Liquid cooling heat dissipation dynamic flow control method, device, medium and product

Abstract

The application discloses a liquid cooling heat dissipation dynamic flow control method, equipment, medium and product, and relates to the technical field of liquid cooling heat dissipation. According to the scheme, a corresponding secondary liquid cooling distribution unit is arranged for each computing device in the liquid cooling heat dissipation system, load information of each component of the corresponding computing device can be collected, and the cooling liquid flow sum required by the corresponding computing device in the next collection cycle is determined according to the load information, so that the corresponding cooling liquid flow is controlled. Meanwhile, the secondary liquid cooling distribution unit reports the corresponding cooling liquid flow sum to the liquid cooling distribution unit, so that the liquid cooling distribution unit controls the total cooling liquid flow according to the cooling liquid flow sum of each computing device, and overall flow regulation and control are realized. More sensors need not be arranged for control, cost is reduced, liquid cooling dynamic flow control can be realized in combination with specific business load conditions, meanwhile, dynamic adjustment of the cooling liquid flow can be carried out according to the differences of heat dissipation pipelines of different components, and component-level fine flow control is realized.

Description

Technical Field

[0001] This invention relates to the field of liquid cooling technology, and in particular to a method, device, medium, and product for dynamic flow control of liquid cooling. Background Technology

[0002] Cold plate liquid cooling technology, with its highly efficient heat exchange characteristics through direct contact with the heat source, has become a key solution for high-density computing infrastructure. To achieve closed-loop control of coolant flow, current methods typically involve deploying multiple types of sensors to monitor system status in real time, combined with the coordinated control of solenoid valves and pump sets to achieve dynamic adjustment.

[0003] However, while the above methods can achieve real-time status monitoring, there is a delay in the sensor's temperature control response, and the main control program also has a response cycle after data reporting, making it impossible to accurately achieve the preset target for flow regulation. Furthermore, the Cooling Distribution Units (CDUs) use a centralized liquid cooling distribution scheme, which cannot perform liquid cooling heat dissipation control at the component-level cold plates, resulting in insufficient precision in heat dissipation. At the same time, more and more sophisticated sensors are required, further increasing the cost on top of the already high cost of cold plate-based liquid cooling, leading to a low cost-performance ratio.

[0004] Given the above, how to solve the problems faced by current liquid cooling distribution methods, such as sensor latency, response cycle of the main control program, inaccurate flow control, insufficient heat dissipation due to centralized liquid cooling distribution, and high cost, are urgent issues for technical personnel in this field. Summary of the Invention

[0005] This invention provides a method, device, medium, and product for dynamic flow control of liquid cooling, which at least solves the problems of sensor delay, response cycle of the main control program, inaccurate flow control, insufficient heat dissipation due to centralized liquid cooling distribution, and high cost in current liquid cooling distribution methods.

[0006] This invention provides a dynamic flow control method for liquid cooling, applied to the main control module of a secondary liquid cooling distribution unit. The liquid cooling system includes multiple secondary liquid cooling distribution units, each corresponding to a computing device. Each secondary liquid cooling distribution unit also includes a flow control component. The main control module is communicatively connected to the corresponding flow control component and various components of the corresponding computing device. The flow control component, the cold plate piping of the corresponding computing device, and the distributor are connected via piping. The method includes: Determine the acquisition cycle of each component of the corresponding computing device, and acquire the load information of the corresponding component according to each acquisition cycle; Determine the required coolant flow rate for each component in the next acquisition cycle based on the load information. Sum the flow rates of each coolant to determine the total coolant flow rate required by the corresponding computing device; The total coolant flow rate is controlled by the corresponding flow control component based on the total coolant flow rate, and the total coolant flow rate is reported to the liquid cooling distribution unit so that the liquid cooling distribution unit can control the total coolant flow rate based on the total coolant flow rate of each computing device.

[0007] The present invention also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described liquid cooling dynamic flow control methods.

[0008] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above-described liquid cooling dynamic flow control methods.

[0009] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any of the above-described liquid cooling dynamic flow control methods.

[0010] The beneficial effects of this invention are as follows: In the liquid cooling system, a corresponding secondary liquid cooling distribution unit is set up for each computing device. This unit can collect the load information of each component of the corresponding computing device, determine the total coolant flow rate required by the corresponding computing device in the next collection cycle based on the load information, and control the corresponding coolant flow rate accordingly. Simultaneously, the secondary liquid cooling distribution unit can report the corresponding total coolant flow rate to the liquid cooling distribution unit, enabling the liquid cooling distribution unit to control the total coolant flow rate based on the total coolant flow rate of each computing device, thereby achieving overall control of the coolant flow rate. This solution eliminates the need to deploy more sensors in the liquid cooling system, reducing costs. It can achieve dynamic liquid cooling flow control based on specific business load conditions, and can also dynamically adjust the coolant flow rate according to the differences in the heat dissipation pipes of different components, achieving fine-grained control of flow rate at the component level.

[0011] In addition, the present invention also provides a liquid-cooled heat dissipation dynamic flow control device, medium and product, with the same effect as above. Attached Figure Description

[0012] To more clearly illustrate the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1A schematic diagram of a liquid cooling heat dissipation system including a secondary liquid cooling distribution unit provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the secondary liquid-cooled distribution unit provided in an embodiment of the present invention; Figure 3 A flowchart of a liquid cooling dynamic flow control method provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a liquid-cooled heat dissipation dynamic flow control device provided in an embodiment of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present invention.

[0015] It should be noted that, in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., used in this invention are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0016] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] Figure 1 This is a schematic diagram of a liquid cooling system including a secondary liquid cooling distribution unit, provided as an embodiment of the present invention. Figure 1 As shown, the liquid cooling system includes multiple secondary liquid cooling distribution units (Sub-CDUs), each corresponding to a computing device. In this embodiment, the computing device is not limited; it can be a server, host, etc.

[0018] Figure 2 This is a schematic diagram of the secondary liquid-cooled distribution unit provided in an embodiment of the present invention. Figure 2As shown, the secondary liquid-cooled distribution unit consists of a main control module and a flow control component. The main control module is communicatively connected to the corresponding flow control component and various components of the corresponding computing device; the flow control component, the cold plate piping of the corresponding computing device, and the distributor are connected via piping. In this embodiment, the specific composition of the flow control component is not limited; for example, it can consist of a miniature water pump and a flow regulating valve. The method provided by this invention is applied to the main control module of the secondary liquid-cooled distribution unit. The following detailed description of the liquid-cooled heat dissipation dynamic flow control method, in conjunction with specific embodiments, is provided: Figure 3 This is a flowchart illustrating a dynamic flow control method for liquid cooling heat dissipation provided in an embodiment of the present invention. Figure 3 As shown, the method includes: S10: Determine the acquisition cycle of each component of the corresponding computing device, and acquire the load information of the corresponding component according to each acquisition cycle.

[0019] First, the main control module determines the acquisition cycle for each component of the corresponding computing device and collects the load information of the corresponding component according to each acquisition cycle. It should be noted that the components in this embodiment refer to those requiring liquid cooling, including but not limited to the Central Processing Unit (CPU), memory, input / output (I / O) devices, and Graphics Processing Unit (GPU). Since the acquisition cycles for each component are different, the acquisition cycle needs to be adaptively adjusted based on the load variation patterns of each component to provide more accurate load data. This embodiment does not limit the specific method for determining the acquisition cycle for each component; it depends on the specific implementation. Furthermore, this embodiment does not limit the specific load information to be collected. For example, the load information to be collected is shown in the table below: Table 1 Load Information Collection Table

[0020] It is understandable that Table 1 only provides an example of load information collection. In actual implementation, more or less load information may be collected, depending on the specific implementation situation.

[0021] S11: Determine the required coolant flow rate for each component in the next acquisition cycle based on the load information.

[0022] S12: Sum the coolant flow rates of each coolant to determine the total coolant flow rate required by the corresponding computing device.

[0023] Furthermore, based on the load information of each component, the required coolant flow rate of each component in the next acquisition cycle is calculated, and the coolant flow rates of each component are summed to determine the total coolant flow rate required by the corresponding computing device.

[0024] It should be noted that this embodiment does not limit the specific process for determining the required coolant flow rate of each component in the next acquisition cycle, and it depends on the specific implementation.

[0025] S13: Control the coolant flow rate of the corresponding flow control component based on the total coolant flow rate, and report the total coolant flow rate to the liquid cooling distribution unit so that the liquid cooling distribution unit can control the total coolant flow rate based on the total coolant flow rate of each computing device.

[0026] Finally, the main control module reports the total coolant flow rate of the corresponding computing devices to the liquid cooling distribution unit. At this time, the liquid cooling distribution unit receives the total coolant flow rate of multiple computing devices. In order to implement coolant flow control in the next cycle and achieve the goal of refined coolant flow control, the liquid cooling distribution unit will control the total coolant flow rate of the next cycle based on the total coolant flow rate of each computing device. At the same time, the main control module of each computing device will also control the coolant flow rate of the corresponding flow control component based on the corresponding total coolant flow rate, such as controlling the speed of the corresponding micro water pump and the opening of the flow regulating valve. This achieves refined control of the coolant flow rate required by each component in the next cycle, ultimately achieving the goal of refined heat dissipation for each component. This solves the problems of untimely flow response and inconsistent heat dissipation modes of different components leading to heat dissipation failure and component damage during liquid cooling heat dissipation, ensuring the stability and reliability of liquid-cooled data center production and reducing the component failure rate caused by liquid cooling heat dissipation.

[0027] It should be noted that this embodiment does not limit the specific process by which the liquid cooling distribution unit controls the total coolant flow rate based on the sum of the coolant flow rates of each computing device; it depends on the specific implementation.

[0028] In this embodiment, a corresponding secondary liquid cooling distribution unit is set up for each computing device in the liquid cooling system. This unit can collect the load information of each component of the corresponding computing device, determine the total coolant flow required by the corresponding computing device in the next collection cycle based on the load information, and control the corresponding coolant flow accordingly. Simultaneously, the secondary liquid cooling distribution unit can report the corresponding total coolant flow to the liquid cooling distribution unit, allowing the liquid cooling distribution unit to control the total coolant flow based on the total coolant flow of each computing device, thus achieving overall control of coolant flow. This solution eliminates the need for deploying more sensors in the liquid cooling system, reducing costs. It can achieve dynamic liquid cooling flow control based on specific business load conditions, and can also dynamically adjust the coolant flow according to the differences in the heat dissipation pipes of different components, achieving fine-grained flow control at the component level.

[0029] Based on the above embodiments, in some embodiments, the acquisition cycle of each component of the corresponding computing device is determined, including: S101: Determine the initial acquisition cycle for each component.

[0030] Each initial acquisition cycle is the same; the device components include at least a central processing unit, memory, input / output devices, and a graphics processing unit.

[0031] S102: Collect the initial load information of each component according to the initial acquisition cycle; S103: Determine the acquisition cycle of the corresponding device component based on the initial load information.

[0032] To determine the acquisition cycle for each component, it is first necessary to identify the specific component types involved in liquid cooling. In this embodiment, the components include at least a CPU, memory, I / O devices, and a GPU. Simultaneously, the initial acquisition cycle for each component is determined. It should be noted that the initial acquisition cycle is the same for all components. This embodiment does not impose any restrictions on the specific size of the initial acquisition cycle; for example, it could be 5 minutes.

[0033] Furthermore, initial load information for each component is collected according to the initial acquisition cycle, and the acquisition cycle for the corresponding component is then determined based on this initial load information. It should be noted that, in order to comprehensively analyze the load variation patterns of each component and provide more accurate load data, the acquisition cycles for each component determined based on the initial load information may not be identical.

[0034] Specifically, in order to determine the acquisition period of the corresponding device component based on the initial load information, the load standard deviation and current load change rate of the corresponding device component are determined based on the initial load information; the base period, dynamic amplification factor, and component weight value of each device component are obtained; and the acquisition period of each device component is determined based on the base period, dynamic amplification factor, component weight value of each device component, load standard deviation, and current load change rate, using the following formula: ; in, Let be the acquisition cycle (s) of the i-th component. The base period is set to 1 second by default, and must meet the minimum monitoring requirements. This is the dynamic amplification factor, recommended to be 0.5~1.0, used to balance fluctuation sensitivity; The load standard deviation of component i is a normalized value of 0 or 1, reflecting historical volatility. This represents the current load change rate, specifically in absolute value. The component weight values ​​are detailed in the table below: Table 2 Component Weights and Parameter Reference Table

[0035] As shown in Table 2 and the above formulas, when the load standard deviation or the rate of change of the load of a component increases, the numerator increases, the corresponding acquisition cycle shortens, and the component weight value reflects the priority. Furthermore, to achieve boundary control, the minimum cycle Tmin = 0.2s can be further set to avoid excessive overhead; simultaneously, the maximum cycle Tmax = 5s can be set to ensure real-time control.

[0036] In this embodiment, the initial load information of each component is collected according to the initial acquisition period, and the acquisition period of the corresponding component is determined according to the initial load information. The acquisition period is adaptively changed by taking into account the load variation pattern of each component in order to provide more accurate load data.

[0037] Based on the above embodiments, in some embodiments, the required coolant flow rate for each component in the corresponding next acquisition cycle is determined according to the load information, including: S110: Determine the heat generated by each component in the next acquisition cycle based on the load information; S111: Determine the required coolant flow rate for each component in the next data collection cycle based on the heat generated.

[0038] In practical implementation, to determine the required coolant flow rate for each component in the next data acquisition cycle, it is first necessary to determine the heat generated by each component in the next data acquisition cycle based on the load information, and then determine the required coolant flow rate for each component in the next data acquisition cycle based on the heat generated. The specific process of determining the heat generated by each component in the next data acquisition cycle is described in detail below with reference to a specific embodiment: (1) Determine the heat generated by the CPU in the next acquisition cycle; In some embodiments, determining the heat generated by each component in the corresponding next acquisition cycle based on the load information includes: S112: Determine the first static power consumption, first dynamic power consumption, maximum utilization, nonlinear correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification coefficient, and first liquid cooling efficiency factor corresponding to the central processing unit.

[0039] S113: Based on the first static power consumption, first dynamic power consumption, maximum utilization, nonlinear correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification coefficient, first liquid cooling efficiency factor and corresponding acquisition cycle of the central processing unit, determine the heat generated by the central processing unit in the next acquisition cycle.

[0040] Specifically, to determine the heat generated by the CPU in the next sampling cycle, it is first necessary to determine the CPU's first static power consumption, first dynamic power consumption, maximum utilization rate, nonlinearity correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification factor, and first liquid cooling efficiency factor. Based on the CPU's first static power consumption, first dynamic power consumption, maximum utilization rate, nonlinearity correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification factor, first liquid cooling efficiency factor, and the corresponding sampling cycle, the heat generated by the CPU in the next sampling cycle is determined using the following formula: ; in, The heat generated by the CPU (J) during the corresponding acquisition period Ti. The first static power consumption (W) is related to process and temperature, such as power consumption caused by leakage current. This is the first dynamic power consumption (W). This represents the maximum CPU utilization (100%). The non-linear correction coefficient (1.1~1.5) for the CPU reflects the accelerated power consumption growth characteristic under high load. CPU core temperature ( ); The ambient temperature corresponding to the CPU ( Specifically, it corresponds to the coolant inlet temperature (e.g., 40). ). For reference temperature difference (e.g., 50) ), used to normalize the effect of temperature difference. and These are all temperature amplification factors corresponding to the CPU. The range is 0.2 to 0.5. The range is 1.2 to 1.5, characterizing the feedback enhancement effect of temperature difference on heat production; This is the first liquid cooling heat dissipation efficiency factor.

[0041] It should be noted that, in order to determine the CPU's first dynamic power consumption, in practice, it is necessary to obtain the CPU's load capacitance, voltage, frequency, and current utilization. Based on the CPU's load capacitance, voltage, frequency, and current utilization, the CPU's first dynamic power consumption is determined using the following formula: ; in, This is the first dynamic power consumption; This is the CPU's load capacitor; This refers to the CPU voltage. For CPU frequency, This represents the current CPU utilization (user mode + system mode).

[0042] To determine the CPU core temperature, in practice, it is necessary to obtain the sensor temperature, heat transfer coefficient, and load intensity parameters set on the CPU. Based on these parameters, the CPU core temperature is determined using the following formula: ; in, The CPU core temperature. and These are all CPU sensor temperatures. and Both are heat transfer coefficients. For load strength parameters, The average temperature difference under CPU load (from zero load to full load).

[0043] To determine the CPU's first liquid cooling efficiency factor, in practice, it is necessary to obtain the CPU's actual thermal resistance, design thermal resistance, and optimized parameters. Based on the CPU's actual thermal resistance, design thermal resistance, and optimized parameters, the CPU's first liquid cooling efficiency factor is determined using the following formula: ; in, The first liquid cooling efficiency factor for the CPU. The actual thermal resistance corresponding to the CPU ( ), The thermal resistance corresponding to the CPU. and These are all optimized parameters corresponding to the CPU. The range is from 0.1 to 0.3. The range is from 1.0 to 1.5.

[0044] (2) Determine the heat generated in the next acquisition cycle corresponding to the memory; In some embodiments, determining the heat generated by each component in the corresponding next acquisition cycle based on the load information includes: S114: Determine the second static power consumption, memory utilization, utilization amplification factor, second dynamic power consumption, page switching frequency, switching frequency nonlinearity coefficient, basic switching power consumption, and second liquid cooling efficiency factor corresponding to the memory.

[0045] S115: Based on the second static power consumption, memory utilization, utilization amplification factor, second dynamic power consumption, page switching frequency, switching frequency nonlinearity coefficient, basic switching power consumption, second liquid cooling heat dissipation efficiency factor and the corresponding acquisition cycle, determine the heat generated by the memory in the next acquisition cycle.

[0046] Specifically, to determine the heat generated by the memory in the next acquisition cycle, it is first necessary to determine the memory's second static power consumption, memory utilization rate, utilization amplification factor, second dynamic power consumption, page swapping frequency, swapping frequency nonlinearity factor, base swapping power consumption, and second liquid cooling efficiency factor. Further, based on the memory's second static power consumption, memory utilization rate, utilization amplification factor, second dynamic power consumption, page swapping frequency, swapping frequency nonlinearity factor, base swapping power consumption, second liquid cooling efficiency factor, and the corresponding acquisition cycle, the heat generated by the memory in the next acquisition cycle is determined using the following formula: ; in, The heat (J) generated in memory during the corresponding acquisition period Ti. The second static power consumption (W) is related to memory leakage current and is affected by temperature. For example, increased leakage current at high temperatures leads to increased power consumption. For memory utilization; The memory utilization amplification factor (0.1~0.3) characterizes the accelerated growth effect of leakage current under high load; The second dynamic power consumption (W / GB / s) is related to the memory bandwidth Bi (GB / s) and reflects the energy consumption of data transmission. For example, the typical power consumption of DDR4-3200 with a bandwidth of 34.1GB / s is about 0.35W / GB / s. This is the page swapping frequency (times / s), calculated using pswpin and pswpout in / proc / vmstat. The nonlinear coefficient of the switching frequency (0.5~1.0) characterizes the exponential increase in power consumption caused by frequent switching, such as the superlinear increase in power consumption when the switching frequency doubles; Basic switching power consumption (W) reflects the average energy consumption of a single page switch (e.g., 5~10W). This is the second liquid cooling heat dissipation efficiency factor.

[0047] To determine the memory utilization rate, in practice, it is necessary to obtain the total memory capacity, free memory capacity, and used memory capacity. Based on the total memory capacity, free memory capacity, and used memory capacity, the memory utilization rate is determined using the following formula: ; in, This refers to the memory utilization rate corresponding to the memory. This represents the total memory capacity. This refers to the free memory capacity. This represents the amount of memory currently in use.

[0048] To determine the second liquid cooling efficiency factor for the memory, in practice, it is necessary to obtain the memory temperature, coolant inlet temperature, reference temperature difference, actual thermal resistance, design thermal resistance, and correction factor. Based on the memory temperature, coolant inlet temperature, reference temperature difference, actual thermal resistance, design thermal resistance, and correction factor, the second liquid cooling efficiency factor for the memory is determined.

[0049] ; in, The second liquid cooling efficiency factor corresponds to the memory. For memory temperature ( ), The ambient temperature corresponding to the memory, i.e., the coolant inlet temperature (e.g., 40). ); For reference temperature difference (50) ), The actual thermal resistance corresponding to the memory ( ), The thermal resistance is designed for the memory. , and These are all correction parameters corresponding to memory. The range is 1.2 to 1.5. The range is from 0.8 to 1.0. The range is from 0.1 to 0.2.

[0050] (3) Determine the heat generated by the IO device in the next acquisition cycle; In some embodiments, determining the heat generated by each component in the corresponding next acquisition cycle based on the load information includes: S116: Determine the third static power consumption, third dynamic power consumption, and third liquid cooling efficiency factor corresponding to the input / output devices.

[0051] S117: Determine the heat generated by the input / output device in the next acquisition cycle based on the third static power consumption, third dynamic power consumption, third liquid cooling efficiency factor and the corresponding acquisition cycle of the input / output device.

[0052] Specifically, to determine the heat generated by the I / O device in the next acquisition cycle, it is first necessary to determine the third static power consumption, third dynamic power consumption, and third liquid cooling efficiency factor of the I / O device. Then, based on the third static power consumption, third dynamic power consumption, third liquid cooling efficiency factor, and the corresponding acquisition cycle, the heat generated by the I / O device in the next acquisition cycle is determined, using the following formula: ; in, In order to correspond to the next acquisition cycle The heat generated by internal I / O devices This is the third static power consumption. This is the third dynamic power consumption. This is the third liquid cooling heat dissipation efficiency factor.

[0053] To determine the third static power consumption of an I / O device, in practice, it is necessary to obtain the device's base static power consumption, core temperature, coolant inlet temperature, reference temperature difference, and temperature coefficient. Based on these parameters, the third static power consumption of the I / O device is determined using the following formula: ; in, This is the third static power consumption corresponding to the I / O device. The base static power consumption (W) of the I / O device. The core temperature of the I / O device. This refers to the coolant inlet temperature corresponding to the IO device. For reference temperature difference, This refers to the temperature coefficient corresponding to the I / O device.

[0054] To determine the third dynamic power consumption of an I / O device, in practice, it is necessary to obtain the power consumption coefficient, read / write rate, input / output queue depth, device busy rate, and busy rate nonlinearity coefficient corresponding to the I / O device. Based on the power consumption coefficient, read / write rate, input / output queue depth, device busy rate, and busy rate nonlinearity coefficient corresponding to the I / O device, the third dynamic power consumption of the I / O device is determined, and the specific formula is as follows: ; in, The third dynamic power consumption (W) for I / O devices. The power consumption factor (W / MB / s) for I / O devices. This refers to the read / write speed (MB / s) of the I / O device. The input / output queue depth of the I / O device (e.g., the maximum queue depth of NVMe SSD is 64K). The device utilization rate (%util) of the I / O device. The busy rate nonlinear coefficient for I / O devices is 0.5~1.0.

[0055] To determine the third liquid cooling efficiency factor for I / O devices, in practice, it is necessary to obtain the actual thermal resistance, design thermal resistance, and correction factor of the I / O device. Based on the actual thermal resistance, design thermal resistance, correction factor, core temperature, coolant inlet temperature, and reference temperature difference of the I / O device, the third liquid cooling efficiency factor for the I / O device is determined, as shown in the following formula: ; in, The third liquid cooling efficiency factor for I / O devices. This refers to the actual thermal resistance corresponding to the I / O device. To design thermal resistance, , and All are correction factors. The range is from 0.1 to 0.3. The range is 1.0 to 1.5. The range is from 0.05 to 0.1.

[0056] (4) Determine the heat generated by the GPU in the next acquisition cycle; In some embodiments, determining the heat generated by each component in the corresponding next acquisition cycle based on the load information includes: S118: Determine the fourth static power consumption, fourth dynamic power consumption, and fourth liquid cooling efficiency factor corresponding to the graphics processing unit.

[0057] S119: Determine the heat generated by the graphics processing unit in the next acquisition cycle based on the fourth static power consumption, fourth dynamic power consumption, fourth liquid cooling efficiency factor and the corresponding acquisition cycle of the graphics processing unit.

[0058] Specifically, to determine the heat generated by the GPU in the next acquisition cycle, it is first necessary to determine the GPU's fourth static power consumption, fourth dynamic power consumption, and fourth liquid cooling efficiency factor. Then, based on the GPU's fourth static power consumption, fourth dynamic power consumption, fourth liquid cooling efficiency factor, and the corresponding acquisition cycle, the heat generated by the GPU in the next acquisition cycle is determined, using the following formula: ; in, To address the heat generated by the GPU during the next acquisition cycle, This is the fourth static power consumption for the GPU. This is the fourth dynamic power consumption. This is the fourth liquid cooling efficiency factor.

[0059] To determine the fourth static power consumption for the GPU, in practice, it is necessary to obtain the GPU's base static power consumption, core temperature, coolant inlet temperature, reference temperature difference, and temperature coefficient. Based on these parameters, the fourth static power consumption for the input / output devices is determined using the following formula: ; in, This is the fourth static power consumption for the GPU. This represents the GPU's base static power consumption (W). The core temperature of the GPU. This refers to the coolant inlet temperature for the GPU. For reference temperature difference, This represents the temperature coefficient corresponding to the GPU.

[0060] To determine the fourth dynamic power consumption of the GPU, in practice, it is necessary to obtain the GPU's memory utilization, compute unit utilization, task queue length, memory nonlinearity coefficient, task queue nonlinearity coefficient, and power consumption coefficient. Based on these parameters, the fourth dynamic power consumption of the input / output device is determined using the following formula: ; in, This is the fourth dynamic power consumption for the GPU. This represents the GPU's memory utilization (used memory / total memory), reflecting the data transfer load. The GPU's compute unit utilization (%) is obtained via nvidia-smi or ROCmSMI. This represents the length of the GPU's task queue (number of kernels to be processed), indicating the degree of task backlog. This is the non-linearity coefficient of the GPU's memory, reflecting the accelerated increase in power consumption under high load. is the non-linearity coefficient of the GPU's task queue, representing the exponential increase in power consumption when the queue is saturated; The power consumption factor (W) for the GPU needs to be calibrated through actual measurement.

[0061] To determine the fourth liquid cooling efficiency factor for the GPU, in practice, it is necessary to obtain the GPU's actual thermal resistance, design thermal resistance, correction factor, actual coolant flow rate, design flow rate, and flow rate compensation factor. Based on these parameters, the fourth liquid cooling efficiency factor for the GPU is determined using the following formula: ; in, This is the fourth liquid cooling efficiency factor for GPUs. The actual thermal resistance of the GPU ( (), affected by the cold plate flow channel design; For the design thermal resistance of the GPU ( ), cold plate optimization target value. and All are correction factors. The range is from 0.1 to 0.3. The range is from 1.0 to 1.5. The actual coolant flow rate (L / min) for the GPU is obtained via a CDU sensor. The design flow rate (L / min) for the GPU is required to meet the maximum thermal load requirements. The flow compensation coefficient represents the effect of flow rate reduction on heat dissipation efficiency, and its value ranges from 0.05 to 0.1.

[0062] In summary, the coolant flow rates of the CPU, memory, I / O devices, and GPU in the next acquisition cycle were obtained respectively. , , and Taking into account the specific load conditions of each component, the system accurately acquires the corresponding heat, so as to determine the required coolant flow rate of each component in the next acquisition cycle based on the heat.

[0063] Based on the above embodiments, in some embodiments, the required coolant flow rate for each component in the corresponding next acquisition cycle is determined according to the heat generated, including: S121: Determine the theoretical coolant flow rate required for each component in the next acquisition cycle based on the laws of conservation of heat and thermodynamic energy.

[0064] S122: Obtain the flow redundancy coefficient, pipeline resistance correction coefficient, total pipeline resistance value, and design resistance target value.

[0065] S123: Based on the flow redundancy coefficient, pipeline resistance correction coefficient, total pipeline resistance value, design resistance target value, and theoretical coolant flow rate, determine the actual coolant flow rate required by each component in the next acquisition cycle.

[0066] S124: Set the target speed of the water pump and the opening degree of the flow control component according to the actual coolant flow rate.

[0067] S125: Determine the required coolant flow rate for each component in the next data collection cycle based on the actual coolant flow rate, target pump speed, and flow control valve opening.

[0068] After obtaining the heat values, the theoretical coolant flow rate required for each component in the next data acquisition cycle is determined based on the heat values ​​and the law of conservation of thermodynamic energy. Specifically, based on the law of conservation of thermodynamic energy, the theoretical coolant flow rate is... satisfy: ; in, Theoretical coolant flow rate The density of the coolant; This refers to the specific heat capacity of the coolant. The maximum allowable temperature rise; The acquisition cycle is the period corresponding to the device component.

[0069] Further redundancy and system loss corrections are applied to the theoretical coolant flow rate. Specifically, the flow redundancy coefficient, pipe resistance correction coefficient, total pipe resistance value, and design resistance target value are obtained. Based on the flow redundancy coefficient, pipe resistance correction coefficient, total pipe resistance value, design resistance target value, and each theoretical coolant flow rate, the actual coolant flow rate required by each component in the corresponding next data acquisition cycle is determined, using the following formula: ; in, This represents the actual coolant flow rate. This is the traffic redundancy coefficient. This is the pipeline resistance correction factor. This represents the total resistance value of the pipeline. To design the target resistance value.

[0070] Then, the relationship between the pump speed and flow rate in the flow control component was adjusted. The pump flow rate is directly proportional to the pump speed, as shown in the following formula: ; in, For water pump flow rate, For rated flow rate, Rated speed, This refers to the current rotational speed. In this embodiment, it is necessary to ensure... And the operating point is located in the high-efficiency zone of the water pump (70%-110% of the rated flow).

[0071] At the same time, the actual flow rate is adjusted by changing the valve opening of the flow regulating valve. ,satisfy: ; in, For valve characteristic functions (such as linear or equal percentage characteristics), V needs to be dynamically adjusted to match. This allows for the setting of the target pump speed and flow regulating valve opening of the flow control component based on the actual coolant flow rate.

[0072] Finally, taking into account the above factors, based on the actual coolant flow rate, the target speed of the water pump, and the opening of the flow control valve, the required coolant flow rate for each component in the next data collection cycle is determined. The specific formula is as follows: ; in, This refers to the coolant flow rate required by the instrument components in the next data acquisition cycle. This represents the actual coolant flow rate. For the target speed of the water pump, it must meet the following requirements. ; The opening degree of the flow regulating valve is dynamically adjusted based on temperature feedback.

[0073] In this embodiment, by setting the water pump speed and the opening of the flow regulating valve to control the flow rate and maintain the coolant flow rate in the next cycle, the accuracy and timeliness of liquid cooling heat dissipation of the cold plates of each component can be guaranteed.

[0074] Correspondingly, the required coolant flow rates for each component are summed, the total coolant flow rate required by the calculation equipment in the next cycle is calculated, and then sent to the CDU. The formula is as follows: ; in, This represents the total coolant flow rate for a single computing device. This specifies the required coolant flow rate for each component (CPU, GPU, memory, I / O devices, etc.) in the next acquisition cycle.

[0075] It should also be noted that, after determining the total coolant flow rate for a single computing device... Next, based on the next acquisition cycle / load change patterns of each component (CPU, GPU, memory, I / O devices, etc.), the overall dynamic acquisition cycle of the computing device needs to be determined. Understandably, this occurs when the computing device is in the next dynamic acquisition cycle. At this time, it is equivalent to the corresponding components (CPU, GPU, memory, I / O devices, etc.) being in the next acquisition cycle. In this embodiment, the dynamic acquisition period is... The specific determination process is not restricted.

[0076] Based on the above embodiments, in some embodiments, the liquid-cooled distribution unit controls the total coolant flow rate according to the sum of the coolant flow rates corresponding to each computing device, including: S131: Obtain the baseline period, global redundancy coefficient, computing device priority coefficient, and dynamic acquisition period corresponding to each computing device; S132: Determine the total coolant flow rate based on the baseline period, global redundancy coefficient, computing device priority coefficient, dynamic acquisition period of each computing device, and total coolant flow rate.

[0077] The CDU calculates the total coolant flow required for the next cycle based on the dynamic flow demand reported by different computing devices within the cycle. This can be combined with dynamic load forecasting, redundancy design, and real-time feedback mechanisms. The specific formula is as follows: ; in, This is the total coolant flow rate. The dynamic acquisition period corresponding to the i-th computing device (s) is predicted in real time by load volatility (such as the rate of change in GPU memory utilization). A baseline period (e.g., 1 second) is used to standardize traffic demand for different periods. This represents the total flow rate required by each computing device in the next dynamic acquisition cycle, i.e., the total coolant flow rate. This is a global redundancy factor (recommended 10%~20%) to compensate for the cumulative effect of load fluctuations across multiple servers. Prioritize critical servers (0.1~0.3) and reserve extra bandwidth for high-priority tasks (such as AI training clusters).

[0078] It should be noted that the above formula can be used to calculate the total coolant flow rate for each computing device. According to its cycle Adjust to the base period .like In high-frequency monitoring scenarios, it is necessary to increase the instantaneous throughput to cope with rapid load changes, such as a sudden surge in AI training tasks; if In low-frequency monitoring scenarios, instantaneous flow can be reduced, but the cooling duration needs to be extended.

[0079] In this embodiment, the total flow rate of the CDU in the next cycle is dynamically calculated to meet the cooling needs of multiple computing devices, thereby achieving fine-grained control over the cooling flow rate of each computing device.

[0080] Based on the above embodiments, in some embodiments, after determining the total coolant flow rate, the method further includes: S133: Obtain the flow deviation record, response sensitivity coefficient, baseline period, global redundancy coefficient, and baseline flow rate within the historical period.

[0081] S134: Determine the redundant flow rate based on the flow deviation record, response sensitivity coefficient, reference period, global redundancy coefficient, reference flow rate, and the sum of the dynamic acquisition period and coolant flow rate of each computing device.

[0082] S135: Redundant correction of total coolant flow rate based on redundant flow rate.

[0083] To avoid insufficient coolant flow when the load suddenly increases and a sudden increase in coolant flow is required, after determining the total coolant flow rate, it is necessary to further obtain historical flow deviation records, response sensitivity coefficients, reference periods, global redundancy coefficients, and reference flow rates. Based on the flow deviation records, response sensitivity coefficients, reference periods, global redundancy coefficients, reference flow rates, and the sum of the acquisition periods for each computing device and the total coolant flow rate, the redundant flow rate is determined. The specific formula is as follows: ; ; in, For redundant traffic; This records the flow deviation over a historical period. This is the response sensitivity coefficient. The dynamic acquisition period corresponding to the i-th computing device (s); It serves as a baseline period for standardizing traffic demand across different periods; The required flow rate for each computing device in the next cycle; The global redundancy factor (recommended 10%~20%) (20%) to compensate for the cumulative effect of load fluctuations across multiple servers. The baseline flow rate is the average standard flow rate within the baseline period.

[0084] Finally, the total coolant flow rate is redundantly adjusted based on the redundant flow rate. This avoids the problem of insufficient coolant flow when the load suddenly increases and a sudden increase in coolant flow rate is required.

[0085] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0086] Figure 4 This is a schematic diagram of a liquid cooling dynamic flow control device provided in an embodiment of the present invention. The device is applied to the main control module of a secondary liquid cooling distribution unit; wherein, the liquid cooling system includes multiple secondary liquid cooling distribution units, each corresponding to a computing device; each secondary liquid cooling distribution unit also includes a flow control component; the main control module is communicatively connected to the corresponding flow control component and the components of the corresponding computing device; the flow control component, the cold plate piping of the corresponding computing device, and the distributor are connected via piping; as shown... Figure 4 As shown, the device includes: The acquisition module 10 is used to determine the acquisition cycle of each component of the corresponding computing device, and to acquire the load information of the corresponding component according to each acquisition cycle. The first determining module 11 is used to determine the required coolant flow rate of each component in the corresponding next acquisition cycle based on the load information. The second determining module 12 is used to sum up the flow rates of each coolant to determine the total coolant flow rate required by the corresponding computing device. The control module 13 is used to control the coolant flow of the corresponding flow control component according to the total coolant flow, and to report the total coolant flow to the liquid cooling distribution unit so that the liquid cooling distribution unit can control the total coolant flow according to the total coolant flow of each computing device.

[0087] In some embodiments, the acquisition module 10 includes: The first determining submodule is used to determine the initial acquisition cycle of each component; wherein, each initial acquisition cycle is the same; the component includes at least a central processing unit, memory, input / output devices, and a graphics processing unit; The first acquisition submodule is used to acquire the initial load information of each component according to the initial acquisition cycle; The second determination submodule is used to determine the acquisition cycle of the corresponding device component based on the initial load information.

[0088] In some embodiments, the second determining submodule includes: The third determination submodule is used to determine the load standard deviation and current load change rate of the corresponding component based on the initial load information. The first acquisition submodule is used to acquire the basic period, dynamic amplification factor, and component weight values ​​of each component. The fourth determination submodule is used to determine the acquisition cycle of each component based on the base cycle, dynamic amplification factor, component weight value of each component, load standard deviation and current load change rate.

[0089] In some embodiments, the first determining module 11 includes: The fifth determination submodule is used to determine the heat generated by each component in the next acquisition cycle based on the load information. The sixth determination submodule is used to determine the required coolant flow rate for each component in the next acquisition cycle based on the heat.

[0090] In some embodiments, the fifth determining submodule includes: The CPU parameter determination module is used to determine the first static power consumption, first dynamic power consumption, maximum utilization, nonlinear correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification coefficient, and first liquid cooling efficiency factor corresponding to the central processing unit. The CPU heat determination module is used to determine the heat generated by the central processing unit in the next acquisition cycle based on the first static power consumption, first dynamic power consumption, maximum utilization, nonlinear correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification coefficient, first liquid cooling heat dissipation efficiency factor and the corresponding acquisition cycle. The memory parameter determination module is used to determine the second static power consumption, memory utilization, utilization amplification factor, second dynamic power consumption, page switching frequency, switching frequency nonlinearity coefficient, basic switching power consumption, and second liquid cooling heat dissipation efficiency factor corresponding to the memory. The memory heat determination module is used to determine the heat generated by the memory in the next acquisition cycle based on the memory's second static power consumption, memory utilization rate, utilization rate amplification factor, second dynamic power consumption, page switching frequency, switching frequency nonlinearity coefficient, basic switching power consumption, second liquid cooling heat dissipation efficiency factor and the corresponding acquisition cycle. The IO device parameter determination module is used to determine the third static power consumption, third dynamic power consumption, and third liquid cooling efficiency factor corresponding to the input and output devices. The IO device heat determination module is used to determine the heat generated by the input / output device in the next acquisition cycle based on the third static power consumption, third dynamic power consumption, third liquid cooling heat dissipation efficiency factor and the corresponding acquisition cycle of the input / output device. The GPU parameter determination module is used to determine the fourth static power consumption, the fourth dynamic power consumption, and the fourth liquid cooling efficiency factor corresponding to the graphics processing unit. The GPU heat determination module is used to determine the heat generated by the graphics processing unit in the next acquisition cycle based on the fourth static power consumption, the fourth dynamic power consumption, the fourth liquid cooling efficiency factor and the corresponding acquisition cycle of the graphics processing unit.

[0091] In some embodiments, the CPU parameter determination module includes: The second acquisition submodule is used to acquire the load capacitance, voltage, frequency and current utilization of the central processing unit; The seventh determination submodule is used to determine the first dynamic power consumption of the central processing unit based on the load capacitance, voltage, frequency and current utilization of the central processing unit. The third acquisition submodule is used to acquire the sensor temperature, heat transfer coefficient, load intensity parameters and average temperature difference under load intensity set in the central processing unit; The eighth determination submodule is used to determine the core temperature of the central processing unit based on the sensor temperature, heat transfer coefficient, load intensity parameters, and average temperature difference of the load intensity. The fourth acquisition submodule is used to acquire the actual thermal resistance, design thermal resistance, and optimization parameters of the central processing unit; The ninth determination submodule is used to determine the first liquid cooling efficiency factor of the central processing unit based on the actual thermal resistance, design thermal resistance and optimization parameters of the central processing unit.

[0092] In some embodiments, the memory parameter determination module includes: The fifth submodule is used to obtain the total memory capacity, free memory capacity, and used memory capacity. The tenth determination submodule is used to determine the memory utilization rate based on the total memory capacity, free memory capacity, and used memory capacity. The sixth acquisition submodule is used to acquire the memory temperature, coolant inlet temperature, reference temperature difference, actual thermal resistance, design thermal resistance and correction factor corresponding to the memory. The eleventh determination submodule is used to determine the second liquid cooling efficiency factor corresponding to the memory based on the memory temperature, coolant inlet temperature, reference temperature difference, actual thermal resistance, design thermal resistance, and correction factor.

[0093] In some embodiments, the IO device parameter determination module includes: The seventh acquisition submodule is used to acquire the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference and temperature coefficient of the input and output devices. The twelfth determination submodule is used to determine the third static power consumption of the input / output device based on the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference and temperature coefficient of the input / output device. The eighth acquisition submodule is used to acquire the power consumption coefficient, read / write rate, input / output queue depth, device busy rate, and busy rate nonlinearity coefficient corresponding to the input / output device; The thirteenth determination submodule is used to determine the third dynamic power consumption of the input and output devices based on the power consumption coefficient, read and write rate, input and output queue depth, device busy rate and busy rate nonlinearity coefficient of the input and output devices. The ninth acquisition submodule is used to acquire the actual thermal resistance, design thermal resistance, and correction factor of the input and output devices. The fourteenth determination submodule is used to determine the third liquid cooling efficiency factor corresponding to the input and output devices based on the actual thermal resistance, design thermal resistance, correction factor, core temperature, coolant inlet temperature and reference temperature difference of the input and output devices.

[0094] In some embodiments, the GPU parameter determination module includes: The tenth acquisition submodule is used to acquire the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference and temperature coefficient corresponding to the graphics processing unit; The fifteenth determination submodule is used to determine the fourth static power consumption of the input / output device based on the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference and temperature coefficient corresponding to the graphics processing unit. The eleventh acquisition submodule is used to acquire the video memory utilization rate, computing unit utilization rate, task queue length, video memory nonlinearity coefficient, task queue nonlinearity coefficient and power consumption coefficient corresponding to the graphics processing unit. The sixteenth determination submodule is used to determine the fourth dynamic power consumption of the input / output device based on the graphics processing unit's memory utilization, computing unit utilization, task queue length, memory nonlinearity coefficient, task queue nonlinearity coefficient, and power consumption coefficient. The twelfth acquisition submodule is used to acquire the actual thermal resistance, design thermal resistance, correction factor, actual coolant flow rate, design flow rate and flow compensation factor corresponding to the graphics processing unit. The seventeenth determination submodule is used to determine the fourth liquid cooling heat dissipation efficiency factor corresponding to the graphics processing unit based on the actual thermal resistance, design thermal resistance, correction coefficient, actual coolant flow rate, design flow rate, and flow compensation coefficient corresponding to the graphics processing unit.

[0095] In some embodiments, the sixth determining submodule includes: The eighteenth determination submodule is used to determine the theoretical coolant flow rate required by each component in the next acquisition cycle based on the laws of conservation of heat and thermodynamic energy. The thirteenth acquisition submodule is used to acquire the flow redundancy coefficient, pipeline resistance correction coefficient, total pipeline resistance value, and design resistance target value; The nineteenth determination submodule is used to determine the actual coolant flow required by each component in the next acquisition cycle based on the flow redundancy coefficient, pipeline resistance correction coefficient, total pipeline resistance value, design resistance target value, and theoretical coolant flow rate. The setting module is used to set the target speed of the water pump and the opening degree of the flow control valve of the flow control component according to the actual coolant flow rate. The twentieth determination submodule is used to determine the required coolant flow rate for each component in the next acquisition cycle based on the actual coolant flow rate, the target speed of the water pump, and the opening degree of the flow regulating valve.

[0096] In some embodiments, the liquid-cooled distribution unit controls the total coolant flow rate based on the sum of coolant flow rates corresponding to each computing device, including: obtaining a reference period, a global redundancy coefficient, a computing device priority coefficient, and a dynamic acquisition period corresponding to each computing device; and determining the total coolant flow rate based on the reference period, the global redundancy coefficient, the computing device priority coefficient, the dynamic acquisition period corresponding to each computing device, and the sum of coolant flow rates.

[0097] In some embodiments, the liquid-cooled distribution unit also acquires flow deviation records, response sensitivity coefficients, reference periods, global redundancy coefficients, and reference flow rates within historical periods; determines redundant flow rates based on the flow deviation records, response sensitivity coefficients, reference periods, global redundancy coefficients, reference flow rates, and the sum of dynamic acquisition periods and coolant flow rates corresponding to each computing device; and performs redundancy correction on the total coolant flow rate based on the redundant flow rates.

[0098] For a description of the features in the embodiment of the liquid cooling dynamic flow control device, please refer to the relevant description of the embodiment of the liquid cooling dynamic flow control method, which will not be repeated here.

[0099] Embodiments of the present invention also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above embodiments of the liquid cooling dynamic flow control method.

[0100] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above embodiments of the liquid cooling dynamic flow control method when running.

[0101] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0102] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the liquid cooling dynamic flow control method.

[0103] Embodiments of the present invention also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above embodiments of the liquid cooling dynamic flow control method.

[0104] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0105] The present invention provides a detailed description of a liquid cooling dynamic flow control method, device, medium, and product. Specific examples have been used to illustrate the principles and implementation methods of the invention. These examples are merely illustrative to aid in understanding the method and core concepts of the invention. It should be noted that those skilled in the art can make various improvements and modifications to the invention without departing from its principles, and these improvements and modifications also fall within the scope of protection of the claims.

Claims

1. A method for dynamic flow control of liquid cooling heat dissipation, characterized in that, A main control module is applied to a secondary liquid-cooled distribution unit; wherein, the liquid-cooled heat dissipation system includes multiple secondary liquid-cooled distribution units, each secondary liquid-cooled distribution unit corresponding to a computing device; each secondary liquid-cooled distribution unit further includes a flow control component; the main control module is communicatively connected to the corresponding flow control component and the corresponding components of the computing device; the flow control component, the cold plate piping of the corresponding computing device, and the liquid distributor are connected via piping; the method includes: Determine the acquisition cycle of each component of the corresponding computing device, and acquire the load information of the corresponding component according to each acquisition cycle; Determine the required coolant flow rate for each component in the next acquisition cycle based on the load information described above. The coolant flow rates of each item are summed to determine the total coolant flow rate required by the corresponding computing device. The total coolant flow rate is controlled by the corresponding flow control component based on the total coolant flow rate, and the total coolant flow rate is reported to the liquid cooling distribution unit so that the liquid cooling distribution unit can control the total coolant flow rate based on the total coolant flow rate of each computing device. Determine the acquisition cycle of each component of the corresponding computing device, including: The initial acquisition period for each of the aforementioned components is determined; wherein, the initial acquisition periods for each component are the same; the aforementioned components include at least a central processing unit, memory, input / output devices, and a graphics processing unit; The initial load information of each component is collected according to the initial acquisition period; The acquisition cycle of the corresponding device component is determined based on the initial load information of each component; Determining the acquisition cycle of the corresponding device component based on the initial load information includes: The load standard deviation and current load change rate of the corresponding device component are determined based on the initial load information of each component. Obtain the basic period, dynamic amplification factor, and component weight values ​​of each of the device components; The acquisition period of each device is determined based on the base period, the dynamic amplification factor, the component weight value of each device, the load standard deviation, and the current load change rate. The liquid cooling distribution unit controls the total coolant flow rate based on the sum of the coolant flow rates corresponding to each computing device, including: Obtain the baseline period, global redundancy coefficient, computing device priority coefficient, and dynamic acquisition period for each computing device; The total coolant flow rate is determined based on the baseline period, the global redundancy coefficient, the computing device priority coefficient, the dynamic acquisition period corresponding to each computing device, and the sum of the coolant flow rates.

2. The liquid cooling heat dissipation dynamic flow control method according to claim 1, characterized in that, Determining the required coolant flow rate for each component in the next acquisition cycle based on the load information includes: The heat generated by each component in the next acquisition cycle is determined based on the load information. The required coolant flow rate for each of the aforementioned components in the corresponding next acquisition cycle is determined based on the heat generated.

3. The liquid cooling heat dissipation dynamic flow control method according to claim 2, characterized in that, Determining the heat generated by each component in the corresponding next acquisition cycle based on the load information includes: Determine the first static power consumption, first dynamic power consumption, maximum utilization, nonlinearity correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification coefficient, and first liquid cooling efficiency factor corresponding to the central processing unit. Based on the first static power consumption, first dynamic power consumption, maximum utilization, nonlinear correction coefficient, core temperature, ambient temperature, reference temperature difference, temperature amplification coefficient, first liquid cooling efficiency factor, and the corresponding acquisition cycle, determine the heat generated by the central processing unit in the next acquisition cycle. Determine the second static power consumption, memory utilization, utilization amplification factor, second dynamic power consumption, page switching frequency, switching frequency nonlinearity coefficient, basic switching power consumption, and second liquid cooling efficiency factor corresponding to the memory. Based on the second static power consumption, memory utilization, utilization amplification factor, second dynamic power consumption, page switching frequency, switching frequency nonlinearity coefficient, basic switching power consumption, second liquid cooling heat dissipation efficiency factor and the corresponding acquisition cycle, determine the heat generated by the memory in the next acquisition cycle. Determine the third static power consumption, the third dynamic power consumption, and the third liquid cooling efficiency factor for the input and output devices; Based on the third static power consumption, third dynamic power consumption, third liquid cooling efficiency factor and the corresponding acquisition cycle of the input / output device, determine the heat generated by the input / output device in the next acquisition cycle. Determine the fourth static power consumption, fourth dynamic power consumption, and fourth liquid cooling efficiency factor corresponding to the graphics processing unit; Based on the fourth static power consumption, fourth dynamic power consumption, fourth liquid cooling efficiency factor, and the corresponding acquisition cycle of the graphics processing unit, the heat generated by the graphics processing unit in the next acquisition cycle is determined.

4. The liquid cooling heat dissipation dynamic flow control method according to claim 3, characterized in that, Determine the central processing unit's first dynamic power consumption, core temperature, and first liquid cooling efficiency factor, including: Obtain the CPU's load capacitance, voltage, frequency, and current utilization. The first dynamic power consumption of the central processing unit is determined based on the load capacitance, voltage, frequency and current utilization of the central processing unit. Obtain the sensor temperature, heat transfer coefficient, load intensity parameters, and average temperature difference under load intensity, which are set in the central processing unit; The core temperature of the central processing unit is determined based on the sensor temperature, heat transfer coefficient, load intensity parameters, and average temperature difference under load intensity. Obtain the actual thermal resistance, design thermal resistance, and optimization parameters of the central processing unit; The first liquid cooling efficiency factor of the central processing unit is determined based on the actual thermal resistance, design thermal resistance, and optimization parameters of the central processing unit.

5. The liquid cooling heat dissipation dynamic flow control method according to claim 3, characterized in that, Determine the memory utilization rate and the second liquid cooling efficiency factor corresponding to the memory, including: Get the total memory capacity, free memory capacity, and used memory capacity; Determine the memory utilization rate based on the total memory capacity, free memory capacity, and used memory capacity; Obtain the memory temperature, coolant inlet temperature, reference temperature difference, actual thermal resistance, design thermal resistance, and correction factor corresponding to the memory. The second liquid cooling efficiency factor for the memory is determined based on the memory temperature, coolant inlet temperature, reference temperature difference, actual thermal resistance, design thermal resistance, and correction factor.

6. The liquid cooling heat dissipation dynamic flow control method according to claim 3, characterized in that, Determine the third static power consumption, third dynamic power consumption, and third liquid cooling efficiency factor for the input / output devices, including: Obtain the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference, and temperature coefficient corresponding to the input and output devices; The third static power consumption of the input / output device is determined based on the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference, and temperature coefficient of the input / output device. Obtain the power consumption coefficient, read / write rate, input / output queue depth, device busy rate, and busy rate nonlinearity coefficient corresponding to the input / output devices; The third dynamic power consumption of the input / output device is determined based on the power consumption coefficient, read / write rate, input / output queue depth, device busy rate, and busy rate nonlinearity coefficient of the input / output device. Obtain the actual thermal resistance, design thermal resistance, and correction factor for the input and output devices; The third liquid cooling efficiency factor for the input / output devices is determined based on the actual thermal resistance, design thermal resistance, correction factor, core temperature, coolant inlet temperature, and reference temperature difference.

7. The liquid cooling heat dissipation dynamic flow control method according to claim 3, characterized in that, Determine the fourth static power consumption, fourth dynamic power consumption, and fourth liquid cooling efficiency factor corresponding to the graphics processing unit, including: Obtain the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference, and temperature coefficient corresponding to the graphics processing unit; The fourth static power consumption of the graphics processing unit is determined based on the basic static power consumption, core temperature, coolant inlet temperature, reference temperature difference, and temperature coefficient of the graphics processing unit. Obtain the memory utilization, computing unit utilization, task queue length, memory nonlinearity coefficient, task queue nonlinearity coefficient, and power consumption coefficient corresponding to the graphics processing unit; The fourth dynamic power consumption of the graphics processing unit is determined based on the memory utilization, computing unit utilization, task queue length, memory nonlinearity coefficient, task queue nonlinearity coefficient, and power consumption coefficient corresponding to the graphics processing unit. Obtain the actual thermal resistance, design thermal resistance, correction factor, actual coolant flow rate, design flow rate, and flow compensation factor corresponding to the graphics processing unit; The fourth liquid cooling efficiency factor for the graphics processing unit is determined based on the actual thermal resistance, design thermal resistance, correction factor, actual coolant flow rate, design flow rate, and flow compensation factor corresponding to the graphics processing unit.

8. The liquid cooling heat dissipation dynamic flow control method according to claim 3, characterized in that, Determining the required coolant flow rate for each of the aforementioned components in the corresponding next acquisition cycle based on the heat generated includes: The theoretical coolant flow rate required for each of the aforementioned components in the next corresponding acquisition cycle is determined based on the aforementioned heat and the thermodynamic energy conservation law. Obtain the flow redundancy coefficient, pipeline resistance correction coefficient, total pipeline resistance value, and design resistance target value; Based on the flow redundancy coefficient, the pipeline resistance correction coefficient, the total pipeline resistance value, the design resistance target value, and the theoretical coolant flow rate of each component, the actual coolant flow rate required by each component in the next acquisition cycle is determined. The target speed of the water pump and the opening degree of the flow control component are set according to the actual coolant flow rate; Based on the actual coolant flow rate, the target speed of the water pump, and the opening degree of the flow regulating valve, the required coolant flow rate for each component in the next corresponding acquisition cycle is determined.

9. The liquid cooling dynamic flow control method according to claim 1, characterized in that, After determining the total coolant flow rate, the process also includes: Obtain historical data on traffic deviation, response sensitivity coefficient, baseline period, global redundancy coefficient, and baseline traffic. The redundant flow rate is determined based on the flow deviation record, response sensitivity coefficient, reference period, global redundancy coefficient, reference flow rate, dynamic acquisition period corresponding to each computing device, and the total coolant flow rate. The total coolant flow rate is redundantly corrected based on the redundancy flow rate.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the liquid cooling dynamic flow control method as described in any one of claims 1 to 9 when executing the computer program.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the liquid cooling dynamic flow control method as described in any one of claims 1 to 9.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the liquid cooling dynamic flow control method as described in any one of claims 1 to 9.

Images