption is estimated
About the 25% of nominal power, i.e. 275W, the rated disspation 4000W of cabinet, if not considering the factors such as power supply safety, heat dissipation,
Cabinet can configure R430 server 14 of Dell.If comprehensively considering cabinet power supply safety, it can configure 12 cabinets.And the Dell
R430 server valuation power consumption be 140W, it is seen that than it is described estimation it is low in energy consumption, if do not consider cabinet power supply, heat dissipation etc. influence
Factor only considers to configure cabinet with the rated disspation of cabinet and the valuation power consumption of server, then cabinet is 28 at most configurable
The R430 server of Dell.Comprehensively consider the factors such as cabinet power supply safety, heat dissipation, can configure the R430 server of 20 Dell.
Although with the above arrangement, server is increased in cabinet, since its actual power loss is all in zone of reasonableness
It is interior, the maximum deploying servers of cabinet can be made, can make full use of cabinet power supply, improve the use effect of cabinet power supply
Rate reduces PUE (power supply service efficiency) value, it is also possible that the power supply of cabinet maintains in the range of powering threshold value.
In one alternate embodiment, the actual power loss of each server of real-time monitoring is gone back, and seeks the reality of each server
The difference of the average value of the actual power loss of Servers-all in border power consumption and place cabinet, if difference is higher than difference limit value,
To the CPU down conversion process of the highest server of power consumption, until the difference of the actual power loss of server is in difference limits, such as
Fruit difference is lower than difference limit value, then handles the CPU raising frequency of the highest server of power consumption, so that each server mean allocation is negative
It carries, power consumption reaches balanced.
In one alternate embodiment, in the valuation power consumption or monthly average actual power loss or this season according to same class server
Average actual power loss is greater than during forming cabinet configuration server scheme using actual power loss in conjunction with cabinet rated disspation
The one-to-one mode of server that the server and actual power loss of given threshold are less than given threshold, which to configure for cabinet, to be serviced
Device, wherein the given threshold is the average value of the actual power loss of inhomogeneous server.
In one alternate embodiment, in the valuation power consumption or monthly average actual power loss or this season according to same class server
Average actual power loss after forming cabinet configuration server scheme, is also assisted using machine learning method in conjunction with cabinet rated disspation
Cabinet configuration server scheme receives the cabinet configuration server scheme of input using neural network model and handles described defeated
The cabinet configuration server scheme entered according to scoring situation to select to the cabinet configuration server schemes generation corresponding scores
Optimal cabinet configuration server scheme.
Further, using machine learning method extension cabinet configuration server scheme the step of, is as follows:
It constructs cabinet and configures disaggregated model;
The training dataset for training cabinet configuration disaggregated model is obtained, training data concentration includes that correspondence is not of the same trade or business
The allocation plan of the cabinet of business requirement description, and the score value judged according to customized standards of grading;
Disaggregated model is configured using training dataset training cabinet, is configured by the different cabinets for concentrating training data
Scheme inputs cabinet and configures disaggregated model, and the output that cabinet configures disaggregated model is classified by classifier, then leads to
Loss function is crossed to control the nicety of grading of cabinet configuration disaggregated model, to improve the classification essence of cabinet configuration disaggregated model
Degree.
Further, building cabinet configuration disaggregated model the following steps are included:
Cabinet is set and configures training pattern parameter, the cabinet configuration training pattern is deep neural network model, described
It includes input layer, two-way GRU, softmax layer and full articulamentum that cabinet, which configures training pattern,;
Multiple cabinets configuration training program is inputted into the cabinet and configures training pattern, training pattern is configured to the cabinet
It is trained, updates two-way GRU parameter in the cabinet configuration training pattern;
The two-way of the two-way GRU parameter initialization cabinet configuration disaggregated model of training pattern is configured according to updated cabinet
GRU parameter, while configuring the parameter in addition to two-way GRU parameter of cabinet configuration disaggregated model;
Cabinet configuration training program input cabinet is configured into training pattern, exercise supervision dual training, obtains cabinet configuration
Disaggregated model, and the cabinet after newly-increased cabinet allocation plan input supervision dual training is configured into disaggregated model, it carries out without prison
Virtual dual training is superintended and directed, the parameter of cabinet configuration disaggregated model is updated, cabinet is obtained and configures disaggregated model.
Further, the calculation formula of GRU is as follows:
zt=σ (ftUz+s(t-1)Wz)
rt=σ (ftUr+s(t-1)Wr)
ht=tanh (ftUh+(s(t-1)*rt)Wh)
st=(1-zt)*ft+zt*s(t-1)
Wherein, ztIt is to update door, how many candidate hidden layer h are added in controltInformation;
rtIt is resetting door, for calculating candidate hidden layer ht, the how many previous moment hidden layer s of control reservation(t-1)Information;
htIt is candidate hidden layer;
U, W is weight matrix;
ftIt is the input data of t moment;
s(t-1)It is the activation value of t-1 moment hidden layer neuron;
σ indicates sigmoid activation primitive;
Tanh is activation primitive;
stIt is the activation value of t moment hidden layer neuron.
As shown in fig.2, being the hardware structure schematic diagram of the embodiment of electronic device of the present invention.It is described in the present embodiment
Electronic device 2 be it is a kind of can according to the instruction for being previously set or store, automatic progress numerical value calculating and/or information processing
Equipment.For example, it may be smart phone, tablet computer, laptop, desktop computer, rack-mount server, blade type take
It is engaged in device, tower server or Cabinet-type server (including server set composed by independent server or multiple servers
Group) etc..As shown in Fig. 2, the electronic device 2 includes at least, but it is not limited to, depositing for connection can be in communication with each other by system bus
Reservoir 21, processor 22, network interface 23.Wherein: the memory 21 includes at least a type of computer-readable storage
Medium, the readable storage medium storing program for executing include flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory etc.),
Random access storage device (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable are only
Read memory (EEPROM), programmable read only memory (PROM), magnetic storage, disk, CD etc..In some embodiments
In, the memory 21 can be the internal storage unit of the electronic device 2, such as the hard disk or memory of the electronic device 2.
In further embodiments, the memory 21 is also possible to the External memory equipment of the electronic device 2, such as electronics dress
Set the plug-in type hard disk being equipped on 2, intelligent memory card (Smart Media Card, SMC), secure digital (Secure
Digital, SD) card, flash card (Flash Card) etc..Certainly, the memory 21 can also both include the electronic device 2
Internal storage unit also include its External memory equipment.In the present embodiment, the memory 21 is installed on commonly used in storage
Operating system and types of applications software of the electronic device 2, such as server operation power managed program code etc..This
Outside, the memory 21 can be also used for temporarily storing the Various types of data that has exported or will export.
The processor 22 can be in some embodiments central processing unit (Central Processing Unit,
CPU), controller, microcontroller, microprocessor or other data processing chips.The processor 22 is commonly used in the control electricity
The overall operation of sub-device 2, such as execute control relevant to the electronic device 2 progress data interaction or communication and processing
Deng.In the present embodiment, the processor 22 is for running the program code stored in the memory 21 or processing data, example
Server as described in running runs power managed program.
The network interface 23 may include radio network interface or wired network interface, which is commonly used in
Communication connection is established between the electronic device 2 and other electronic devices.For example, the network interface 23 is used to incite somebody to action by network
The electronic device 2 is connected with push platform, and data transmission channel is established between the electronic device 2 and push platform and is led to
Letter connection etc..The network can be intranet (Intranet), internet (Internet), global system for mobile communications
(Global System of Mobile communication, GSM), wideband code division multiple access (Wideband
CodeDivision Multiple Access, WCDMA), 4G network, 5G network, bluetooth (Bluetooth), Wi-Fi etc. is wireless
Or cable network.
Optionally, which can also include display, and display is referred to as display screen or display unit.
It can be light-emitting diode display, liquid crystal display, touch-control liquid crystal display and Organic Light Emitting Diode in some embodiments
(Organic Light-Emitting Diode, OLED) display etc..Display is used to be shown in handle in electronic device 2
Information and for showing visual user interface.
It should be pointed out that Fig. 2 illustrates only the electronic device 2 with component 21-23, it should be understood that not
It is required that implement all components shown, the implementation that can be substituted is more or less component.
It may include operating system, server operation power managed program 50 in memory 21 comprising readable storage medium storing program for executing
Deng.Processor 22 realizes following steps when executing server operation power managed program 50 in memory 21:
Step S10, by the server of data center according to same brand, same configuration classify, and by same brand,
The server of same configuration is as same class server.
Step S30 acquires the actual power loss value of same class server, and the cpu load according to acquisition time is sampled,
Actual power loss load relationship table is established, actual power loss can be valuation power consumption or monthly average actual power loss or this season is average practical
Power consumption.
Step S50 is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season,
In conjunction with cabinet rated disspation, cabinet configuration server scheme is formed.
In the present embodiment, the server operation power managed program being stored in memory 21 can be divided into
One or more program module, one or more of program modules are stored in memory 21, and can by one or
Multiple processors (the present embodiment is processor 22) are performed, to complete the present invention.For example, Fig. 3 shows the server fortune
The program module schematic diagram of row power managed program, in the embodiment, the server operation power managed program 50 can be by
It is divided into classification server module 501, power consumption load relationship table establishes module 502, cabinet configuration module 503.Wherein, of the invention
So-called program module is the series of computation machine program instruction section for referring to complete specific function, than program more suitable for description
Implementation procedure of the server operation power managed program in the electronic device 2.It will be described below described in specifically introducing
The concrete function of program module.
Classification server module 501 is used to classify the server of data center according to same brand, same configuration,
And using same brand, same configuration server as same class server.
Power consumption load relationship table establishes module 502 for establishing power consumption load relationship table, since current each data center exists
When to cabinet configuration server, all it is to estimate power consumption on the basis of the nominal power of manufacturer server mark, estimates function
Consumption is usually the 25% of nominal power.However the power consumption range of each server usually again be it is relevant to type of business, it is different
It is different that type of business is formed by load.Therefore, the power consumption number for acquiring same class server, the OS according to acquisition time
(CPU) load is sampled, and foundation and power consumption load relationship table, as shown in Table 1.
Wherein, the actual power loss of same class server calculates in the following ways,
Valuation power consumption=cabinet actual power loss/equipment cabinet server number, wherein what is installed on cabinet is all same class service
Device,
Also, the monthly average actual power loss of calculation server is gone back, monthly average actual power loss=cabinet monthly average power consumption/cabinet
Server number, wherein what is installed on cabinet is all same class server,
Also, this season for going back calculation server is averaged actual power loss, and this season is averaged the function that is averaged in actual power loss=cabinet season
Consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server;
Cabinet configuration module 503 is used for flat according to the valuation power consumption or monthly average actual power loss or this season of same class server
Equal actual power loss is cabinet configuration server in conjunction with cabinet rated disspation.
Illustratively the present embodiment is come with specific example below.Data center needs the server of fair amount, configuration
Distribute to it is different apply quotient, so that application quotient can carry out the business of itself using the server of distribution, for example, Jingdone district is leased
The multiple servers of data center are supported the business in its Jingdone district store.In the present embodiment, it is contemplated that same class server
The actual power loss of corresponding different business, can play the efficiency of server to a greater extent.According to type of business, cabinet is set
In server brand, configuration and quantity.For example, the nominal power of the R430 server of Dell is 1100W, power consumption is estimated
About the 25% of nominal power, i.e. 275W, the rated disspation 4000W of cabinet, then comprehensively consider cabinet power supply safety, it can configure 12
The R430 server of platform Dell.And the valuation power consumption of the R430 server of the Dell is 140W, it is seen that it is more low in energy consumption than the estimation,
Comprehensively consider cabinet power supply safety, can configure 20, the R430 server of Dell.
It in one alternate embodiment, further include actual power loss monitoring modular 504, the reality of each server of real-time monitoring
Power consumption, and the difference of the actual power loss and the average value of the actual power loss of Servers-all in the cabinet of place of each server is sought,
If difference is higher than difference limit value, to the CPU down conversion process of the highest server of power consumption, until the actual power loss of server
Difference is in difference limits, if difference is lower than difference limit value, handles the CPU raising frequency of the highest server of power consumption,
So that each server mean allocation load, power consumption reach balanced.
In one alternate embodiment, in the valuation power consumption or monthly average actual power loss or this season according to same class server
Average actual power loss, in conjunction with cabinet rated disspation, during forming cabinet allocation plan, cabinet configuration module 503 is using real
The one-to-one mode of server that border power consumption is greater than the server of given threshold and actual power loss is less than given threshold is come for machine
Cabinet configuration server, wherein the given threshold is the average value of the actual power loss of inhomogeneous server.
It in one alternate embodiment, further include intelligent configuration module 505, in the valuation power consumption according to same class server
Or monthly average actual power loss or this season are averaged actual power loss, and in conjunction with cabinet rated disspation, after forming cabinet configuration server scheme,
Also using the machine learning method extension cabinet configuration server scheme of intelligent configuration module 505, connect using neural network model
It receives the cabinet configuration server scheme of input and handles the cabinet configuration server scheme of the input and taken with being configured to cabinet
Business device schemes generation corresponding scores select optimal allocation plan according to scoring situation.
Further, the step of intelligent configuration module 505 is using machine learning method extension cabinet allocation plan is as follows:
It constructs cabinet and configures disaggregated model;
The training dataset for training cabinet configuration disaggregated model is obtained, training data concentration includes that correspondence is not of the same trade or business
The allocation plan of the cabinet of business requirement description, and the score value judged according to customized standards of grading;
Disaggregated model is configured using training dataset training cabinet, is configured by the different cabinets for concentrating training data
Scheme inputs cabinet and configures disaggregated model, and the output that cabinet configures disaggregated model is classified by classifier, then leads to
Loss function is crossed to control the nicety of grading of cabinet configuration disaggregated model, to improve the classification essence of cabinet configuration disaggregated model
Degree.
Further, intelligent configuration module 505 construct cabinet configuration disaggregated model the following steps are included:
Cabinet is set and configures training pattern parameter, the cabinet configuration training pattern is deep neural network model, described
It includes input layer, two-way GRU, softmax layer and full articulamentum that cabinet, which configures training pattern,;
Multiple cabinets configuration training program is inputted into the cabinet and configures training pattern, training pattern is configured to the cabinet
It is trained, updates two-way GRU parameter in the cabinet configuration training pattern;
The two-way of the two-way GRU parameter initialization cabinet configuration disaggregated model of training pattern is configured according to updated cabinet
GRU parameter, while configuring the parameter in addition to two-way GRU parameter of cabinet configuration disaggregated model;
Cabinet configuration training program input cabinet is configured into training pattern, exercise supervision dual training, obtains cabinet configuration
Disaggregated model, and the cabinet after newly-increased cabinet allocation plan input supervision dual training is configured into disaggregated model, it carries out without prison
Virtual dual training is superintended and directed, the parameter of cabinet configuration disaggregated model is updated, cabinet is obtained and configures disaggregated model.
Further, the calculation formula of GRU is as follows:
zt=σ (ftUz+s(t-1)Wz)
rt=σ (ftUr+s(t-1)Wr)
ht=tanh (ftUh+(s(t-1)*rt)Wh)
st=(1-zt)*ft+zt*s(t-1)
Wherein, ztIt is to update door, how many candidate hidden layer h are added in controltInformation;
rtIt is resetting door, for calculating candidate hidden layer ht, the how many previous moment hidden layer s of control reservation(t-1)Information;
htIt is candidate hidden layer;
U, W is weight matrix;
ftIt is the input data of t moment;
s(t-1)It is the activation value of t-1 moment hidden layer neuron;
σ indicates sigmoid activation primitive;
Tanh is activation primitive;
stIt is the activation value of t moment hidden layer neuron.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
It can be hard disk, multimedia card, SD card, flash card, SMC, read-only memory (ROM), Erasable Programmable Read Only Memory EPROM
(EPROM), any one in portable compact disc read-only memory (CD-ROM), USB storage etc. or several timess
Meaning combination.It include server operation power managed program etc. in the computer readable storage medium, the server runs function
It consumes and realizes following operation when management program 50 is executed by processor 22:
Step S10, by the server of data center according to same brand, same configuration classify, and by same brand,
The server of same configuration is as same class server.
Step S30 acquires the actual power loss value of same class server, and the cpu load according to acquisition time is sampled,
Actual power loss load relationship table is established, it is average practical that actual power loss value can be valuation power consumption, monthly average actual power loss or this season
Power consumption.
Step S50 is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season,
In conjunction with cabinet rated disspation, cabinet configuration server scheme is formed.
The specific embodiment of the computer readable storage medium of the present invention and above-mentioned server run power consumption management method
And the specific embodiment of electronic device 2 is roughly the same, details are not described herein.
The above description is only a preferred embodiment of the present invention, is not intended to restrict the invention, for those skilled in the art
For member, the invention may be variously modified and varied.All within the spirits and principles of the present invention, it is made it is any modification,
Equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of server runs power consumption management method, it is applied to electronic device characterized by comprising
The server of data center is classified according to same brand, same configuration, and by the clothes of same brand, same configuration
Device be engaged in as same class server;
The actual power loss value for acquiring same class server, the cpu load according to acquisition time are sampled, and establish actual power loss
Load relationship table, wherein the actual power loss of same class server calculates in the following ways, the valuation power consumption=practical function of cabinet
Consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server, also, goes back the monthly average of calculation server
Actual power loss, monthly average actual power loss=cabinet monthly average power consumption/equipment cabinet server number, wherein what is installed on cabinet is all
Same class server, also, this season for going back calculation server is averaged actual power loss, this season actual power loss=cabinet season that is averaged are flat
Equal power consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server;
It is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season, it is specified in conjunction with cabinet
Power consumption forms cabinet configuration server scheme.
2. server according to claim 1 runs power consumption management method, which is characterized in that also real-time monitoring each service
The actual power loss of device, and seek the average value of the actual power loss and the actual power loss of Servers-all in the cabinet of place of each server
Difference, if difference is higher than difference limit value, to the CPU down conversion process of the highest server of power consumption, until the reality of server
The difference of border power consumption is in difference limits, if difference is lower than difference limit value, to the CPU liter of the highest server of power consumption
Frequency is handled, so that each server mean allocation load, power consumption reach balanced.
3. server according to claim 1 runs power consumption management method, which is characterized in that according to same class server
Valuation power consumption or monthly average actual power loss or this season actual power loss that is averaged in conjunction with cabinet rated disspation form cabinet configuration clothes
After device scheme of being engaged in, preferred cabinet configuration server scheme is also assisted using machine learning method, is received using neural network model
The cabinet configuration server scheme of input and handle the cabinet configuration server scheme of the input with to the cabinet configure
Server schemes generation corresponding scores select optimal cabinet configuration server scheme according to scoring situation.
4. server according to claim 3 runs power consumption management method, which is characterized in that excellent using machine learning method
Change cabinet configuration server scheme the step of include:
It constructs cabinet and configures disaggregated model;
The training dataset for being trained to cabinet configuration disaggregated model is obtained, training data concentration includes corresponding different
The cabinet configuration server scheme of business demand description, and the score value judged according to customized standards of grading;
Disaggregated model is configured using training dataset training cabinet, passes through the different cabinet configuration services for concentrating training data
Device scheme inputs cabinet and configures disaggregated model, and the output that cabinet configures disaggregated model is classified by classifier, then
The nicety of grading of cabinet configuration disaggregated model is controlled by loss function, to improve the classification essence of cabinet configuration disaggregated model
Degree;
The cabinet configuration disaggregated model completed using training is cabinet configuration server for business demand.
5. server according to claim 4 runs power consumption management method, which is characterized in that building cabinet configuration classification mould
For type the following steps are included: setting cabinet configures training pattern parameter, the cabinet configuration training pattern is deep neural network mould
Type, the cabinet configuration training pattern includes input layer, two-way GRU, softmax layer and full articulamentum;Multiple cabinets are matched
It sets training program and inputs the cabinet configuration training pattern, cabinet configuration training pattern is trained, the machine is updated
Cabinet configures two-way GRU parameter in training pattern;The two-way GRU parameter initialization machine of training pattern is configured according to updated cabinet
Cabinet configures the two-way GRU parameter of disaggregated model, while configuring the parameter in addition to two-way GRU parameter of cabinet configuration disaggregated model;
Cabinet configuration training program input cabinet is configured into training pattern, exercise supervision dual training, cabinet configuration disaggregated model is obtained,
And the cabinet after newly-increased cabinet configuration server scheme input supervision dual training is configured into disaggregated model, carry out unsupervised void
Quasi- dual training updates the parameter of cabinet configuration disaggregated model, obtains cabinet and configures disaggregated model.
6. server according to claim 5 runs power consumption management method, which is characterized in that the calculation formula of GRU is as follows:
zt=σ (ftUz+s(t-1)Wz)
rt=σ (ftUr+s(t-1)Wr)
ht=tanh (ftUh+(s(t-1)*rt)Wh)
st=(1-zt)*ft+zt*s(t-1)
Wherein, ztIt is to update door, how many candidate hidden layer h are added in controltInformation;
rtIt is resetting door, for calculating candidate hidden layer ht, the how many previous moment hidden layer s of control reservation(t-1)Information;
htIt is candidate hidden layer;
U, W is weight matrix;
ftIt is the input data of t moment;
s(t-1)It is the activation value of t-1 moment hidden layer neuron;
σ indicates sigmoid activation primitive;
Tanh is activation primitive;
stIt is the activation value of t moment hidden layer neuron.
7. server according to claim 1 runs power consumption management method, which is characterized in that according to same class server
Valuation power consumption or monthly average actual power loss or this season actual power loss that is averaged in conjunction with cabinet rated disspation form cabinet configuration clothes
During device scheme of being engaged in, it is less than the service of given threshold greater than the server and actual power loss of given threshold using actual power loss
The one-to-one mode of device is come for cabinet configuration server, wherein the given threshold is the practical function of inhomogeneous server
The average value of consumption.
8. a kind of electronic device, which is characterized in that the electronic device includes: memory and processor, is stored in the memory
There is server to run power managed program, the server operation power managed program is realized as follows when being executed by the processor
Step:
The server of data center is classified according to same brand, same configuration, and by the clothes of same brand, same configuration
Device be engaged in as same class server;
The actual power loss value for acquiring same class server, the cpu load according to acquisition time are sampled, and establish actual power loss
Load relationship table, wherein the actual power loss of same class server calculates in the following ways, the valuation power consumption=practical function of cabinet
Consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server, also, goes back the monthly average of calculation server
Actual power loss, monthly average actual power loss=cabinet monthly average power consumption/equipment cabinet server number, wherein what is installed on cabinet is all
Same class server, also, this season for going back calculation server is averaged actual power loss, this season actual power loss=cabinet season that is averaged are flat
Equal power consumption/equipment cabinet server number, wherein what is installed on cabinet is all same class server;
It is averaged actual power loss according to the valuation power consumption of same class server or monthly average actual power loss or this season, it is specified in conjunction with cabinet
Power consumption forms cabinet configuration server scheme.
9. electronic device according to claim 8, which is characterized in that the also actual power loss of each server of real-time monitoring,
And the difference of the actual power loss and the average value of the actual power loss of Servers-all in the cabinet of place of each server is sought, if poor
Value is higher than difference limit value, then to the CPU down conversion process of the highest server of power consumption, until the difference of the actual power loss of server exists
In difference limits, if difference is lower than difference limit value, the CPU raising frequency of the highest server of power consumption is handled, so that respectively
The load of platform server mean allocation, power consumption reach balanced.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer journey
Sequence, the computer program include that program instruction is realized in claim 1 to 7 and appointed when described program instruction is executed by processor
Server described in one runs power consumption management method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811550110.XA CN109840180B (en) | 2018-12-18 | 2018-12-18 | Server operation power consumption management method and device and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811550110.XA CN109840180B (en) | 2018-12-18 | 2018-12-18 | Server operation power consumption management method and device and computer readable storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109840180A true CN109840180A (en) | 2019-06-04 |
CN109840180B CN109840180B (en) | 2022-07-08 |
Family
ID=66883224
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811550110.XA Active CN109840180B (en) | 2018-12-18 | 2018-12-18 | Server operation power consumption management method and device and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109840180B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110389637A (en) * | 2019-06-26 | 2019-10-29 | 苏州浪潮智能科技有限公司 | A kind of automatic setting method, BMC and the server of server radiating strategy |
CN110837724A (en) * | 2019-11-07 | 2020-02-25 | 交控科技股份有限公司 | Automatic processing method and device for data communication cabinet |
CN111198799A (en) * | 2019-12-31 | 2020-05-26 | 苏州浪潮智能科技有限公司 | Machine room power consumption early warning method, system, terminal and storage medium based on LSTM |
CN111736549A (en) * | 2020-06-11 | 2020-10-02 | 上海申瑞继保电气有限公司 | Production line energy-saving auxiliary equipment control method |
CN113748398A (en) * | 2019-09-06 | 2021-12-03 | 阿里巴巴集团控股有限公司 | Data processing and task scheduling method, device, system and storage medium |
CN114064282A (en) * | 2021-11-23 | 2022-02-18 | 北京百度网讯科技有限公司 | Resource mining method and device and electronic equipment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101277200A (en) * | 2007-03-30 | 2008-10-01 | 联想(北京)有限公司 | Method and device for managing multiserver power supply |
US20130289789A1 (en) * | 2009-06-25 | 2013-10-31 | Server Technology, Inc. | Power distribution apparatus with input and output power sensing and method of use |
US20150256386A1 (en) * | 2014-03-06 | 2015-09-10 | Dell Products, Lp | System and Method for Providing a Server Rack Management Controller |
CN105930253A (en) * | 2016-05-05 | 2016-09-07 | 华中科技大学 | Non-invasive server energy consumption decomposition system and method |
CN105975047A (en) * | 2016-04-29 | 2016-09-28 | 北京百度网讯科技有限公司 | Method and system for adjusting power consumption of cloud data centre |
CN107329811A (en) * | 2017-06-09 | 2017-11-07 | 北京云集智造科技有限公司 | A kind of power consumption of data center adjusting method and device |
CN108874112A (en) * | 2018-06-28 | 2018-11-23 | 郑州云海信息技术有限公司 | A kind of whole machine cabinet power consumption limit method, apparatus, equipment and storage medium |
-
2018
- 2018-12-18 CN CN201811550110.XA patent/CN109840180B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101277200A (en) * | 2007-03-30 | 2008-10-01 | 联想(北京)有限公司 | Method and device for managing multiserver power supply |
US20130289789A1 (en) * | 2009-06-25 | 2013-10-31 | Server Technology, Inc. | Power distribution apparatus with input and output power sensing and method of use |
US20150256386A1 (en) * | 2014-03-06 | 2015-09-10 | Dell Products, Lp | System and Method for Providing a Server Rack Management Controller |
CN105975047A (en) * | 2016-04-29 | 2016-09-28 | 北京百度网讯科技有限公司 | Method and system for adjusting power consumption of cloud data centre |
CN105930253A (en) * | 2016-05-05 | 2016-09-07 | 华中科技大学 | Non-invasive server energy consumption decomposition system and method |
CN107329811A (en) * | 2017-06-09 | 2017-11-07 | 北京云集智造科技有限公司 | A kind of power consumption of data center adjusting method and device |
CN108874112A (en) * | 2018-06-28 | 2018-11-23 | 郑州云海信息技术有限公司 | A kind of whole machine cabinet power consumption limit method, apparatus, equipment and storage medium |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110389637A (en) * | 2019-06-26 | 2019-10-29 | 苏州浪潮智能科技有限公司 | A kind of automatic setting method, BMC and the server of server radiating strategy |
CN113748398A (en) * | 2019-09-06 | 2021-12-03 | 阿里巴巴集团控股有限公司 | Data processing and task scheduling method, device, system and storage medium |
CN113748398B (en) * | 2019-09-06 | 2023-01-17 | 阿里云计算有限公司 | Data processing and task scheduling method, device, system and storage medium |
CN110837724A (en) * | 2019-11-07 | 2020-02-25 | 交控科技股份有限公司 | Automatic processing method and device for data communication cabinet |
CN110837724B (en) * | 2019-11-07 | 2023-09-29 | 交控科技股份有限公司 | Automatic processing method and device for data communication cabinet |
CN111198799A (en) * | 2019-12-31 | 2020-05-26 | 苏州浪潮智能科技有限公司 | Machine room power consumption early warning method, system, terminal and storage medium based on LSTM |
CN111736549A (en) * | 2020-06-11 | 2020-10-02 | 上海申瑞继保电气有限公司 | Production line energy-saving auxiliary equipment control method |
CN114064282A (en) * | 2021-11-23 | 2022-02-18 | 北京百度网讯科技有限公司 | Resource mining method and device and electronic equipment |
CN114064282B (en) * | 2021-11-23 | 2023-07-25 | 北京百度网讯科技有限公司 | Resource mining method and device and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN109840180B (en) | 2022-07-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109840180A (en) | Server runs power consumption management method, device and computer readable storage medium | |
Kołodziej et al. | Energy efficient genetic‐based schedulers in computational grids | |
CN102780759B (en) | Based on the cloud computing resource scheduling method in regulation goal space | |
Simaria et al. | A genetic algorithm based approach to the mixed-model assembly line balancing problem of type II | |
Guazzone et al. | A game-theoretic approach to coalition formation in green cloud federations | |
CN104038540B (en) | Method and system for automatically selecting application proxy server | |
CN104065745A (en) | Cloud computing dynamic resource scheduling system and method | |
CN108108224B (en) | Virtual machine placement method in cloud data center based on ant colony optimization algorithm | |
CN108416523A (en) | Method for scheduling task, device, electronic equipment and storage medium | |
CN109800082A (en) | In conjunction with the method, apparatus and storage medium of actual power loss buying server | |
CN110162390B (en) | Task allocation method and system for fog computing system | |
Han et al. | Scaling up cooperative game theory-based energy management using prosumer clustering | |
Çakmak et al. | Optimal scheduling of time shiftable loads in a task scheduling based demand response program by symbiotic organisms search algorithm | |
CN107870810A (en) | Using method for cleaning, device, storage medium and electronic equipment | |
CN104268769A (en) | Time-of-use electricity price design method based on model clustering method | |
Huang et al. | Applications of machine learning to resource management in cloud computing | |
CN107239853B (en) | Intelligent housekeeper system based on cloud computing and working method thereof | |
CN109800125A (en) | Cabinet configuring management method, device and computer readable storage medium | |
Abbasi et al. | Evolutionary green computing solutions for distributed cyber physical systems | |
Zhou et al. | Designing pricing incentive mechanism for proactive demand response in smart grid | |
KR101517996B1 (en) | VPP Operation System based on Multi-Agent System | |
CN110533218A (en) | A kind of resident's multiple-objection optimization electricity consumption strategy and system based on data mining | |
Zhang et al. | Structure-aware stochastic load management in smart grids | |
CN104917298A (en) | Automatic demand response based load control device and method for high energy consumption type enterprise | |
Xu et al. | VMs placement strategy based on distributed parallel ant colony optimization algorithm |
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 |