Method and device for executing heat dissipation operation, computer equipment and storage medium

Abstract

The invention relates to the technical field of servers, and discloses a method and device for executing heat dissipation operation, computer equipment and a storage medium, and the method comprises the steps: in a current period, obtaining actual temperature values corresponding to each heating component in multiple periods; inputting the actual temperature values corresponding to each heating part in the plurality of periods into a pre-constructed temperature prediction model to obtain a predicted temperature value of each heating part in the next period; according to the actual temperature value of the target heating component in the current period and the predicted temperature value of the target heating component in the next period, whether heat dissipation operation is executed on the target heating component or not is determined; when it is determined that the heat dissipation operation is executed on the target heating component, the actual temperature value of the target heating component in the current period is input into a pre-constructed heat dissipation control model, and a target control quantity is obtained; and according to the target control quantity, the heat dissipation operation quantity of the target heat dissipation component within the preset distance range of the target heating component is increased. According to the invention, resource waste can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of servers, and in particular to a method, an apparatus, a computer device and a storage medium for performing heat dissipation operations. Background Art

[0002] In the field of server technology, servers are an important part of modern information technology infrastructure. Since servers usually need to undertake heavy computing tasks, they generate a lot of heat inside. If the heat cannot be dissipated in a timely and effective manner, it may cause the server to overheat, thereby affecting its performance and even causing hardware damage. Therefore, various heat dissipation components are usually required in the server to dissipate heat from the server in a timely manner. For example, the heat dissipation component can be a fan.

[0003] In the related art, the server generally collects its own temperature value periodically, and compares the temperature value collected in the current period with the preset temperature value. If the temperature value collected in the current period is greater than the preset temperature value, the control amount is calculated based on the temperature value collected in the current period and the control model, and the fan speed is increased according to the control amount to reduce the temperature.

[0004] However, the related technology may lead to the problem of excessive heat dissipation or insufficient heat dissipation, thereby causing the problem of waste of resources. Summary of the invention

[0005] In view of this, the present invention provides a method, apparatus, computer device, storage medium and program product for performing heat dissipation operations to solve the problem of excessive heat dissipation or insufficient heat dissipation caused by related technologies.

[0006] In a first aspect, the present invention provides a method for performing a heat dissipation operation, the method being applied to a server, the server comprising at least one heat generating component, the method comprising:

[0007] In the current cycle, obtaining actual temperature values ​​of each of the heat generating components corresponding to a plurality of cycles, wherein the plurality of cycles include the current cycle and at least one cycle before the current cycle;

[0008] Inputting actual temperature values ​​corresponding to each of the heat generating components in a plurality of cycles into a pre-built temperature prediction model, and obtaining a predicted temperature value of each of the heat generating components in a next cycle corresponding to the current cycle output by the temperature prediction model;

[0009] Determine whether to perform a heat dissipation operation on the target heat-generating component according to the actual temperature value of the target heat-generating component in the current cycle and the predicted temperature value in the next cycle, wherein the target heat-generating component is any heat-generating component of at least one of the heat-generating components;

[0010] When it is determined to perform a heat dissipation operation on the target heat-generating component, the actual temperature value of the target heat-generating component in the current cycle is input into a pre-built heat dissipation control model to obtain a target control amount;

[0011] According to the target control amount, the heat dissipation operation amount of the target heat dissipation component within the preset distance range of the target heat dissipation component is increased.

[0012] The method for performing heat dissipation operation provided by the present invention has the following advantages:

[0013] This solution can periodically collect the actual temperature value of each heating component. For the current cycle, the server can predict the temperature value of the heating component in the next cycle based on the actual temperature value of each heating component in the current cycle and a series of cycles before it. Then, for the target heating component, the future trend of the target heating component, that is, whether the temperature is rising or falling, can be determined by the actual temperature value and predicted temperature value of the target heating component in the current cycle, and then the heat dissipation operation is performed by the actual temperature value of the current cycle. In this way, when the future temperature trend is rising, the server can dissipate heat in advance to avoid the need for more power to dissipate heat when the subsequent heat accumulation is large, which wastes resources. When the future temperature trend is decreasing, if the heat dissipation operation is taken in the current cycle, it will lead to the problem of waste of resources caused by excessive heat dissipation. In addition, this solution uses the actual temperature value of the target heating component in the current cycle to calculate the target control amount, rather than the predicted temperature value. In this way, when the heat dissipation operation is taken, the actual temperature value of the current cycle is used for heat dissipation, which can save resources.

[0014] In an optional embodiment, the temperature prediction model includes multiple weight matrices, multiple activation functions and multiple bias vectors, the weight matrix is ​​used to perform a linear transformation on the input vector, the activation function is used to perform a nonlinear transformation on the input vector, and the bias vector is used to adjust the activation degree of the corresponding activation function;

[0015] The step of inputting the actual temperature values ​​corresponding to each of the heat generating components in a plurality of cycles into a pre-built temperature prediction model to obtain the predicted temperature value of each of the heat generating components in the next cycle corresponding to the current cycle output by the temperature prediction model comprises:

[0016] Generate a first data input vector corresponding to the first cycle according to the actual temperature value of each of the heat generating components in the first cycle, wherein the first cycle is any cycle among the multiple cycles except the first cycle;

[0017] splicing a second data output vector generated in a second cycle and the first data input vector to obtain a total data input vector corresponding to the first cycle, wherein the second cycle is a cycle previous to the first cycle;

[0018] Taking each of the activation functions as a target activation function, inputting the total data input vector corresponding to the first period, the target weight matrix and the target bias vector corresponding to the target activation function into the target activation function to obtain a transformation vector output by the target activation function;

[0019] Acquire a first state vector corresponding to the second period;

[0020] Generate a first data output vector of the first period according to the transformation vector output by each of the activation functions and the first state vector;

[0021] After generating a data output vector corresponding to the last cycle among the multiple cycles, the data output vector corresponding to the last cycle is dimensionally transformed according to a first weight matrix in at least one of the weight matrices to obtain a predicted temperature vector, wherein the predicted temperature vector includes the predicted temperature value of each of the heat-generating components in the next cycle.

[0022] Specifically, during the prediction process, the temperature prediction model can combine the previous temperature and the temperature of the current cycle to jointly predict the temperature of the next cycle, so that it can predict a more accurate temperature value through the evolution trend of the temperature. In addition, by processing the total data input vector multiple times through multiple weight matrices, multiple activation functions, and multiple bias vectors, the temperature prediction model can learn the characteristics of temperature changes and predict a more accurate temperature value. In summary, this solution can provide a more accurate basis for the subsequent execution of heat dissipation operations and avoid the problem of resource waste caused by performing incorrect heat dissipation operations.

[0023] In an optional implementation, generating the first data output vector of the first period according to the transformation vector output by each of the activation functions and the first state vector includes:

[0024] Determine a second state vector corresponding to the first period according to transformation vectors respectively output by activation functions other than the preset activation function among the plurality of activation functions and the first state vector;

[0025] The first data output vector is determined according to the transformation vector output by the preset activation function and the second state vector.

[0026] Specifically, by combining the transformation vector output by the activation function with the state vector, the temperature prediction model can learn the characteristics of temperature changes and predict a more accurate temperature value.

[0027] In an optional implementation, each of the activation functions is used as a target activation function, and the total data input vector corresponding to the first period, the target weight matrix and the target bias vector corresponding to the target activation function are input into the target activation function to obtain the transformation vector output by the target activation function, using the following expression:

[0028] q t = J(W h ·[h t-1 ,x t ]+b h )

[0029] Among them, q t is the transformation vector output by the target activation function, J is the target activation function, and the target activation function is a Sigmoid function or a hyperbolic tangent function, [h t-1 ,x t ] is the input vector of the current time step, x t is the output vector of the current time step, h t-1 is the output vector of the previous time step, b h is the target bias vector.

[0030] In an optional implementation, the second state vector corresponding to the first period is determined according to the transformation vectors respectively output by the activation functions other than the preset activation function among the plurality of activation functions and the first state vector, using the following expression:

[0031]

[0032] Among them, c t is the second state vector, f t is the transformation vector output by the first activation function among the plurality of activation functions, c t-1 is the first state vector, i t is the transformation vector output by the second activation function among the plurality of activation functions, A transformation vector output by a third activation function among the plurality of activation functions.

[0033] In an optional implementation, the first data output vector is determined according to the transformation vector output by the preset activation function and the second state vector, using the following expression:

[0034] h t=o t ·tanh(c t )

[0035] Among them, h t is the first data output vector, o t is the transformation vector output by the preset activation function, c t is the second state vector.

[0036] In an optional implementation, determining whether to perform a heat dissipation operation on the target heat-generating component according to the actual temperature value of the target heat-generating component in the current cycle and the predicted temperature value in the next cycle includes:

[0037] Subtract the actual temperature value of the target heating component in the current cycle from the predicted temperature value of the target heating component in the next cycle to obtain a temperature change value of the target heating component;

[0038] Determine whether the temperature change value of the target heat-generating component is greater than or equal to a preset temperature change threshold;

[0039] When it is determined that the temperature change value of the target heat-generating component is greater than or equal to the temperature change threshold, determining to perform a heat dissipation operation on the target heat-generating component;

[0040] or,

[0041] When it is determined that the temperature change value of the target heat generating component is less than the temperature change threshold, it is determined not to perform a heat dissipation operation on the target heat generating component.

[0042] Specifically, by comparing the temperature change value with the temperature change threshold, the server can start taking heat dissipation measures before reaching a dangerous level, which can effectively prevent the heating components from overheating and protect the safety of the server.

[0043] In a second aspect, the present invention provides a device for performing a heat dissipation operation, the device being applied to a server, the server comprising at least one heat generating component, the device comprising:

[0044] In the current cycle, obtaining actual temperature values ​​of each of the heat generating components corresponding to a plurality of cycles, wherein the plurality of cycles include the current cycle and at least one cycle before the current cycle;

[0045] Inputting actual temperature values ​​corresponding to each of the heat generating components in a plurality of cycles into a pre-built temperature prediction model, and obtaining a predicted temperature value of each of the heat generating components in a next cycle corresponding to the current cycle output by the temperature prediction model;

[0046] Determine whether to perform a heat dissipation operation on the target heat-generating component according to the actual temperature value of the target heat-generating component in the current cycle and the predicted temperature value in the next cycle, wherein the target heat-generating component is any heat-generating component of at least one of the heat-generating components;

[0047] When it is determined to perform a heat dissipation operation on the target heat-generating component, the actual temperature value of the target heat-generating component in the current cycle is input into a pre-built heat dissipation control model to obtain a target control amount;

[0048] According to the target control amount, the heat dissipation operation amount of the target heat dissipation component within the preset distance range of the target heat dissipation component is increased.

[0049] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the method for performing heat dissipation operations of the above-mentioned first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0050] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for performing a heat dissipation operation according to the first aspect or any corresponding embodiment thereof.

[0051] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the method for performing a heat dissipation operation according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 is a schematic diagram of the architecture of a server according to an embodiment of the present invention;

[0054] Figure 2 is a flow chart of a method for performing a heat dissipation operation according to an embodiment of the present invention;

[0055] Figure 3 is a structural block diagram of an apparatus for performing heat dissipation operation according to an embodiment of the present invention;

[0056] Figure 4It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0058] The present invention provides a server, such as Figure 1As shown, the server may include various heat-generating components that generate heat during operation, such as a central processing unit (CPU), a graphics processing unit (GPU), a memory module, and a hard disk. The surface of the heat-generating component may be sprayed with a graphene coating to improve the heat dissipation efficiency of the heat-generating component itself. In order to collect the temperature of the heat-generating component, a temperature sensor may be integrated on each heat-generating component, such as a thermistor (Negative Temperature Coefficient, NTC) resistor, a resistance temperature detector (Resistance Temperature Detector, RTD), etc. In order to dissipate the heat of the above-mentioned heat-generating components, the server may also include heat-dissipating components, such as a fan, a pump, a solenoid valve, a coolant pipe, etc. In addition, the server may also include at least one thermoelectric converter, which is mainly used to absorb the heat energy generated inside the server and convert it into electrical energy. The principle followed by the thermoelectric converter is the Seebeck effect, that is, V=SΔT, where V is voltage, S is the Seebeck coefficient, and ΔT is the temperature difference. In some optional embodiments, the structure of the server can be obtained by finite element analysis (FEA) software, specifically: Since the CPU undertakes more computing tasks and is one of the more important components in the server, the temperature reduction threshold of the CPU can be pre-set, for example, 10%. The technician can pre-set that the technician can create a three-dimensional geometric model inside the server, use FEA software to mesh the created three-dimensional geometric model, and define the material properties of the heat-generating components, such as thermal conductivity, specific heat capacity, and define boundary conditions, the maximum power consumption of the CPU, the ambient temperature range, and the working state of the cooling system. The working state of the cooling system can be the speed range of the fan under different loads, the flow rate, pressure and initial temperature of the liquid in the liquid cooling system, etc. Furthermore, the shape, quantity, distribution, etc. of the heat dissipation components inside the server can be changed, and the temperature change rate of the CPU under different server structures can be simulated by FEA software, and the server structure with the largest temperature change rate can be selected according to the heat dissipation of the CPU under different server structures. In the process of simulation, the FEA software can use the Fourier heat conduction formula: Wherein, T represents the temperature value, α is the preset thermal diffusion coefficient, and t represents time.

[0059] An embodiment of the present invention provides a method for performing a heat dissipation operation, which predicts the temperature value of the next cycle through the actual temperature values ​​detected in the current cycle and before the current cycle, thereby judging the subsequent temperature trend and performing a heat dissipation operation in advance to achieve the effect of saving resources.

[0060] According to an embodiment of the present invention, a method embodiment for performing a heat dissipation operation is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0061] In this embodiment, a method for performing a heat dissipation operation is provided, which can be used for the above-mentioned server. Figure 2 is a flow chart of a method for performing a heat dissipation operation according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0062] Step S201, in the current cycle, obtaining actual temperature values ​​of each heating component corresponding to multiple cycles.

[0063] The multiple cycles may include the current cycle and at least one cycle before the current cycle.

[0064] Specifically, the server can periodically collect the actual temperature value of the corresponding heat-generating component from each temperature sensor according to the pre-configured cycle duration and store it. After completing the collection operation of the current cycle, the server can predict the temperature value of each heat-generating component in the next cycle based on the actual temperature value of each heat-generating component collected in the current cycle and at least one cycle before the current cycle.

[0065] Step S202, inputting the actual temperature values ​​corresponding to each heat-generating component in multiple cycles into a pre-built temperature prediction model, and obtaining the predicted temperature value of each heat-generating component in the next cycle corresponding to the current cycle output by the temperature prediction model.

[0066] Specifically, the temperature prediction model may include multiple weight matrices, multiple activation functions, and multiple bias vectors. Among them, the temperature prediction model may be a long short-term memory network (Long Short-Term Memory, LSTM). The weight matrix may be used to perform a linear transformation on the input vector, the activation function may be used to perform a nonlinear transformation on the input vector, and the bias vector may be used to adjust the activation degree of the corresponding activation function. Accordingly, the process of calculating the predicted temperature value by the temperature prediction model may be as follows:

[0067] Step 1: Generate a first data input vector corresponding to the first period according to the actual temperature value of each heat-generating component in the first period.

[0068] The first cycle is any cycle among the multiple cycles except the first cycle.

[0069] For example, there are three heat-generating components, and in the first cycle, the actual temperature values ​​corresponding to the three heat-generating components are 25, 30, and 35 respectively, and the first data input vector is [25, 20, 35].

[0070] Step 2: concatenate the second data output vector generated in the second cycle with the first data input vector to obtain a total data input vector corresponding to the first cycle.

[0071] The second cycle is the previous cycle of the first cycle.

[0072] Step three, take each activation function as the target activation function, input the total data input vector corresponding to the first cycle, the target weight matrix and the target bias vector corresponding to the target activation function into the target activation function, and obtain the transformation vector output by the target activation function.

[0073] q t = J(W h ·[h t-1 ,x t ]+b h )

[0074] Among them, q t is the transformation vector output by the target activation function, J is the target activation function, the target activation function is the Sigmoid function or the hyperbolic tangent function, [h t-1 ,x t ] is the input vector of the current time step, x t is the output vector of the current time step, h t-1 is the output vector of the previous time step, b h is the target bias vector.

[0075] In some optional embodiments, the plurality of activation functions may include a first activation function, a second activation function, a third activation function, and a fourth activation function, wherein the first activation function, the second activation function, and the fourth activation function may be Sigmoid functions, which may be expressed as The third activation function may be a hyperbolic tangent function. For example, the transformation vector output by the first activation function may be as follows:

[0076] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0077] Among them, f t is the transformation vector output by the first activation function, σ(x) is the first activation function, and x is W f ·[h t-1 ,x t]+b f , W f is the second weight matrix, b f is the first bias vector.

[0078] The transformation vector output by the second activation function can be as follows:

[0079] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0080] Among them, i t is the transformation vector output by the second activation function, W i is the third weight matrix, b i is the second bias vector.

[0081] The transformation vector output by the third activation function can be as follows:

[0082]

[0083] in, is the transformation vector output by the third activation function, W c is the fourth weight matrix, b c is the third bias vector.

[0084] The transformation vector output by the preset activation function can be as follows:

[0085] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0086] Among them, t is the transformation vector output by the preset activation function, W o is the fifth weight matrix, b o is the fourth bias vector.

[0087] Step 4: Obtain a first state vector corresponding to the second period.

[0088] Step five: Generate a first data output vector of a first period according to the transformation vector output by each activation function and the first state vector.

[0089] Step 1: Determine a second state vector corresponding to a first period according to transformation vectors and first state vectors respectively output by activation functions other than a preset activation function among a plurality of activation functions.

[0090] Specifically, step 1 can be expressed as follows:

[0091]

[0092] The first state vector corresponding to the second cycle may be calculated according to the formula during the processing of the second cycle.

[0093] Step 2: Determine a first data output vector according to a transformation vector output by a preset activation function and a second state vector.

[0094] Specifically, step 2 can be expressed as follows:

[0095] h t =o t ·tanh(c t )

[0096] Among them, h t is the first data output vector.

[0097] Step six, after generating a data output vector corresponding to the last cycle among multiple cycles, perform dimension conversion on the data output vector corresponding to the last cycle according to a first weight matrix in at least one weight matrix to obtain a predicted temperature vector.

[0098] The predicted temperature vector includes the predicted temperature value of each heat-generating component in the next cycle.

[0099] The calculation process of the data output vector for the first cycle is similar to the above calculation process, except that in step 2, the server can concatenate the zero vector of the first preset dimension and the data input vector corresponding to the first cycle to obtain the total data input vector corresponding to the first cycle, that is, the above h t-1 It can be a zero vector of the first preset dimension. When calculating the state vector of the first cycle, the zero vector of the second preset dimension can be obtained in step 4, that is, the above c t-1 It can be a zero vector of a second preset dimension. In step five, the second state vector of the first cycle can be determined based on the transformation vectors and the zero vector of the second preset dimension output by the activation functions other than the preset activation function in multiple activation functions.

[0100] Step S203: determining whether to perform a heat dissipation operation on the target heat-generating component according to the actual temperature value of the target heat-generating component in the current cycle and the predicted temperature value in the next cycle.

[0101] The target heat-generating component is any heat-generating component among the at least one heat-generating component.

[0102] Specifically, step S203 may include the following steps:

[0103] Step 1: Subtract the actual temperature value of the target heating component in the current cycle from the predicted temperature value of the target heating component in the next cycle to obtain the temperature change value of the target heating component.

[0104] Step 2: Determine whether the temperature change value of the target heating component is greater than or equal to a preset temperature change threshold.

[0105] Step three: when it is determined that the temperature change value of the target heat-generating component is greater than or equal to the temperature change threshold, determine to perform a heat dissipation operation on the target heat-generating component.

[0106] Step 4: when it is determined that the temperature change value of the target heat-generating component is less than the temperature change threshold, it is determined not to perform a heat dissipation operation on the target heat-generating component.

[0107] In some optional embodiments, when it is determined that the temperature change value of the target heat-generating component is less than the temperature change threshold and the temperature change threshold is a value less than 0, the server may determine not to perform a heat dissipation operation on the target heat-generating component. Alternatively, when it is determined that the temperature change value of the target heat-generating component is less than the temperature change threshold and the temperature change threshold is a value greater than or equal to 0, the server may also determine whether the predicted temperature value of the target heat-generating component is greater than or equal to the preset temperature threshold. When it is determined that the predicted temperature value of the target heat-generating component is greater than the preset temperature threshold, a heat dissipation operation may also be performed on the target heat-generating component. Alternatively, when it is determined that the predicted temperature value of the target heat-generating component is less than or equal to the preset temperature threshold, it is determined not to perform a heat dissipation operation on the target heat-generating component.

[0108] In this way, by further considering the relationship between the predicted temperature value and the preset temperature threshold, it is possible to more accurately determine whether the predicted temperature value is a safe condition, thereby avoiding safety issues caused by a large predicted temperature value.

[0109] Step S204: when it is determined to perform a heat dissipation operation on the target heat generating component, the actual temperature value of the target heat generating component in the current cycle is input into a pre-built heat dissipation control model to obtain a target control amount.

[0110] Specifically, step S204 may adopt the following expression:

[0111]

[0112] Among them, u(t) is the target control quantity, K p is the preset proportional gain coefficient, K i is the preset integral gain coefficient, K d is the preset differential gain coefficient, e(t) is the temperature error value of the target heating component in the current cycle, e(τ) is the temperature error value in any cycle τ between the initial cycle and the current cycle, and the temperature error value is the difference between the actual temperature value and the preset safety temperature value.

[0113] Step S205 , according to the target control amount, increasing the heat dissipation operation amount of the target heat dissipation component within the preset distance range of the target heat-generating component.

[0114] Specifically, when the heat dissipation method adopted by the server includes fan heat dissipation, and the target heat dissipation component is a fan, the heat dissipation operation amount can be the rotational speed; when the heat dissipation method adopted by the server includes liquid cooling, and the target heat dissipation components include a pump and a solenoid valve, the heat dissipation operation amount corresponding to the solenoid valve can be the valve opening, and the heat dissipation operation amount corresponding to the pump can be the rotational speed.

[0115] At least one heat dissipation component can be arranged within a preset distance range of each heat-generating component. For each heat-generating component, the server can obtain identification information of the heat-generating component, and based on the identification information of the heat-generating component, obtain identification information of at least one heat-generating component within the preset distance range of the heat-generating component, and based on the identification information of the heat-generating component and the target control amount, increase the heat dissipation operation amount of the corresponding heat-generating component, for example, increase the speed of the fan, or increase the speed of the pump and increase the opening of the solenoid valve.

[0116] For example, the actual temperature value of the target heat-generating component in the current cycle is 70°C, and the predicted temperature value in the next cycle is 75°C. The target control amount calculated by the heat dissipation control model is 3000RPM (unit of rotational speed). The server can increase the speed of the fan corresponding to the target heat-generating component according to the target control amount.

[0117] In some optional implementations, when it is determined that the heat dissipation operation is not performed on the target heat-generating component, the actual temperature value of the target heat-generating component in the current cycle is input into a pre-built heat dissipation control model to obtain a target control amount, and according to the target control amount, the heat dissipation operation amount of the target heat-generating component within a preset distance range of the target heat-generating component is reduced. In this way, the problem of waste of resources can be avoided.

[0118] The method for performing heat dissipation operation provided in this embodiment can periodically collect the actual temperature value of each heat-generating component. For the current cycle, the server can predict the temperature value of the heat-generating component in the next cycle based on the actual temperature value of each heat-generating component in the current cycle and a series of cycles before it. Then, for the target heat-generating component, the future trend of the target heat-generating component, that is, whether the temperature is rising or falling, can be determined by the actual temperature value and predicted temperature value of the target heat-generating component in the current cycle, and then the heat dissipation operation is performed by the actual temperature value of the current cycle. In this way, when the trend of the future temperature is rising, the server can perform heat dissipation in advance to avoid the need for more power to dissipate heat when the subsequent heat accumulation is large, which wastes resources. When the trend of the future temperature is decreasing, if the heat dissipation operation is taken in the current cycle, it will lead to the problem of waste of resources caused by excessive heat dissipation. In addition, this solution uses the actual temperature value of the target heat-generating component in the current cycle to calculate the target control amount, rather than the predicted temperature value. In this way, when the heat dissipation operation is taken, the actual temperature value of the current cycle is used for heat dissipation, which can save resources.

[0119] In summary, using heat-conducting materials and a server structure with good heat conduction effect for heat-generating components, and using fans and liquid cooling systems to dissipate heat from the server at the same time can improve the heat dissipation efficiency of the server.

[0120] In this embodiment, a device for performing heat dissipation operation is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the term "module" may be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0121] This embodiment provides a device for performing heat dissipation operations, such as Figure 3 As shown, including:

[0122] An acquisition module 301 is used to acquire, in a current cycle, actual temperature values ​​of each heating component corresponding to a plurality of cycles, wherein the plurality of cycles include the current cycle and at least one cycle before the current cycle;

[0123] The determination module 302 is used to input the actual temperature values ​​corresponding to each heat-generating component in multiple cycles into a pre-built temperature prediction model to obtain the predicted temperature value of each heat-generating component in the next cycle corresponding to the current cycle output by the temperature prediction model; determine whether to perform a heat dissipation operation on the target heat-generating component according to the actual temperature value of the target heat-generating component in the current cycle and the predicted temperature value in the next cycle, wherein the target heat-generating component is any heat-generating component among the at least one heat-generating component;

[0124] The improvement module 303 is used to input the actual temperature value of the target heat-generating component in the current cycle into a pre-built heat-dissipation control model to obtain a target control amount when it is determined to perform a heat dissipation operation on the target heat-generating component; and improve the heat dissipation operation amount of the target heat-dissipating component within a preset distance range of the target heat-generating component according to the target control amount.

[0125] In some optional embodiments, the temperature prediction model includes multiple weight matrices, multiple activation functions, and multiple bias vectors, the weight matrix is ​​used to perform a linear transformation on the input vector, the activation function is used to perform a nonlinear transformation on the input vector, and the bias vector is used to adjust the activation degree of the corresponding activation function;

[0126] The determination module 302 is specifically configured to:

[0127] Generate a first data input vector corresponding to the first cycle according to the actual temperature value of each heat-generating component in the first cycle, wherein the first cycle is any cycle among the multiple cycles except the first cycle;

[0128] splicing the second data output vector generated in the second cycle and the first data input vector to obtain a total data input vector corresponding to the first cycle, wherein the second cycle is a cycle previous to the first cycle;

[0129] Taking each activation function as the target activation function, the total data input vector corresponding to the first cycle, the target weight matrix and the target bias vector corresponding to the target activation function are input into the target activation function to obtain the transformation vector output by the target activation function;

[0130] Acquire a first state vector corresponding to the second period;

[0131] Generate a first data output vector of a first cycle according to the transformation vector output by each activation function and the first state vector;

[0132] After generating a data output vector corresponding to the last cycle among multiple cycles, the data output vector corresponding to the last cycle is dimensionally transformed according to a first weight matrix in at least one weight matrix to obtain a predicted temperature vector, wherein the predicted temperature vector includes the predicted temperature value of each heat-generating component in the next cycle.

[0133] In some optional implementations, the determination module 302 is specifically configured to:

[0134] Determine a second state vector corresponding to the first period according to transformation vectors and the first state vector respectively output by activation functions other than the preset activation function among the plurality of activation functions;

[0135] A first data output vector is determined according to the transformation vector output by a preset activation function and the second state vector.

[0136] In some optional implementations, each activation function is used as the target activation function, and the total data input vector corresponding to the first period, the target weight matrix and the target bias vector corresponding to the target activation function are input into the target activation function to obtain the transformation vector output by the target activation function, using the following expression:

[0137] q t = J(W h ·[h t-1 ,x t ]+b h )

[0138] Among them, q t is the transformation vector output by the target activation function, J is the target activation function, the target activation function is the Sigmoid function or the hyperbolic tangent function, [h t-1 ,x t ] is the input vector of the current time step, x t is the output vector of the current time step, h t-1 is the output vector of the previous time step, b h is the target bias vector.

[0139] In some optional implementations, the second state vector corresponding to the first period is determined according to the transformation vectors and the first state vector respectively output by the activation functions other than the preset activation function in the plurality of activation functions, using the following expression:

[0140]

[0141] Among them, c t is the second state vector, f t is the transformation vector output by the first activation function among multiple activation functions, c t-1 is the first state vector, i t is the transformation vector output by the second activation function among the multiple activation functions, A transformation vector output by a third activation function among the plurality of activation functions.

[0142] In some optional implementations, the first data output vector is determined according to the transformation vector output by the preset activation function and the second state vector, using the following expression:

[0143] h t =o t ·tanh(c t )

[0144] Among them, ht is the first data output vector, o t is the transformation vector output by the preset activation function, c t is the second state vector.

[0145] In some optional implementations, the determination module 302 is specifically configured to:

[0146] Subtract the predicted temperature value of the target heating component in the next cycle from the actual temperature value of the target heating component in the current cycle to obtain a temperature change value of the target heating component;

[0147] Determine whether the temperature change value of the target heat-generating component is greater than or equal to a preset temperature change threshold;

[0148] When it is determined that the temperature change value of the target heat-generating component is greater than or equal to the temperature change threshold, determining to perform a heat dissipation operation on the target heat-generating component;

[0149] or,

[0150] When it is determined that the temperature change value of the target heat generating component is less than the temperature change threshold, it is determined not to perform the heat dissipation operation on the target heat generating component. The further functional description of the above modules and units is the same as that of the above corresponding embodiments, and will not be repeated here.

[0151] The device for performing heat dissipation operations in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0152] The embodiment of the present invention also provides a computer device having the above Figure 3 The device shown performs heat dissipation operations.

[0153] See also Figure 4 , Figure 4 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 4As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0154] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware integrated circuit. The hardware integrated circuit may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable logic gate array, a general purpose array logic or any combination thereof.

[0155] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0156] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0157] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0158] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0159] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0160] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0161] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for performing a heat dissipation operation, characterized in that: The method is applied to a server, the server includes at least one heat generating component, and the method includes: In the current cycle, obtaining actual temperature values ​​of each of the heat generating components corresponding to a plurality of cycles, wherein the plurality of cycles include the current cycle and at least one cycle before the current cycle; Inputting actual temperature values ​​corresponding to each of the heat generating components in a plurality of cycles into a pre-built temperature prediction model, and obtaining a predicted temperature value of each of the heat generating components in a next cycle corresponding to the current cycle output by the temperature prediction model; Determine whether to perform a heat dissipation operation on the target heat-generating component according to the actual temperature value of the target heat-generating component in the current cycle and the predicted temperature value in the next cycle, wherein the target heat-generating component is any heat-generating component of at least one of the heat-generating components; When it is determined to perform a heat dissipation operation on the target heat-generating component, the actual temperature value of the target heat-generating component in the current cycle is input into a pre-built heat dissipation control model to obtain a target control amount; According to the target control amount, the heat dissipation operation amount of the target heat dissipation component within the preset distance range of the target heat dissipation component is increased.

2. The method according to claim 1, characterized in that The temperature prediction model includes a plurality of weight matrices, a plurality of activation functions and a plurality of bias vectors, wherein the weight matrix is ​​used to perform a linear transformation on the input vector, the activation function is used to perform a nonlinear transformation on the input vector, and the bias vector is used to adjust the activation degree of the corresponding activation function; The step of inputting the actual temperature values ​​corresponding to each of the heat generating components in a plurality of cycles into a pre-built temperature prediction model to obtain the predicted temperature value of each of the heat generating components in the next cycle corresponding to the current cycle output by the temperature prediction model comprises: Generate a first data input vector corresponding to the first cycle according to the actual temperature value of each of the heat generating components in the first cycle, wherein the first cycle is any cycle among the multiple cycles except the first cycle; splicing a second data output vector generated in a second cycle and the first data input vector to obtain a total data input vector corresponding to the first cycle, wherein the second cycle is a cycle previous to the first cycle; Taking each of the activation functions as a target activation function, inputting the total data input vector corresponding to the first period, the target weight matrix and the target bias vector corresponding to the target activation function into the target activation function to obtain a transformation vector output by the target activation function; Acquire a first state vector corresponding to the second period; Generate a first data output vector of the first period according to the transformation vector output by each of the activation functions and the first state vector; After generating a data output vector corresponding to the last cycle among the multiple cycles, the data output vector corresponding to the last cycle is dimensionally transformed according to a first weight matrix in at least one of the weight matrices to obtain a predicted temperature vector, wherein the predicted temperature vector includes the predicted temperature value of each of the heat-generating components in the next cycle.

3. The method according to claim 2, characterized in that The step of generating a first data output vector of the first cycle according to the transformation vector output by each activation function and the first state vector comprises: Determine a second state vector corresponding to the first period according to transformation vectors respectively output by activation functions other than the preset activation function among the plurality of activation functions and the first state vector; The first data output vector is determined according to the transformation vector output by the preset activation function and the second state vector.

4. The method according to claim 2 or 3, characterized in that: Each of the activation functions is used as the target activation function, and the total data input vector corresponding to the first period, the target weight matrix and the target bias vector corresponding to the target activation function are input into the target activation function to obtain the transformation vector output by the target activation function, using the following expression: q t =J(W h ·[h t-1 ,x t ]+b h ) Among them, q t is the transformation vector output by the target activation function, J is the target activation function, and the target activation function is a Sigmoid function or a hyperbolic tangent function, [h t-1 ,x t ] is the input vector of the current time step, x t is the output vector of the current time step, h t-1 is the output vector of the previous time step, b h is the target bias vector.

5. The method according to claim 3, characterized in that: The second state vector corresponding to the first period is determined according to the transformation vectors respectively output by the activation functions other than the preset activation function among the plurality of activation functions and the first state vector, using the following expression: Among them, c t is the second state vector, f t is the transformation vector output by the first activation function among the plurality of activation functions, c t-1 is the first state vector, i t is the transformation vector output by the second activation function among the plurality of activation functions, A transformation vector output by a third activation function among the plurality of activation functions.

6. The method according to claim 3 or 5, characterized in that: The first data output vector is determined according to the transformation vector output by the preset activation function and the second state vector, using the following expression: h t =o t ·tanh(c t ) Among them, h t is the first data output vector, o t is the transformation vector output by the preset activation function, c t is the second state vector.

7. The method according to any one of claims 1 to 3 or claim 5, characterized in that: The step of determining whether to perform a heat dissipation operation on the target heat generating component according to the actual temperature value of the target heat generating component in the current cycle and the predicted temperature value of the target heat generating component in the next cycle includes: Subtract the actual temperature value of the target heating component in the current cycle from the predicted temperature value of the target heating component in the next cycle to obtain a temperature change value of the target heating component; Determine whether the temperature change value of the target heat-generating component is greater than or equal to a preset temperature change threshold; When it is determined that the temperature change value of the target heat-generating component is greater than or equal to the temperature change threshold, determining to perform a heat dissipation operation on the target heat-generating component; or, When it is determined that the temperature change value of the target heat generating component is less than the temperature change threshold, it is determined not to perform a heat dissipation operation on the target heat generating component.

8. A device for performing a heat dissipation operation, characterized in that: The device is applied to a server, the server includes at least one heat generating component, and the device includes: An acquisition module, used for acquiring, in a current cycle, actual temperature values ​​of each of the heat generating components corresponding to a plurality of cycles, wherein the plurality of cycles include the current cycle and at least one cycle before the current cycle; A determination module, used to input the actual temperature values ​​corresponding to each of the heat-generating components in multiple cycles into a pre-built temperature prediction model, and obtain the predicted temperature value of each of the heat-generating components in the next cycle corresponding to the current cycle output by the temperature prediction model; determine whether to perform a heat dissipation operation on the target heat-generating component according to the actual temperature value of the target heat-generating component in the current cycle and the predicted temperature value in the next cycle, wherein the target heat-generating component is any heat-generating component among at least one of the heat-generating components; The improvement module is used to input the actual temperature value of the target heat-generating component in the current cycle into a pre-built heat-dissipation control model to obtain a target control amount when it is determined to perform a heat dissipation operation on the target heat-generating component; and improve the heat dissipation operation amount of the target heat-dissipating component within a preset distance range of the target heat-generating component according to the target control amount.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for performing a heat dissipation operation according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for performing a heat dissipation operation according to any one of claims 1 to 7.

Images