A server power consumption management method, device, computer equipment and storage medium

Abstract

The present application relates to a server power consumption management method, device, computer equipment and storage medium. The method comprises: obtaining server real-time data, wherein the server real-time data comprises component real-time operation data and real-time ambient temperature data; based on a heat dissipation correction model, obtaining a component heat dissipation correction factor according to the component real-time operation data and the real-time ambient temperature data; based on a combined prediction model and the component real-time operation data, obtaining a component normal predicted power consumption; according to the component heat dissipation correction factor, correcting the component normal predicted power consumption to obtain a component corrected predicted power consumption; and determining a power consumption adjustment method of the server according to the component corrected predicted power consumption and the server real-time operation data. This method can accurately predict server power consumption, improve the efficiency and accuracy of optimizing server power consumption, and thus achieve efficient and flexible power consumption adjustment.

Description

Technical Field

[0001] The present application relates to the technical field of server power consumption management, and in particular to a server power consumption management method, apparatus, computer equipment and storage medium. Background Art

[0002] With the rapid development of technologies such as cloud computing, big data analysis and artificial intelligence, the number of servers and processing power in data centers have grown exponentially, and server power consumption has become an important part of operating costs. Server power consumption management not only directly affects the power consumption and operating costs of data centers, but also has an important impact on environmental sustainability.

[0003] For the power consumption management of servers, it is usually necessary to base it on power consumption prediction. Accurately predicting the power consumption of servers not only helps to reduce the long-term energy consumption of servers, but also increases the service life of the equipment. However, traditional power consumption prediction methods mainly rely on empirical formulas and statistical analysis. These methods often have difficulty achieving high prediction accuracy when facing complex and changeable server workloads. In recent years, although deep learning technology has provided new ideas for server power consumption prediction, most of the power consumption prediction methods currently used by technicians are limited to a single model and cannot fully capture the complexity and nonlinear relationship of the data, resulting in large errors in the model's power consumption prediction results. As a result, the optimization efficiency and accuracy of power consumption management based on traditional power consumption prediction methods are not ideal. Summary of the invention

[0004] Based on this, it is necessary to provide a server power consumption management method, device, computer equipment and storage medium that can accurately predict server power consumption, improve the efficiency and accuracy of optimizing server power consumption, and thus achieve efficient and flexible power consumption adjustment in response to the above technical problems.

[0005] In one aspect, a method for managing server power consumption is provided, the method comprising:

[0006] Acquire server real-time data, wherein the server real-time data includes component real-time operation data and real-time ambient temperature data;

[0007] Based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained;

[0008] Based on the combined prediction model and the real-time operation data of the component, a normal predicted power consumption of the component is obtained;

[0009] According to the component heat dissipation correction factor, the normal predicted power consumption of the component is corrected to obtain the component corrected predicted power consumption;

[0010] A power consumption adjustment method of the server is determined according to the component revised predicted power consumption and the real-time operation data of the server.

[0011] In one embodiment, the server includes a service resource component and a heat dissipation component, and the heat dissipation correction model is obtained by:

[0012] Acquire server historical data, wherein the server historical data includes: component historical operation data and server historical ambient temperature data;

[0013] Constructing an original power consumption model corresponding to the service resource component, and calculating the original power consumption of the service resource component;

[0014] Obtaining actual power consumption of the service resource component according to the historical operation data of the component;

[0015] Calculating the difference between the original power consumption and the actual power consumption of the service resource component to obtain a power consumption deviation correction factor;

[0016] Constructing a heat dissipation efficiency formula based on the original power consumption of the service resource component and the historical data of the server;

[0017] According to the heat dissipation efficiency formula and in combination with the power consumption deviation correction factor, a heat dissipation correction model is obtained.

[0018] In one embodiment, the service resource component includes at least one of the following: a computing element, a memory element, a network element, and a hard disk element, and the step of constructing an original power consumption model corresponding to the service resource component and calculating the original power consumption of the service resource component includes:

[0019] Constructing an original computing power consumption model of the computing element, wherein the original computing power consumption model includes the following formula:

[0020] P CPU =(P max -P idle )·U CPU +P idle

[0021] Among them, P CPU represents the raw power of the computing element, P max represents the maximum power of the computing element, P idle represents the idle power of the computing element, U CPU represents the utilization rate of the computing element, that is, the percentage of the computing element actually used in a given time;

[0022] Construct an original memory power consumption model of the memory element, wherein the original memory power consumption model includes the following formula:

[0023] P memory =P Mrd ·U Mrd +P Mwr ·U Mwr

[0024] Among them, P memory represents the raw power of the memory element, P Mrd represents the read state power of the memory element, P Mwr represents the write state power of the memory element, U Mrd Indicates the read operation ratio of the memory element, U Mwr Indicates the write operation ratio of the memory element;

[0025] Constructing an original network power consumption model of the network element, wherein the original network power consumption model includes the following formula:

[0026]

[0027] Among them, P net is the original power of the network element, B represents the bandwidth of the network element, ρ represents the bandwidth utilization of the network element, and τ represents the power efficiency of the network element;

[0028] Construct an original hard disk power consumption model of the hard disk element, wherein the original hard disk power consumption model includes the following formula:

[0029] P disk =P Drd ·U Drd +P Dwr ·U Dwr +P Didle ·(1-U Drd -U Dwr )

[0030] Among them, P disk represents the raw power of the hard disk element, P Drd represents the read state power of the hard disk component, P Dwr represents the write power of the hard disk element, P Didle represents the idle state power of the hard disk component, U Drd Indicates the read operation ratio of the hard disk component, U Dwr Indicates the write operation ratio of the hard disk component;

[0031] According to the type of the service resource component, a corresponding original power consumption model is selected to obtain the original power corresponding to the service resource component;

[0032] Obtaining the working time corresponding to the service resource component according to the historical operation data of the component;

[0033] The product of the original power and the working time is calculated to obtain the original power consumption of the service resource component.

[0034] In one embodiment, the component historical operation data further includes: historical working data of the heat dissipation component and historical operating temperature data of the service resource component. The heat dissipation efficiency formula is constructed based on the actual power consumption of the service resource component and the server historical data, including:

[0035] Obtaining the original power consumption of the service resource component;

[0036] Based on the actual power consumption of the service resource component, combined with the historical operating temperature data of the service resource component and the historical ambient temperature data of the server, a basic heat dissipation formula of the service resource component is constructed, wherein the basic heat dissipation formula of the service resource component is:

[0037]

[0038] Among them, η component represents the basic heat dissipation efficiency of the service resource component, P real represents the actual power consumption corresponding to the service resource component, T component represents the operating temperature of the service resource component, T env Indicates the server ambient temperature;

[0039] The working efficiency of the heat dissipation component is obtained according to the historical working data of the heat dissipation component, and the ambient temperature correction value is obtained by fitting. The basic heat dissipation formula of the service resource component is corrected by the ambient temperature correction value to obtain a first corrected heat dissipation formula, wherein the first corrected heat dissipation formula is:

[0040]

[0041] in, represents the first modified heat dissipation efficiency of the service resource component, T' env represents the corrected server ambient temperature, where T' env Satisfy the formula: T' env =T env -ΔT env , where ΔT env Indicates that the ambient temperature correction value is positively correlated with the working efficiency of the heat dissipation component;

[0042] Based on the working efficiency of the heat dissipation component, the first corrected heat dissipation formula is corrected again to obtain a second corrected heat dissipation formula, wherein the second corrected heat dissipation formula includes:

[0043]

[0044] in, represents the second modified heat dissipation efficiency of the service resource component, η heat Indicates the working efficiency of the heat dissipation component;

[0045] The heat dissipation correction model is obtained according to the heat dissipation efficiency formula and in combination with the power consumption deviation correction factor, including:

[0046] Fitting is performed according to the historical operation data of the component to obtain the weight of the power consumption deviation factor;

[0047] Based on the second corrected heat dissipation formula, combined with the power consumption deviation factor and the weight of the power consumption deviation factor, the heat dissipation correction model is obtained, wherein the heat dissipation correction model includes the following formula:

[0048]

[0049] Among them, the η component represents the heat dissipation efficiency of the service resource component, ΔP represents the power consumption deviation factor, and k represents the weight of the power consumption deviation factor.

[0050] In one embodiment, based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained, including:

[0051] Based on the heat dissipation correction model, determining several input parameters to be acquired of the service resource component;

[0052] Based on the several input parameters to be obtained of the service resource component, extract and analyze the real-time operation data of the component and the real-time ambient temperature data to obtain several real-time input parameters of the service resource component, wherein the several real-time input parameters correspond one to one to the several input parameters to be obtained;

[0053] The real-time heat dissipation efficiency of the service resource component is obtained by calculating the real-time heat dissipation efficiency of the service resource component through the heat dissipation efficiency formula.

[0054] Based on the type of the service resource component and according to the historical operation data of the component, fitting the heat dissipation correction weight of the service resource component;

[0055] The component heat dissipation correction factor is obtained according to the heat dissipation correction weight and the real-time heat dissipation efficiency of the service resource component.

[0056] In one embodiment, the combined prediction model includes: a primary model and a secondary model, and the combined prediction model is obtained by:

[0057] Obtaining historical operation data of the component;

[0058] According to the historical operation data of the component, the primary model is constructed, wherein the primary model is a convolutional neural network model, including: an input layer, a convolution layer, a pooling layer and a fully connected layer;

[0059] The component historical operation data is converted into a multi-dimensional feature matrix through the input layer to obtain a plurality of input feature graphs;

[0060] Through the convolution layer, convolution operations are performed on the plurality of input feature maps respectively to obtain a plurality of convolution feature maps;

[0061] Through the pooling layer, the plurality of convolutional feature maps are respectively subjected to pooling operations to obtain a plurality of pooling feature maps;

[0062] Through the fully connected layer, the plurality of pooling feature graphs are respectively expanded and fully connected to obtain primary predicted power consumption values ​​of the plurality of components, wherein the primary predicted power consumption vectors of the plurality of components correspond one-to-one to the service resource components and the heat dissipation components respectively;

[0063] Construct a secondary model, wherein the secondary model is a linear regression model, including:

[0064] Obtaining primary predicted power consumption values ​​of several components;

[0065] Obtaining real-time operating data of the components to obtain actual power consumption of several components;

[0066] Pairing a number of the primary predicted power consumption values ​​of the components with a number of the actual power consumptions of the components to obtain a number of training samples to form a secondary model training set;

[0067] A basic linear regression model is constructed, and based on the secondary model training set, the basic linear regression model is improved to obtain an improved linear regression model, which is a combined prediction model.

[0068] In one embodiment, determining a power consumption adjustment method of the server according to the component revised predicted power consumption and the real-time operation data of the server includes:

[0069] In response to obtaining the component revised predicted power consumption of all components in the server, integrating them to obtain a server predicted power consumption of the server;

[0070] Based on a preset server power consumption threshold, judging whether the server power consumption is normal, and in response to the judgment result being normal, no adjustment is required, and in response to the judgment result being abnormal, performing component operation status verification according to the real-time operation data of the server to obtain a component operation status verification result;

[0071] Determine the power consumption adjustment strategy of the server according to the component operation status verification result, wherein the power consumption adjustment strategy includes at least one of the following: performance priority and energy saving priority;

[0072] According to the determined power consumption adjustment strategy, the corresponding operations to be performed by the service resource component and the heat dissipation component are respectively executed.

[0073] On the other hand, a server power consumption management device is provided, the device comprising:

[0074] A data acquisition module is used to obtain real-time data of the server, wherein the real-time data of the server includes: real-time operation data of components and real-time ambient temperature data;

[0075] A heat dissipation factor generation module, used to obtain a component heat dissipation correction factor based on a heat dissipation correction model according to the component real-time operation data and the real-time ambient temperature data;

[0076] A power consumption prediction module, used for obtaining a normal predicted power consumption of a component based on a combined prediction model and real-time operation data of the component;

[0077] A power consumption correction module, used to correct the normal predicted power consumption of the component according to the heat dissipation correction factor of the component to obtain a corrected predicted power consumption of the component;

[0078] The power consumption adjustment decision module is used to determine the power consumption adjustment mode of the server according to the component correction predicted power consumption and the real-time operation data of the server.

[0079] In another aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented:

[0080] Acquire server real-time data, wherein the server real-time data includes component real-time operation data and real-time ambient temperature data;

[0081] Based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained;

[0082] Based on the combined prediction model and the real-time operation data of the component, a normal predicted power consumption of the component is obtained;

[0083] According to the component heat dissipation correction factor, the normal predicted power consumption of the component is corrected to obtain the component corrected predicted power consumption;

[0084] A power consumption adjustment method of the server is determined according to the component revised predicted power consumption and the real-time operation data of the server.

[0085] In another aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0086] Acquire server real-time data, wherein the server real-time data includes component real-time operation data and real-time ambient temperature data;

[0087] Based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained;

[0088] Based on the combined prediction model and the real-time operation data of the component, a normal predicted power consumption of the component is obtained;

[0089] According to the component heat dissipation correction factor, the normal predicted power consumption of the component is corrected to obtain the component corrected predicted power consumption;

[0090] A power consumption adjustment method of the server is determined according to the component revised predicted power consumption and the real-time operation data of the server.

[0091] The above-mentioned server power consumption management method, device, computer equipment and storage medium predict normal power consumption through a combined prediction model, and reflect the actual working status of the current server based on real-time data. Compared with the single model and static prediction of the traditional method, it can combine the advantages of multiple models, reduce errors, and thus improve the accuracy of power consumption prediction; at the same time, through the heat dissipation correction model, the heat dissipation correction factor is introduced, which can not only introduce the influence of temperature change on power consumption based on the ambient temperature, but also consider the synergy between the service resource component and the heat dissipation component, increase the comprehensiveness of power consumption prediction, make it more in line with the actual situation, and help to improve the efficiency and accuracy of subsequent power consumption management; in addition, the model has been constructed in advance and can be directly called when used, which speeds up the efficiency of power consumption prediction, and selects an appropriate power consumption adjustment strategy based on the predicted power consumption obtained by the heat dissipation correction model and the combined prediction model, which can more accurately adjust the server power consumption, ensuring the efficient operation of the server while improving energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1A diagram of an application environment of a method for managing power consumption of a server in an embodiment;

[0093] Figure 2 A schematic diagram of a process flow of a method for managing power consumption of a server in one embodiment;

[0094] Figure 3 A structural block diagram of a server power consumption management device in one embodiment;

[0095] Figure 4 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0096] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0097] The present application provides a method for managing server power consumption, which can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The server 104 has a service resource component and a heat dissipation component. When the server 104 receives the power consumption management request sent by the terminal 102, the heat dissipation correction model and the combined prediction model are called to predict the server power consumption according to the real-time data of the server, and the corresponding power consumption adjustment method is selected based on the predicted power consumption to adjust, and the adjustment result is returned to the terminal 102. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server 104 can be implemented with an independent server or a server cluster composed of multiple servers.

[0098] In one embodiment, Figure 2 As shown, a server power consumption management method is provided, and the method is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0099] Step 201, obtaining server real-time data, wherein the server real-time data includes: component real-time operation data and real-time ambient temperature data.

[0100] Specifically, the service resource component and the heat dissipation component are monitored in real time to obtain real-time data of the service resource component and real-time data of the heat dissipation component; the server ambient temperature is monitored in real time to obtain real-time ambient temperature data.

[0101] Step 202 , based on a heat dissipation correction model, obtain a component heat dissipation correction factor according to the component real-time operation data and the real-time ambient temperature data.

[0102] Specifically, according to the type of the service resource component, the original power consumption of the service resource component is obtained; the historical data of the server is obtained to obtain the actual power consumption of the service resource component; according to the original power consumption of the service resource component, combined with the historical data of the server, a heat dissipation efficiency formula is constructed; according to the heat dissipation efficiency formula, correction is made in combination with the original power consumption of the service resource component to obtain a heat dissipation correction model, wherein the heat dissipation efficiency formula is derived from the basic heat dissipation formula of the service resource component, the first corrected heat dissipation formula and the second corrected heat dissipation formula; based on the heat dissipation correction model, several input parameters to be obtained of the service resource component are obtained; through the heat dissipation efficiency formula, several input parameters to be obtained of the service resource component are calculated to obtain the real-time heat dissipation efficiency of the service resource component; according to the real-time heat dissipation efficiency of the service resource component and combined with the type of the service resource component, a component heat dissipation correction factor is obtained.

[0103] Step 203: obtaining a normal predicted power consumption of the component based on the combined prediction model and the real-time operation data of the component.

[0104] Specifically, the combined prediction model includes: a primary model and a secondary model, and a combined prediction model is obtained; component historical operation data is obtained; a primary model is constructed according to the component historical operation data, wherein the primary model is a convolutional neural network model and the secondary model is a linear regression model; according to the component historical operation data, primary predicted power consumption values ​​of several components are obtained through the primary model; according to the primary predicted power consumption values ​​of several components, the constructed linear regression formula is corrected through the secondary model to obtain a corrected linear regression formula, i.e., a component normal power consumption prediction formula; based on the combined prediction model, several input parameters of the service resource component and the heat dissipation component are obtained; through the component normal power consumption prediction formula, the normal predicted power consumption of the service resource component and the heat dissipation component are respectively calculated.

[0105] Step 204: correct the normal predicted power consumption of the component according to the component heat dissipation correction factor to obtain the component corrected predicted power consumption.

[0106] Specifically, the normal predicted power consumption of the component includes: the normal predicted power consumption of the service resource component and the normal predicted power consumption of the heat dissipation component; the corrected predicted power consumption of the component includes: the corrected predicted power consumption of the service resource component and the corrected predicted power consumption of the heat dissipation component; according to the component heat dissipation correction factor, the normal predicted power consumption of the service resource component is corrected to obtain the corrected predicted power consumption of the service resource component; based on the corrected predicted power consumption of the service resource component and combined with the component heat dissipation correction factor, the normal predicted power consumption of the heat dissipation component is corrected to obtain the corrected predicted power consumption of the heat dissipation component.

[0107] Step 205 : determining a power consumption adjustment method of the server according to the component revised predicted power consumption and the real-time operation data of the server.

[0108] Specifically, the corrected predicted power consumption of the service resource component and the corrected predicted power consumption of the heat dissipation component are obtained, integrated, and the server predicted power consumption of the server is obtained; based on a preset server power consumption threshold, whether the server power consumption is normal is judged, and in response to the judgment result being normal, no adjustment is required, and in response to the judgment result being abnormal, based on the real-time operation data of the server, the component operation status is checked based on the corrected predicted power consumption of the service resource component and the corrected predicted power consumption of the heat dissipation component, and the component operation status check result is obtained; based on the component operation status check result, the power consumption adjustment strategy of the server is determined, wherein the power consumption adjustment strategy includes at least one of the following: performance priority and energy saving priority; according to the determined power consumption adjustment strategy, the corresponding operations to be executed of the service resource component and the heat dissipation component are respectively executed.

[0109] In the above-mentioned server power consumption management method, the normal power consumption is predicted by a combined prediction model, and the actual working status of the current server is reflected based on real-time data. Compared with the single model and static prediction of the traditional method, it can combine the advantages of multiple models, reduce errors, and thus improve the accuracy of power consumption prediction; at the same time, through the heat dissipation correction model, the heat dissipation correction factor is introduced, which can not only introduce the impact of temperature changes on power consumption based on ambient temperature, but also consider the synergy between service resource components and heat dissipation components, increase the comprehensiveness of power consumption prediction, make it more in line with actual conditions, and help improve the efficiency and accuracy of subsequent power consumption management; in addition, the model has been constructed in advance and can be directly called when used, which speeds up the efficiency of power consumption prediction, and selects an appropriate power consumption adjustment strategy based on the predicted power consumption obtained by the heat dissipation correction model and the combined prediction model, which can more accurately adjust the server power consumption, ensuring efficient operation of the server while improving energy efficiency.

[0110] In one embodiment, the server includes a service resource component and a heat dissipation component, and the heat dissipation correction model is obtained in the following manner:

[0111] Acquire server historical data, wherein the server historical data includes: component historical operation data and server historical ambient temperature data;

[0112] Construct an original power consumption model corresponding to the service resource component, and calculate the original power consumption of the service resource component;

[0113] According to the historical operation data of the components, the actual power consumption of the service resource components is obtained;

[0114] Calculate the difference between the original power consumption and the actual power consumption of the service resource component to obtain the power consumption deviation correction factor;

[0115] Based on the original power consumption of service resource components and combined with historical server data, a heat dissipation efficiency formula is constructed;

[0116] According to the heat dissipation efficiency formula and the power consumption deviation correction factor, a heat dissipation correction model is obtained.

[0117] Specifically, by constructing a heat dissipation correction model based on historical data, this embodiment can comprehensively consider the interaction between the operating temperature of the service resource component, the ambient temperature, the efficiency of the heat dissipation component and the power consumption, and obtain the complex correlation between the components associated with the ambient temperature, reflecting how the temperature change of the service resource component affects its power consumption, and considering the impact of the heat dissipation component efficiency on the ambient temperature, thereby reflecting the indirect impact of the heat dissipation component working efficiency on the actual power consumption of the component. The introduced heat dissipation correction model is conducive to improving the accuracy of power consumption prediction in the future, ensuring that the power consumption management strategy is more accurate and effective.

[0118] In one embodiment, the service resource component includes at least one of the following: a computing element, a memory element, a network element, and a hard disk element. The original power consumption model corresponding to the service resource component is constructed, and the original power consumption of the service resource component is calculated, including:

[0119] An original computing power consumption model of the computing element is constructed, wherein the original computing power consumption model includes the following formula:

[0120] P CPU =(P max -P idle )·U CPU +P idle

[0121] Among them, P CPU represents the raw power of the computing element, P max represents the maximum power of the computing element, P idle represents the idle power of the computing element, U CPU It indicates the utilization of computing elements, that is, the percentage of computing elements actually used in a given time;

[0122] Construct an original memory power consumption model of the memory element, wherein the original memory power consumption model includes the following formula:

[0123] P memory =P Mrd ·U Mrd +P Mwr ·U Mwr

[0124] Among them, P memory Represents the raw power of the memory element, P Mrd Represents the read state power of the memory element, P Mwr represents the write power of the memory element, U Mrd Indicates the proportion of read operations of memory components, U Mwr Indicates the write operation ratio of the memory element;

[0125] An original network power consumption model of the network element is constructed, wherein the original network power consumption model includes the following formula:

[0126]

[0127] Among them, P net is the raw power of the network element, B represents the bandwidth of the network element, ρ represents the bandwidth utilization of the network element, and τ represents the power efficiency of the network element;

[0128] An original hard disk power consumption model of the hard disk component is constructed, wherein the original hard disk power consumption model includes the following formula:

[0129] P disk =P Drd ·U Drd +P Dwr ·U Dwr +P Didle ·(1-U Drd -U Dwr )

[0130] Among them, P disk Represents the raw power of the hard disk component, P Drd Indicates the read state power of the hard disk component, P Dwr Indicates the write power of the hard disk component, P Didle Indicates the idle power of the hard disk component, U Drd Indicates the read operation ratio of the hard disk component, U Dwr Indicates the write operation ratio of the hard disk component;

[0131] According to the type of the service resource component, a corresponding original power consumption model is selected to obtain the original power corresponding to the service resource component;

[0132] According to the historical operation data of the component, the corresponding working time of the service resource component is obtained;

[0133] The product of the original power and the working time is calculated to obtain the original power consumption of the service resource component.

[0134] Specifically, in this embodiment, by constructing original power consumption models suitable for different service resource components, such as computing elements, memory elements, network elements and hard disk elements, the power consumption of each component under different working conditions can be accurately calculated, providing basic support for the introduction of the heat dissipation correction model, thereby ensuring that the power consumption of components under different load conditions can be accurately predicted, and the normal power consumption prediction can be corrected in combination with the heat dissipation correction factor, making power consumption management more accurate.

[0135] In one embodiment, the component historical operation data also includes: historical working data of the heat dissipation component and historical operating temperature data of the service resource component. According to the actual power consumption of the service resource component and the historical data of the server, a heat dissipation efficiency formula is constructed, including:

[0136] Get the raw power consumption of the service resource component;

[0137] Based on the actual power consumption of the service resource component, combined with the historical operating temperature data of the service resource component and the historical ambient temperature data of the server, a basic heat dissipation formula for the service resource component is constructed. The basic heat dissipation formula for the service resource component is:

[0138]

[0139] Among them, η component represents the basic heat dissipation efficiency of the service resource component, P real Indicates the actual power consumption corresponding to the service resource component, T component Represents the operating temperature of the service resource component, T env Indicates the server ambient temperature;

[0140] According to the historical working data of the heat dissipation component, the working efficiency of the heat dissipation component is obtained, and the ambient temperature correction value is obtained by fitting. The basic heat dissipation formula of the service resource component is corrected by the ambient temperature correction value to obtain the first corrected heat dissipation formula, where the first corrected heat dissipation formula is:

[0141]

[0142] in, represents the first modified heat dissipation efficiency of the service resource component, T' env represents the corrected server ambient temperature, where T' env Satisfy the formula: T' env =T env -ΔT env , where ΔT env Indicates the ambient temperature correction value, which is positively correlated with the working efficiency of the heat dissipation component;

[0143] Based on the working efficiency of the heat dissipation component, the first corrected heat dissipation formula is corrected again to obtain a second corrected heat dissipation formula, wherein the second corrected heat dissipation formula includes:

[0144]

[0145] in, represents the second modified heat dissipation efficiency of the service resource component, η heat Indicates the working efficiency of the heat dissipation component;

[0146] According to the heat dissipation efficiency formula and combined with the power consumption deviation correction factor, the heat dissipation correction model is obtained, including:

[0147] According to the historical operation data of the component, fitting is performed to obtain the weight of the power consumption deviation factor;

[0148] Based on the second corrected heat dissipation formula, combined with the power consumption deviation factor and the weight of the power consumption deviation factor, a heat dissipation correction model is obtained, wherein the heat dissipation correction model includes the following formula:

[0149]

[0150] Among them, η component represents the heat dissipation efficiency of the service resource component, ΔP represents the power consumption deviation factor, and k represents the weight of the power consumption deviation factor.

[0151] It is worth noting that the service resource component will cause the temperature to rise during operation, and its own temperature interacts with the ambient temperature, which in turn affects the power consumption. The power consumption and efficiency of the heat dissipation component will affect the ambient temperature, which in turn indirectly affects the temperature and power consumption of the service resource component. The power consumption of the service resource component, the operating temperature of the service resource component, the server ambient temperature, the power consumption of the heat dissipation component, and the working efficiency of the heat dissipation component, even in this case, have complex correlations and mutual influences. Therefore, they need to be considered and corrected, which is also the reason for introducing the heat dissipation correction model.

[0152] Specifically, the method in this embodiment constructs a heat dissipation efficiency formula, combines the actual power consumption, operating temperature and ambient temperature of the service resource components, etc., to accurately evaluate the heat dissipation effect, and dynamically corrects the ambient temperature and operating temperature according to the working efficiency of the heat dissipation components; at the same time, the heat dissipation efficiency formula constructed in this embodiment can take into account the impact of temperature fluctuations on the heat dissipation effect, and combine the power consumption deviation factor obtained from the actual power consumption and the original power consumption, so that the prediction process is more in line with the real environment, thereby optimizing heat dissipation management and power consumption prediction, reducing energy waste, and improving the overall energy efficiency of the server system.

[0153] In one embodiment, based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained, including:

[0154] Based on the heat dissipation correction model, determine several input parameters to be obtained of the service resource component;

[0155] Based on the input parameters to be obtained of the service resource component, the real-time operation data of the component and the real-time ambient temperature data are extracted and analyzed to obtain the real-time input parameters of the service resource component, wherein the real-time input parameters correspond one to one to the input parameters to be obtained;

[0156] Through the heat dissipation efficiency formula, several real-time input parameters of the service resource component are calculated to obtain the real-time heat dissipation efficiency of the service resource component.

[0157] Based on the type of service resource component and the historical operation data of the component, the heat dissipation correction weight of the service resource component is obtained by fitting;

[0158] According to the heat dissipation correction weight and real-time heat dissipation efficiency of the service resource component, the component heat dissipation correction factor is obtained.

[0159] Specifically, the heat dissipation correction model is used in this embodiment to dynamically calculate the heat dissipation correction factor through real-time component operation data and ambient temperature data, which can correct the normal power consumption prediction output by the combined prediction model and make up for the fact that the combined prediction model does not consider the complex interactions between service resource components and the influence of ambient temperature, so that the power consumption prediction can be more in line with the actual situation, thereby achieving more accurate power consumption management and optimizing server energy efficiency.

[0160] In one embodiment, the combined prediction model includes: a primary model and a secondary model, and the combined prediction model is obtained in the following manner:

[0161] Get component historical operation data;

[0162] According to the historical operation data of the component, a primary model is constructed, where the primary model is a convolutional neural network model, including: input layer, convolution layer, pooling layer and fully connected layer;

[0163] Through the input layer, the component historical operation data is converted into a multi-dimensional feature matrix to obtain several input feature graphs;

[0164] Through the convolution layer, several input feature maps are convolved to obtain several convolution feature maps;

[0165] Through the pooling layer, several convolution feature maps are pooled respectively to obtain several pooling feature maps;

[0166] Through the fully connected layer, several pooled feature maps are expanded and fully connected respectively to obtain the primary predicted power consumption values ​​of several components, wherein the primary predicted power consumption vectors of several components correspond one-to-one to the service resource components and the heat dissipation components respectively;

[0167] Construct a secondary model, where the secondary model is a linear regression model, including:

[0168] Obtaining primary predicted power consumption values ​​of several components;

[0169] Obtain real-time operation data of components and obtain the actual power consumption of several components;

[0170] Pairing the primary predicted power consumption values ​​of several components with the actual power consumption of several components to obtain several training samples, which constitute a secondary model training set;

[0171] A basic linear regression model is constructed, and based on the secondary model training set, the basic linear regression model is improved to obtain an improved linear regression model, which is the combined prediction model.

[0172] It is worth noting that the basic linear regression model includes: basic linear regression formula, building a basic linear regression model, including:

[0173] Construct a basic linear regression formula, where the basic linear regression formula is:

[0174] y rep =β0+β1x

[0175] Among them, y rep represents the secondary predicted power consumption of the component, x represents the primary predicted power consumption value of the component, β0 represents the bias parameter, and β1 represents the regression coefficient;

[0176] The improved linear regression model includes: an improved linear regression formula, that is, a combined normal power consumption prediction formula. Based on the secondary model training set, the basic linear regression model is improved to obtain an improved linear regression model, that is, a combined prediction model, including:

[0177] Based on several training samples, a model loss function is constructed, where the model loss function is:

[0178]

[0179] Where Q represents the predicted deviation of power consumption, y i represents the actual power consumption of the component of the i-th training sample, represents the secondary predicted power consumption of the component of the i-th training sample, n represents the number of training samples, x i represents the primary predicted power consumption value of the component of the i-th training sample;

[0180] According to the partial differential equation, the bias parameter and regression coefficient are solved to obtain the corrected bias parameter and the corrected regression coefficient. The partial derivatives of the bias parameter β0 and the regression coefficient β1 are calculated respectively, and the corrected bias parameter is calculated based on the extreme point being 0. and the corrected regression coefficient

[0181] Based on the corrected bias parameters and the corrected regression coefficients, the basic linear regression formula is improved to obtain the improved linear regression formula, that is, the normal power consumption prediction formula of the component, which is:

[0182]

[0183] in, Predict the power consumption of the component under normal conditions, x * The primary predicted power consumption value output by the primary model, wherein the data input into the primary model is the real-time operation data of the component during the actual prediction;

[0184] Based on the improved linear regression formula, a combined prediction model is obtained.

[0185] Specifically, in this embodiment, a combined prediction model is constructed by combining a primary convolutional neural network (CNN) model with a secondary linear regression model, which can accurately predict the normal power consumption of each component. The primary model extracts deep features based on the convolutional neural network, can capture the operating mode of the service resource component, and extract deep features related to the power consumption impact. The secondary model corrects the prediction results of the primary model through linear regression, and can quickly capture and correct the characteristic relationship between the component power consumption and other features. The combination of the two can not only improve the prediction accuracy, but also improve the model interpretability, which is conducive to the subsequent analysis of the influence of each feature on power consumption, thereby improving the accuracy of subsequent power consumption adjustment.

[0186] In one embodiment, determining a power consumption adjustment method of a server according to component correction predicted power consumption and real-time operation data of the server includes:

[0187] In response to obtaining component modified predicted power consumptions of all components in the server, integrating them to obtain server predicted power consumption of the server;

[0188] Based on the preset server power consumption threshold, determine whether the server power consumption is normal. If the judgment result is normal, no adjustment is required. If the judgment result is abnormal, perform component operation status verification according to the real-time operation data of the server to obtain the component operation status verification result.

[0189] Determine the power consumption adjustment strategy of the server according to the component operation status verification result, wherein the power consumption adjustment strategy includes at least one of the following: performance priority, energy saving priority;

[0190] According to the determined power consumption adjustment strategy, the operations to be performed of the corresponding service resource components and heat dissipation components are respectively executed;

[0191] It is worth noting that, under the performance-first power consumption adjustment strategy, the pending operations of the service resource component include: increasing the main frequency of the computing element to improve task processing capabilities, dynamically allocating more memory to high-priority tasks, the hard disk component to speed up disk access, and the network component to allocate bandwidth to critical tasks. The pending operations of the heat dissipation component include: improving the working efficiency of the heat dissipation component; under the energy-saving priority power consumption adjustment strategy, the pending operations of the service resource component include: lowering the main frequency of the computing element or enabling dynamic frequency adjustment, enabling memory compression or low power consumption mode for the memory component, enabling the energy-saving mode of the hard disk component, and reducing the power usage mode of the network interface for the network component. The pending operations of the heat dissipation component include: reducing the working efficiency of the heat dissipation component while keeping the server ambient temperature within a safe range.

[0192] Specifically, this embodiment determines whether the power consumption management strategy needs to be adjusted based on the corrected predicted power consumption and the real-time operation data of the server. It can determine the appropriate power consumption adjustment method based on the results of different power consumption predictions and select performance priority or energy saving priority. Moreover, by checking the operating status of the components, it can select the appropriate adjustment strategy in the event of abnormal power consumption, ensuring that the system runs efficiently under different loads and ambient temperatures, while achieving energy saving goals.

[0193] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0194] In one embodiment, Figure 3 As shown, a server power consumption management device is provided, the device comprising: a data acquisition module, a heat dissipation factor generation module, a power consumption prediction module, a power consumption correction module and a power consumption adjustment decision module, wherein:

[0195] A data acquisition module is used to obtain real-time data of the server, wherein the real-time data of the server includes: real-time operation data of components and real-time ambient temperature data;

[0196] A heat dissipation factor generation module, used to obtain a component heat dissipation correction factor based on a heat dissipation correction model according to the component real-time operation data and the real-time ambient temperature data;

[0197] A power consumption prediction module, used for obtaining a normal predicted power consumption of a component based on a combined prediction model and real-time operation data of the component;

[0198] A power consumption correction module, used to correct the normal predicted power consumption of the component according to the heat dissipation correction factor of the component to obtain a corrected predicted power consumption of the component;

[0199] The power consumption adjustment decision module is used to determine the power consumption adjustment mode of the server according to the component correction predicted power consumption and the real-time operation data of the server.

[0200] The data acquisition module is used to monitor the service resource component and the heat dissipation component in real time to obtain the real-time data of the service resource component and the heat dissipation component; and to monitor the server ambient temperature in real time to obtain the real-time ambient temperature data.

[0201] The heat dissipation factor generation module is used to obtain server historical data, wherein the server historical data includes: component historical operation data and server historical ambient temperature data; construct an original power consumption model corresponding to a service resource component, and calculate the original power consumption of the service resource component; obtain the actual power consumption of the service resource component according to the component historical operation data; calculate the difference between the original power consumption and the actual power consumption of the service resource component to obtain the power consumption deviation correction factor; construct a heat dissipation efficiency formula according to the original power consumption of the service resource component in combination with the server historical data; and obtain a heat dissipation correction model according to the heat dissipation efficiency formula in combination with the power consumption deviation correction factor.

[0202] The heat dissipation factor generation module is also used to determine several input parameters to be obtained of the service resource component based on the heat dissipation correction model; based on the several input parameters to be obtained of the service resource component, extract and analyze the real-time operation data and real-time ambient temperature data of the component to obtain several real-time input parameters of the service resource component, wherein the several real-time input parameters correspond one-to-one to the several input parameters to be obtained; calculate the several real-time input parameters of the service resource component through the heat dissipation efficiency formula to obtain the real-time heat dissipation efficiency of the service resource component, and based on the type of the service resource component and according to the historical operation data of the component, fit the heat dissipation correction weight of the service resource component; and obtain the component heat dissipation correction factor according to the heat dissipation correction weight and the real-time heat dissipation efficiency of the service resource component.

[0203] The power consumption prediction module is used to obtain component historical operation data; construct a primary model based on the component historical operation data, wherein the primary model is a convolutional neural network model, including: an input layer, a convolution layer, a pooling layer and a fully connected layer; through the input layer, the component historical operation data is converted into a multi-dimensional feature matrix to obtain a number of input feature maps; through the convolution layer, a number of input feature maps are convoluted to obtain a number of convolution feature maps; through the pooling layer, a number of convolution feature maps are pooled to obtain a number of pooling feature maps; through the fully connected layer, a number of pooling feature maps are expanded and fully connected to obtain a number of component primary predicted power consumption values, wherein the primary predicted power consumption vectors of the components are respectively related to the service The service resource components and the heat dissipation components correspond one to one; a secondary model is constructed, wherein the secondary model is a linear regression model, including: obtaining primary predicted power consumption values ​​of several components; obtaining real-time operation data of components to obtain actual power consumption of several components; pairing the primary predicted power consumption values ​​of several components with the actual power consumption of several components to obtain several training samples to form a secondary model training set; constructing a linear regression formula; based on the several training samples, constructing a model loss function; according to the partial differential equation, solving the bias parameter and the regression coefficient to obtain the corrected bias parameter and the corrected regression coefficient; based on the corrected bias parameter and the corrected regression coefficient, the linear regression formula is corrected to obtain the corrected linear regression formula, that is, the normal power consumption prediction formula of the component.

[0204] The power consumption prediction module is also used to obtain several input parameters of the service resource component and the heat dissipation component based on the combined prediction model; and to calculate the normal predicted power consumption of the service resource component and the heat dissipation component respectively through the component normal power consumption prediction formula.

[0205] The power consumption correction module is used to correct the normal predicted power consumption of the service resource component according to the component heat dissipation correction factor to obtain the corrected predicted power consumption of the service resource component; based on the corrected predicted power consumption of the service resource component and combined with the component heat dissipation correction factor, the normal predicted power consumption of the heat dissipation component is corrected to obtain the corrected predicted power consumption of the heat dissipation component.

[0206] The power consumption adjustment decision module is used to respond to the component corrected predicted power consumption of all components in the server, integrate them, and obtain the server predicted power consumption of the server; based on the preset server power consumption threshold, judge whether the server power consumption is normal, and in response to the judgment result being normal, no adjustment is required; in response to the judgment result being abnormal, perform component operation status verification based on the real-time operation data of the server to obtain the component operation status verification result; determine the power consumption adjustment strategy of the server based on the component operation status verification result, wherein the power consumption adjustment strategy includes at least one of the following: performance priority, energy saving priority; according to the determined power consumption adjustment strategy, respectively execute the corresponding service resource component and the to-be-executed operations of the heat dissipation component.

[0207] For the specific definition of the server power consumption management device, please refer to the definition of the server power consumption management method above, which will not be repeated here. Each module in the above-mentioned server power consumption management device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0208] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store server power consumption management data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a server power consumption management method is implemented.

[0209] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0210] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0211] Acquire server real-time data, wherein the server real-time data includes component real-time operation data and real-time ambient temperature data;

[0212] Based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained;

[0213] Based on the combined prediction model and the real-time operation data of the component, a normal predicted power consumption of the component is obtained;

[0214] According to the component heat dissipation correction factor, the normal predicted power consumption of the component is corrected to obtain the component corrected predicted power consumption;

[0215] A power consumption adjustment method of the server is determined according to the component revised predicted power consumption and the real-time operation data of the server.

[0216] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0217] Acquire server real-time data, wherein the server real-time data includes component real-time operation data and real-time ambient temperature data;

[0218] Based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained;

[0219] Based on the combined prediction model and the real-time operation data of the component, a normal predicted power consumption of the component is obtained;

[0220] According to the component heat dissipation correction factor, the normal predicted power consumption of the component is corrected to obtain the component corrected predicted power consumption;

[0221] A power consumption adjustment method of the server is determined according to the component revised predicted power consumption and the real-time operation data of the server.

[0222] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0223] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0224] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A server power consumption management method, characterized in that: The method comprises: Acquire server real-time data, wherein the server real-time data includes component real-time operation data and real-time ambient temperature data; Based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained; Based on the combined prediction model and the real-time operation data of the component, a normal predicted power consumption of the component is obtained; According to the component heat dissipation correction factor, the normal predicted power consumption of the component is corrected to obtain the component corrected predicted power consumption; A power consumption adjustment method of the server is determined according to the component revised predicted power consumption and the real-time operation data of the server.

2. A server power consumption management method according to claim 1, characterized in that: The server includes a service resource component and a heat dissipation component, and the heat dissipation correction model is obtained in the following manner: Acquire server historical data, wherein the server historical data includes: component historical operation data and server historical ambient temperature data; Constructing an original power consumption model corresponding to the service resource component, and calculating the original power consumption of the service resource component; Obtaining actual power consumption of the service resource component according to the historical operation data of the component; Calculating the difference between the original power consumption and the actual power consumption of the service resource component to obtain a power consumption deviation correction factor; Constructing a heat dissipation efficiency formula based on the original power consumption of the service resource component and the historical data of the server; According to the heat dissipation efficiency formula and in combination with the power consumption deviation correction factor, a heat dissipation correction model is obtained.

3. A server power consumption management method according to claim 2, characterized in that: The service resource component includes at least one of the following: a computing element, a memory element, a network element, and a hard disk element. The constructing of the original power consumption model corresponding to the service resource component and calculating the original power consumption of the service resource component include: Constructing an original computing power consumption model of the computing element, wherein the original computing power consumption model includes the following formula: P CPU =(P max -P idle )·U CPU +P idle Among them, P CPU represents the raw power of the computing element, P max represents the maximum power of the computing element, P idle represents the idle power of the computing element, U CPU represents the utilization rate of the computing element, that is, the percentage of the computing element actually used in a given time; Construct an original memory power consumption model of the memory element, wherein the original memory power consumption model includes the following formula: Q memory =P Mrd ·U Mrd +P Mwr ·U Mwr Among them, P memory represents the raw power of the memory element, P Mrd represents the read state power of the memory element, P Mwr represents the write state power of the memory element, U Mrd represents the read operation ratio of the memory element, U Mwr Indicates the write operation ratio of the memory element; Constructing an original network power consumption model of the network element, wherein the original network power consumption model includes the following formula: Among them, P net is the original power of the network element, B represents the bandwidth of the network element, ρ represents the bandwidth utilization of the network element, and τ represents the power efficiency of the network element; Construct an original hard disk power consumption model of the hard disk element, wherein the original hard disk power consumption model includes the following formula: P disk =P Drd ·IN Drd +P Dwr ·IN Dwr +P Didle ·(1-U Drd -IN Dwr ) Among them, P disk represents the raw power of the hard disk element, P Drd represents the read state power of the hard disk component, P Dwr represents the write state power of the hard disk element, P Didle represents the idle state power of the hard disk component, U Drd Indicates the read operation ratio of the hard disk component, U Dwr Indicates the write operation ratio of the hard disk component; According to the type of the service resource component, a corresponding original power consumption model is selected to obtain the original power corresponding to the service resource component; Obtaining the working time corresponding to the service resource component according to the historical operation data of the component; The product of the original power and the working time is calculated to obtain the original power consumption of the service resource component.

4. A server power consumption management method according to claim 2, characterized in that: The component historical operation data also includes: historical working data of the heat dissipation component and historical operating temperature data of the service resource component. The heat dissipation efficiency formula is constructed based on the actual power consumption of the service resource component and the server historical data, including: Obtaining the original power consumption of the service resource component; Based on the actual power consumption of the service resource component, combined with the historical operating temperature data of the service resource component and the historical ambient temperature data of the server, a basic heat dissipation formula of the service resource component is constructed, wherein the basic heat dissipation formula of the service resource component is: Among them, η component represents the basic heat dissipation efficiency of the service resource component, P real represents the actual power consumption corresponding to the service resource component, T component represents the operating temperature of the service resource component, T env Indicates the server ambient temperature; The working efficiency of the heat dissipation component is obtained according to the historical working data of the heat dissipation component, and the ambient temperature correction value is obtained by fitting. The basic heat dissipation formula of the service resource component is corrected by the ambient temperature correction value to obtain a first corrected heat dissipation formula, wherein the first corrected heat dissipation formula is: in, represents the first modified heat dissipation efficiency of the service resource component, T' env represents the corrected server ambient temperature, where T' env Satisfy the formula: T' env =T env -ΔT env , where ΔT env Indicates that the ambient temperature correction value is positively correlated with the working efficiency of the heat dissipation component; Based on the working efficiency of the heat dissipation component, the first corrected heat dissipation formula is corrected again to obtain a second corrected heat dissipation formula, wherein the second corrected heat dissipation formula includes: in, represents the second modified heat dissipation efficiency of the service resource component, η heat Indicates the working efficiency of the heat dissipation component; The heat dissipation correction model is obtained according to the heat dissipation efficiency formula and in combination with the power consumption deviation correction factor, including: Fitting is performed according to the historical operation data of the component to obtain the weight of the power consumption deviation factor; Based on the second corrected heat dissipation formula, combined with the power consumption deviation factor and the weight of the power consumption deviation factor, the heat dissipation correction model is obtained, wherein the heat dissipation correction model includes the following formula: Among them, the η component represents the heat dissipation efficiency of the service resource component, ΔP represents the power consumption deviation factor, and k represents the weight of the power consumption deviation factor.

5. A server power consumption management method according to claim 2, characterized in that: Based on the heat dissipation correction model, according to the real-time operation data of the component and the real-time ambient temperature data, a component heat dissipation correction factor is obtained, including: Based on the heat dissipation correction model, determining several input parameters to be acquired of the service resource component; Based on the several input parameters to be obtained of the service resource component, extract and analyze the real-time operation data of the component and the real-time ambient temperature data to obtain several real-time input parameters of the service resource component, wherein the several real-time input parameters correspond one to one to the several input parameters to be obtained; The real-time heat dissipation efficiency of the service resource component is obtained by calculating the real-time heat dissipation efficiency of the service resource component through the heat dissipation efficiency formula. Based on the type of the service resource component and according to the historical operation data of the component, fitting the heat dissipation correction weight of the service resource component; The component heat dissipation correction factor is obtained according to the heat dissipation correction weight and the real-time heat dissipation efficiency of the service resource component.

6. A method for managing server power consumption according to claim 2, characterized in that: The combined prediction model includes: a primary model and a secondary model, and the combined prediction model is obtained in the following manner: Obtaining historical operation data of the component; According to the historical operation data of the component, the primary model is constructed, wherein the primary model is a convolutional neural network model, including: an input layer, a convolution layer, a pooling layer and a fully connected layer; The component historical operation data is converted into a multi-dimensional feature matrix through the input layer to obtain a plurality of input feature graphs; Through the convolution layer, convolution operations are performed on the plurality of input feature maps respectively to obtain a plurality of convolution feature maps; Through the pooling layer, the plurality of convolutional feature maps are respectively subjected to pooling operations to obtain a plurality of pooling feature maps; Through the fully connected layer, the plurality of pooling feature graphs are respectively expanded and fully connected to obtain primary predicted power consumption values ​​of the plurality of components, wherein the primary predicted power consumption vectors of the plurality of components correspond one-to-one to the service resource components and the heat dissipation components respectively; Construct a secondary model, wherein the secondary model is a linear regression model, including: Obtaining primary predicted power consumption values ​​of several components; Obtaining real-time operating data of the components to obtain actual power consumption of several components; Pairing a number of the primary predicted power consumption values ​​of the components with a number of the actual power consumptions of the components to obtain a number of training samples to form a secondary model training set; A basic linear regression model is constructed, and based on the secondary model training set, the basic linear regression model is improved to obtain an improved linear regression model, which is a combined prediction model.

7. A server power consumption management method according to claim 2, characterized in that: The determining the power consumption adjustment mode of the server according to the component corrected predicted power consumption and the real-time operation data of the server includes: In response to obtaining the component revised predicted power consumption of all components in the server, integrating them to obtain a server predicted power consumption of the server; Based on a preset server power consumption threshold, judging whether the server power consumption is normal, and in response to the judgment result being normal, no adjustment is required, and in response to the judgment result being abnormal, performing component operation status verification according to the real-time operation data of the server to obtain a component operation status verification result; Determine the power consumption adjustment strategy of the server according to the component operation status verification result, wherein the power consumption adjustment strategy includes at least one of the following: performance priority and energy saving priority; According to the determined power consumption adjustment strategy, the corresponding operations to be performed by the service resource component and the heat dissipation component are respectively executed.

8. A server power consumption management device, characterized in that: The device comprises: A data acquisition module is used to obtain real-time data of the server, wherein the real-time data of the server includes: real-time operation data of components and real-time ambient temperature data; A heat dissipation factor generation module, used to obtain a component heat dissipation correction factor based on a heat dissipation correction model according to the component real-time operation data and the real-time ambient temperature data; A power consumption prediction module, used for obtaining a normal predicted power consumption of a component based on a combined prediction model and real-time operation data of the component; A power consumption correction module, used to correct the normal predicted power consumption of the component according to the heat dissipation correction factor of the component to obtain a corrected predicted power consumption of the component; The power consumption adjustment decision module is used to determine the power consumption adjustment mode of the server according to the component correction predicted power consumption and the real-time operation data of the server.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images