Liquid cooling server intelligent temperature control method based on local software monitoring and liquid cooling server

Abstract

The invention relates to the technical field of server temperature control, and discloses a liquid cooling server intelligent temperature control method based on local software monitoring, and the method comprises the following steps: data collection: collecting hardware temperature data and task load data of a liquid cooling server through an embedded real-time monitoring module; temperature prediction: based on the hardware temperature data and the task load data, utilizing a prediction algorithm to analyze the temperature change trend of the server in a period of time in the future; regulation and control decision making: according to the temperature change trend, generating regulation and control parameters for regulating the flow of the cooling liquid and the target temperature of the cooling liquid; the invention further discloses a liquid cooling server which comprises an embedded real-time monitoring module, a distributed control unit, a cooling loop module and a local off-line temperature control module. Through the dynamic temperature control algorithm, the distributed control unit and the local temperature control module, real-time accurate regulation and control, efficient response and independent operation under the condition of network disconnection are achieved, and the reliability and adaptability of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of server temperature control, and in particular to an intelligent temperature control method for a liquid-cooled server based on local software monitoring and a liquid-cooled server. Background Art

[0002] With the advent of the information age, the rapid development of cloud computing, big data, and artificial intelligence technologies has put forward higher requirements on the computing power of data centers. The deployment of a large number of high-density servers has made the heat dissipation problem more severe. Traditional air cooling technology has been unable to meet the needs of modern servers due to its high energy consumption and low heat dissipation efficiency. Liquid cooling technology has gradually become an important choice for data centers due to its efficient heat dissipation capacity and energy-saving effect. By directly contacting the heat source with the coolant, liquid cooling technology can effectively reduce the operating temperature of the server and improve the stability of the system and the life of the hardware.

[0003] Although existing liquid cooling technology has developed to a certain extent, it is mostly focused on solving heat dissipation efficiency and basic cooling problems. Through efficient heat exchange between coolant and hardware, existing technology can significantly reduce server temperature and has a higher energy efficiency ratio than traditional air cooling technology. On the other hand, the noise level of the liquid cooling system is low, which has obvious advantages in scenarios that require silent operation. In addition, some liquid cooling systems achieve cooling optimization under different load conditions through basic temperature control adjustments, meeting the general needs of medium-density data centers.

[0004] However, existing technologies still have significant deficiencies when dealing with complex operating scenarios. First, most existing technologies use fixed parameters to control coolant flow and target temperature, which makes it difficult to dynamically adapt to the needs of load changes, resulting in frequent overcooling or overheating. Secondly, existing technologies mainly trigger regulation through real-time hardware temperature, lack prediction of temperature rise trends, and often take measures only when the temperature approaches the critical value. The regulation response is delayed, which can easily cause hardware performance fluctuations or thermal damage. In addition, the system with a centralized control architecture has limited response speed in high-density server scenarios, is prone to delays and bottlenecks, and cannot meet complex temperature control requirements. Finally, existing technologies do not provide sufficient support for offline scenarios and rely on remote control systems. It is difficult to achieve temperature control functions in offline situations, which seriously restricts the application of liquid cooling technology in unmanned sites or edge computing scenarios. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an intelligent temperature control method for a liquid-cooled server based on local software monitoring and a liquid-cooled server, which solves the problems in the prior art that the temperature control system of the liquid-cooled server cannot dynamically adapt to load changes, lacks temperature rise trend prediction, has a slow response speed of the centralized control architecture, and does not provide sufficient support for offline scenarios.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a liquid cooling server intelligent temperature control method based on local software monitoring, comprising the following steps: Data collection: The hardware temperature data and task load data of the liquid-cooled server are collected through the embedded real-time monitoring module; Temperature prediction: Based on the hardware temperature data and task load data, a prediction algorithm is used to analyze the temperature change trend of the server in the future; Control decision: generating control parameters for adjusting the coolant flow rate and the coolant target temperature according to the temperature change trend; Execution control: receiving the control parameters through a distributed control unit and independently executing the adjustment of the coolant flow rate and the target temperature; Fault detection and circuit switching: Real-time detection of the operating status of the cooling circuit, and automatic switching to the backup cooling circuit when a cooling circuit fails; Offline operation support: In the server offline scenario, data collection, temperature prediction, regulation decision-making and execution control are completed through the locally deployed temperature control module.

[0007] Preferably, in the data collection step, the following data are monitored in real time: Temperature data of key server hardware, including the temperature of CPU, GPU, and memory; Server task load information, including CPU usage, GPU queue length, and memory bandwidth usage.

[0008] Preferably, the temperature prediction step comprises the following steps: Preprocessing the hardware temperature data and task load data collected in real time, wherein the preprocessing includes noise removal, normalization processing, and feature extraction based on time series; The preprocessed data is input into a temperature prediction model, which is built based on a time series algorithm and combines historical operation data to learn the temperature rise characteristics of the server; Predicting the temperature change trend within a preset time interval in the future, the trend including the direction of temperature change and the magnitude of temperature change; The prediction results are provided to the control decision module for adjusting the coolant flow and target temperature in advance.

[0009] Preferably, in the control decision step, the control parameters of the coolant flow rate and the target temperature are calculated according to the predicted temperature change trend and the safe temperature range of the server, wherein the control parameters meet the following conditions: The coolant flow rate tends to increase as the temperature rises; The target temperature of the coolant is dynamically adjusted according to the direction of temperature change to keep the server operating temperature within the safe temperature range.

[0010] Preferably, in the execution control step, each server independently completes the adjustment of the coolant flow rate and the target temperature through a distributed control unit, and the distributed control unit includes: A cooling pump regulating module for regulating the coolant flow rate; The temperature regulating module is used to adjust the heat exchange efficiency of the coolant.

[0011] Preferably, the fault detection and circuit switching steps include: The flow rate of the coolant is monitored in real time through the flow sensor in the cooling circuit. When the flow rate is detected to be lower than the preset threshold, the system determines that the cooling circuit has a fault; Verify the fault condition by combining temperature sensor data, including detecting whether abnormal coolant flow rate causes server temperature to exceed the limit, to ensure the accuracy of fault judgment; After confirming the cooling circuit failure, sending a start command to the standby cooling circuit, the standby cooling circuit is connected to the main cooling circuit through an electronically controlled valve; Perform circuit switching operation, the electronically controlled valve closes the main cooling circuit and opens the standby cooling circuit to ensure the continuity of coolant flow; After the switch is completed, the system monitors the operating status of the backup cooling circuit in real time, including flow rate and temperature, to ensure that the backup circuit operates normally and meets the heat dissipation requirements of the server; The completion time of the loop switching operation does not exceed 100 milliseconds to avoid overheating of the server due to a malfunction.

[0012] Preferably, the performing loop switching operation comprises the following steps: When a failure occurs in the main cooling circuit, the system automatically closes the coolant flow channel of the main cooling circuit and simultaneously opens the coolant flow channel of the backup cooling circuit; The switching is achieved by an electrically controlled valve arranged on the cooling circuit, and the opening and closing actions of the electrically controlled valve are controlled by a switching instruction triggered by a flow sensor; During the switching process, the system maintains the continuity of the coolant flow rate to avoid overheating of the server due to flow interruption; After the switch is completed, the system monitors the flow rate and cooling effect of the backup circuit in real time to determine whether the backup circuit is operating normally, and further adjusts the cooling parameters based on the monitoring results; If the backup circuit fails to work properly, the system will send out an alarm signal and can choose to reduce the task load of the server to reduce the cooling demand; The backup circuit is in a standby state during normal operation and is connected to the main circuit through a shared coolant circulation pipeline to reduce system resource occupation.

[0013] Preferably, in the offline operation support step, the temperature control module independently completes the following operations through local embedded software when the server and external communication are interrupted: Data collection; Temperature forecast; Regulatory decision-making; Execution control.

[0014] The present invention also provides a liquid cooling server, comprising: Embedded real-time monitoring module, used to collect server hardware temperature data and task load data; A local intelligent control module, used to predict the temperature change trend and generate control parameters based on the hardware temperature data and task load data; Distributed control unit for independently performing the regulation of coolant flow and target temperature; A cooling circuit module, comprising a main cooling circuit and a backup cooling circuit, wherein when a failure occurs in the main cooling circuit, the backup cooling circuit is switched to through an electronically controlled valve; The local offline temperature control module is used to implement temperature control operations in offline scenarios.

[0015] Preferably, the cooling circuit module includes a main cooling circuit and a backup cooling circuit, and the connection and switching between the main cooling circuit and the backup cooling circuit are realized by the following devices: Flow sensors are provided at the inlet of the main cooling circuit and the standby cooling circuit to detect abnormal flow rate of the coolant; An electrically controlled valve is arranged on the main cooling circuit and the backup cooling circuit channel, and the electrically controlled valve is controlled by the fault detection module of the server, and is used to close the main circuit and open the backup circuit when a fault occurs in the main cooling circuit; The operating parameter monitoring module of the standby cooling circuit includes a flow rate sensor and a temperature sensor, which is used to monitor the flow rate and cooling effect of the standby circuit in real time and transmit the monitoring results to the control module of the server; A standby module for the backup cooling circuit, which maintains the low-power standby state of the backup circuit when the main circuit is working normally, and immediately restores normal cooling performance after the backup circuit is started; The cooling circuit fault alarm module is used to send out an alarm signal when the backup circuit fails to switch or operates abnormally, so as to prompt the user to take further maintenance measures; The backup cooling circuit can work in conjunction with the main cooling circuit to ensure the continuity and reliability of server heat dissipation by dynamically adjusting the coolant flow rate and target temperature.

[0016] The present invention provides a liquid cooling server intelligent temperature control method based on local software monitoring and a liquid cooling server. It has the following beneficial effects: 1. The present invention adopts a data acquisition technology solution based on an embedded real-time monitoring module, which can efficiently collect hardware temperature data and task load information, and use these data for real-time analysis and temperature control decisions. It achieves the technical effect of improving data acquisition accuracy and real-time performance. Compared with the technical solution that relies on centralized monitoring in the prior art, it solves the shortcomings of large acquisition delay and untimely response to dynamic temperature changes.

[0017] 2. The present invention introduces a dynamic temperature prediction algorithm and uses historical data and real-time collected temperature and load data to establish a time series model, which can predict the temperature rise trend in advance and generate control parameters. This technical solution realizes forward-looking management of the temperature control system and avoids the problem of temperature overshoot or overcooling caused by passive response in the prior art.

[0018] 3. The present invention adopts a distributed control unit to perform the technical solution of coolant flow regulation and target temperature control, which can independently and quickly respond to the dynamic heat dissipation requirements of the server and realize efficient local temperature control. Compared with the centralized control system in the prior art, it solves the shortcomings of slow response speed and easy formation of system bottlenecks, and significantly improves the stability of the cooling system.

[0019] 4. The present invention uses the technical solution of deploying a local temperature control module in the server offline scenario, so that the system can independently complete the whole process of data collection, temperature prediction, regulation and control decision-making, and execution control. The technical effect of not affecting the temperature control capability in the case of network disconnection is achieved, and the deficiency of the prior art that it cannot operate normally in the offline state due to reliance on remote server control is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a schematic diagram of the device of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Please see attached Figure 1 The embodiment of the present invention provides a liquid cooling server intelligent temperature control method based on local software monitoring, comprising the following steps: S1. Data collection: collect the hardware temperature data and task load data of the liquid cooling server through the embedded real-time monitoring module; Data collection is an important basic step in the entire liquid-cooled server intelligent temperature control system. This step uses an embedded real-time monitoring module to continuously collect temperature data and task load data of key hardware of the liquid-cooled server, providing necessary data support for subsequent temperature prediction and control decisions. The real-time and accuracy of data collection directly affects the accuracy of the temperature prediction model and the response speed of the entire temperature control system.

[0023] In general, the hardware temperature and task load of liquid-cooled servers change dynamically. Especially in high-load computing scenarios, the hardware temperature tends to rise rapidly, and the task load tends to show complex fluctuation trends. To this end, the present invention adopts an embedded real-time monitoring module that can collect multiple key data at the same time. These data provide rich information input for subsequent prediction and control steps, ensuring that the temperature control system can accurately respond to dynamic changes in complex scenarios.

[0024] In this embodiment, data collection is completed through an embedded real-time monitoring module. The module integrates multiple types of sensors to collect temperature data of key hardware of the liquid cooling server and current task load data of the server.

[0025] In some embodiments, the embedded real-time monitoring module includes the following hardware: Temperature sensor: used to monitor the real-time temperature of CPU, GPU and memory.

[0026] Flow sensor: used to detect the coolant flow rate so as to analyze the heat dissipation efficiency of the server in combination with temperature data.

[0027] Task load monitoring unit: collects task load information such as CPU occupancy, GPU task queue length, and memory bandwidth usage through the embedded computing module.

[0028] Specifically, the temperature sensor uses a high-precision digital temperature sensor with a measurement accuracy of ±0.1°C, which can quickly respond to instantaneous temperature changes. In one possible implementation, multiple temperature sensors are deployed on each liquid-cooled server, corresponding to the key parts of the CPU, GPU, and memory, to ensure that the collected data covers the entire hardware operating environment.

[0029] As an option, the task load monitoring unit obtains real-time load data by reading the task scheduling information and performance counters of the server. The CPU occupancy rate can be directly collected through the hardware performance counter, the GPU task queue length can be obtained by querying the GPU task manager, and the memory bandwidth usage rate can be obtained through the real-time statistics function of the memory controller. In this way, the monitoring module can fully grasp the current working status of the server.

[0030] Generally, the real-time monitoring module collects data at a fixed sampling frequency. The specific sampling frequency can be set to once per second or higher as required. For example, when the server is running high-density computing tasks, the sampling frequency can be increased to 5 times per second to capture finer-grained temperature and load changes.

[0031] In another embodiment, in order to improve the accuracy of data collection, the embedded module performs preliminary processing on the collected data. The data processing process includes: Data denoising: Filter the sampled data of the temperature sensor to remove high-frequency noise caused by electromagnetic interference or hardware errors.

[0032] Data calibration: The collected data is corrected for deviations through the built-in calibration algorithm to ensure that the data is consistent with the actual physical quantity.

[0033] Data integration: Align the data collected by multiple sensors according to timestamps to form multi-dimensional time series data for subsequent processing.

[0034] In this embodiment, the collected data mainly includes hardware temperature data and task load data. Among them, the hardware temperature data is recorded as ,in Represents hardware components The temperature, represents the sampling time. The task load data is recorded as Represents task load indicators The value of .

[0035] As a possible implementation, the collected data can be represented as the following multidimensional time series:

[0036] in: for Hardware temperature data; for Task load data; is the sampling time.

[0037] In some embodiments, in order to reduce the energy consumption of the data acquisition module, the system can dynamically adjust the sampling frequency. For example, when the server is in a low-load state, the sampling frequency is reduced to once every 10 seconds; and when a high-load task is detected, the sampling frequency is automatically increased to capture more detailed data.

[0038] In addition, the collected data is transmitted to the temperature control module via a lightweight local communication protocol. 2 C bus protocol to ensure real-time data transmission and low power consumption. The communication delay between the acquisition module and the prediction module can be controlled within 1 millisecond.

[0039] In some embodiments, the data acquisition module can be further expanded to support ambient temperature monitoring. By installing an ambient temperature sensor outside the server, the system can better analyze the correlation between the hardware temperature and the external environment. For example, when the ambient temperature is too high, the system can predict the trend of hardware temperature rise, thereby starting a higher intensity cooling solution in advance.

[0040] In another possible implementation, the data acquisition module can also be combined with the cooling circuit status monitoring module to achieve synchronous acquisition of multi-dimensional data. For example, by simultaneously acquiring the coolant flow rate, coolant inlet and outlet temperatures, and hardware surface temperature, the system can accurately analyze the cooling efficiency.

[0041] S2, temperature prediction: Based on hardware temperature data and task load data, use prediction algorithms to analyze the temperature change trend of the server in the future; Temperature prediction is one of the core steps. It is responsible for analyzing the temperature change trend of the server in the future based on the real-time collected hardware temperature data and task load data, and using the prediction algorithm. The purpose of temperature prediction is to provide forward-looking support for subsequent regulation and control decisions, ensuring that the cooling system can respond to the upcoming temperature rise or drop in advance. Through accurate modeling of the server's working status and precise prediction of temperature changes, the system can achieve a transition from passive response to active regulation, effectively avoiding hardware overheating or waste of cooling resources.

[0042] In general, the temperature change of the server has certain time series characteristics and is highly correlated with the task load fluctuation. The temperature rise of the hardware depends not only on the current task load, but also on the previous operating status. Therefore, in the present invention, the prediction algorithm fully combines the temperature data and the task load data, and introduces historical data to model and analyze future trends. This modeling method can capture complex nonlinear change relationships, especially in scenarios with faster load changes, and shows higher accuracy.

[0043] In this embodiment, the temperature prediction is based on the collected hardware temperature data and task load data, and uses a time series prediction algorithm to analyze the temperature change trend in the future. The prediction algorithm is run through an embedded computing module and can generate a prediction result of temperature rise or temperature drop in real time.

[0044] In some embodiments, the main process of temperature prediction includes four steps: data preprocessing, feature extraction, model building, and real-time prediction. The details are as follows: Data preprocessing: Specifically, the collected temperature data and task load data need to be normalized to improve the quality of model input. Including instantaneous temperature and temperature change rate, task load data Including CPU usage, GPU queue length and memory bandwidth.

[0045] In one possible implementation, data preprocessing uses sliding window technology to process continuous time series data in segments. For example, the data of the past 30 seconds can be used as input segments to form a multidimensional vector The preprocessing also includes denoising operation, which removes high-frequency interference signals through a low-pass filter.

[0046] In some embodiments, the feature extraction of temperature data and task load data combines statistical characteristics and dynamic change trends. For example: Statistical characteristics include the average, maximum, and minimum values ​​of temperature, as well as the peak and fluctuation range of load; Dynamic trend characteristics include the slope of temperature change, the frequency of load change, etc.

[0047] Through feature extraction, complex raw data is compressed into a set of key features, thereby reducing the computational complexity of the prediction model.

[0048] In this embodiment, the temperature prediction model is constructed using a long short-term memory network (LSTM), which is suitable for sequence data with time-dependent characteristics. The LSTM model can remember data changes over a long time range and capture the long-term dependency of temperature changes through hidden layer states.

[0049] As an option, the input of the prediction model is the data vector after preprocessing and feature extraction. , the output is the temperature rise or temperature drop trend in the future .in: It is a multi-dimensional input feature, including historical temperature and load information; Indicates the predicted time point The temperature change; The prediction time interval is usually 10 to 60 seconds.

[0050] In the prediction phase, the model runs continuously in a sliding window manner. At fixed time intervals, the model uses the latest data fragment to generate temperature prediction results. In some embodiments, the prediction results include: Temperature trend (e.g. will the temperature rise or fall); The temperature rise (e.g. the temperature will rise by 5°C in the next 10 seconds).

[0051] Specifically, the prediction formula can be expressed as:

[0052] in: Represents the nonlinear mapping relationship of the prediction model; It is the hardware temperature data; The task load data.

[0053] In one possible implementation, in order to improve the stability of the prediction, the system can combine the results of multiple models for weighted averaging. For example, the output of the LSTM model is fused with the results of the support vector machine (SVM) model to improve the robustness of the prediction.

[0054] In some embodiments, the temperature prediction model may also introduce external environmental parameters as additional inputs. , combined with the hardware temperature and load data , the model can more accurately analyze the server's heat dissipation capacity. For servers in high-temperature environments, the model will predict the accelerating trend of hardware temperature rise in advance and issue higher-intensity control instructions to the cooling system.

[0055] As an option, the prediction model can also dynamically adjust parameters based on the server's task type. For example, for GPU-intensive tasks, the model may give higher weight to GPU temperature data, while for memory-intensive tasks, it may prioritize the changing trend of memory bandwidth.

[0056] S3, control decision: generating control parameters for adjusting the coolant flow rate and the coolant target temperature according to the temperature change trend; In the intelligent temperature control method for liquid-cooled servers, regulation and control decisions are a key link between the preceding and following. Based on the output results of the temperature prediction module, the system needs to generate regulation and control parameters for adjusting the coolant flow rate and the coolant target temperature to ensure that the server hardware temperature always remains within a safe range. The accuracy and real-time performance of regulation and control decisions directly affect the efficiency of the cooling system and the stability of the server. The regulation and control decision module of the present invention combines the temperature change trend of the server, the safe working range, and the response characteristics of the cooling system, and realizes intelligent adjustment of the cooling intensity by dynamically calculating the regulation and control parameters.

[0057] Generally, the safe temperature range of the server is preset. For example, the safe range of the CPU is 35°C to 60°C, and the safe range of the GPU may be slightly different. When the predicted temperature change trend indicates that the hardware may be close to the critical temperature, the control decision module will promptly generate instructions for enhanced cooling. When the temperature gradually drops, the system will reduce the cooling intensity to save resources. This dynamic adjustment method avoids the problem of overcooling or overheating in traditional cooling methods.

[0058] In this embodiment, the control decision module dynamically generates control parameters according to the temperature prediction results. The parameters include the coolant flow rate. and coolant target temperature The core goal of regulating parameter generation is to make the hardware temperature Always maintain a safe working range Inside.

[0059] Specifically, the coolant flow and target temperature The calculation is based on the following logic: In general, the coolant flow rate needs to be positively correlated with the temperature rise and the heat dissipation requirements. It can be calculated by the following formula:

[0060] in: It indicates the predicted temperature rise, that is, the amount of temperature change in the future; is the flow adjustment coefficient, which indicates the flow increment corresponding to the unit temperature rise; The base flow rate value represents the minimum cooling requirement of the hardware in a steady state.

[0061] In some embodiments, and It can be set according to the server model and cooling system characteristics. For example, for a high power consumption server, May take a larger value, and for lightly loaded servers, The value may be smaller.

[0062] Coolant target temperature The setting needs to match the current temperature of the hardware and the predicted temperature rise trend. In this embodiment, the coolant target temperature can be determined by the following formula:

[0063] in: Indicates the current temperature of the hardware; is the target temperature adjustment coefficient, which is used to control the influence of the temperature rise trend on the target temperature.

[0064] As an option, The value can be adjusted dynamically according to the server cooling efficiency. For example, when the coolant flow rate is low, the value can be increased appropriately. value to enhance the heat transfer capacity.

[0065] In one possible implementation, the calculation of the control parameters also takes into account the response time of the cooling system. In order to avoid the impact of frequent coolant flow and temperature adjustments on system stability, the system sets a threshold for changes in the control parameters. For example, when the newly generated coolant flow With current flow When the difference is less than the set threshold, the system will ignore this adjustment.

[0066] In some embodiments, the control decision module generates the control parameters according to the following logic: When the predicted temperature rise is small, e.g. , the system maintains the basic cooling intensity and does not adjust the flow rate and target temperature. When the predicted temperature rise reaches a medium level, such as , the system appropriately increases the coolant flow rate and reduces the coolant target temperature by 1°C to 2°C. When the predicted temperature rise is large, such as , the system significantly increases the coolant flow and reduces the coolant target temperature by at least 3°C.

[0067] Specifically, assuming that the current CPU temperature is 55°C, the safety range is 35°C to 60°C, and the temperature rise is predicted to be 4°C in the next 10 seconds, then: according to the above logic, the system will increase the coolant flow rate from the current 1500 liters / minute to 2000 liters / minute; the coolant target temperature is adjusted from the original 25°C to 22°C.

[0068] In another possible implementation, the control decision module will refine the parameters in combination with the task load information. For example, when the task load prediction indicates that the GPU task will occupy a large amount of computing resources, the system will prioritize the flow rate and target temperature of the GPU cooling loop, while the parameters of the CPU loop remain unchanged.

[0069] In some embodiments, the control decision module can also introduce external environmental parameters as auxiliary references. For example, when the ambient temperature is high, the system may increase the coolant flow rate in advance to compensate for the lack of heat dissipation capacity. In addition, for cooling systems with multi-loop designs, the control decision module can also independently calculate the parameters of different cooling loops. For example, the flow rate and target temperature of the CPU and GPU loops can be adjusted separately to achieve more refined cooling control.

[0070] As an option, the control decision module can also work in conjunction with the server load scheduling module. For example, when the cooling system cannot meet the current heat dissipation requirements, the system can generate a frequency reduction instruction to reduce the task load of the hardware, thereby slowing down the temperature rise.

[0071] S4, execution control: receiving control parameters through the distributed control unit and independently executing the adjustment of the coolant flow rate and the target temperature; Execution control is the key link to achieve the temperature control effect. Based on the control parameters of coolant flow and target temperature generated by the control decision module, execution control independently completes the corresponding coolant flow regulation and target temperature adjustment operations through the distributed control unit; the role of the distributed control unit is to quickly and efficiently convert the control parameters into physical operations to maintain the temperature stability of the liquid-cooled server. In general, each server is equipped with an independent control unit that can autonomously process cooling instructions, thereby avoiding the response delay and load bottleneck problems caused by centralized control.

[0072] Specifically, the execution control module includes a cooling pump adjustment module and a temperature adjustment module. The former is responsible for adjusting the circulation flow rate of the coolant, and the latter dynamically adjusts the coolant temperature by adjusting the heat exchanger parameters. In some embodiments, the control unit can control the flow and temperature of the coolant according to the needs of different hardware components to achieve more fine-grained temperature control management.

[0073] In the embodiment, the distributed control unit receives the coolant flow and target temperature parameters generated by the control decision module, and completes the cooling operation through the built-in control algorithm and execution component. The control unit communicates with the monitoring module and the control module in the server in real time to ensure that the execution process is consistent with the current temperature control requirements.

[0074] In one possible implementation, the coolant flow rate The adjustment is completed by the cooling pump adjustment module. The flow rate adjustment formula of the cooling pump can be expressed as:

[0075] in: is the target flow rate; is the base flow rate, usually the minimum cooling requirement of the server in idle state; is the flow rate that increases dynamically according to the temperature rise trend, ,in is the adjustment coefficient, To predict the temperature rise.

[0076] In general, the cooling pump regulation module adjusts the pump speed according to the above formula. For example, when the predicted temperature rise is small, When the value of is small, the cooling pump only slightly increases the flow rate. If the value of increases significantly, the cooling pump will quickly increase the flow rate to enhance the cooling effect.

[0077] As an option, the temperature control module dynamically adjusts the coolant temperature by adjusting the working parameters of the heat exchanger. Specifically, the heat exchanger can achieve the target temperature by controlling the heat exchange area or the coolant flow time. The temperature regulation formula can be expressed as:

[0078] in: is the coolant target temperature; is the coolant inlet temperature; is the heat transfer efficiency coefficient, which is related to the material, area and other parameters of the heat exchanger; is the heat exchange capacity of the heat exchanger, which is related to the coolant flow rate and heat load.

[0079] In one possible implementation, the temperature control module controls the coolant flow path through the electronically controlled valve, thereby changing the heat exchange area. The system dynamically adjusts the heat exchange path according to the real-time temperature control requirements to ensure that the outlet coolant temperature always meets the requirements of the control decision.

[0080] In some embodiments, the operation flow of executing control is as follows: First, the distributed control unit receives the control parameters and , and compare it with the current system status. For example, when the difference between the current flow rate and the target flow rate is When it is less than the threshold, the system can choose to keep the existing flow rate unchanged.

[0081] The cooling pump regulation module then adjusts the pump speed based on the calculation results. For example, if the target flow The current flow rate is 2000 l / min. If the pump speed is 1500 l / min, the pump speed will be gradually increased at the set acceleration rate to avoid the impact of instantaneous adjustment on the coolant circulation system.

[0082] At the same time, the temperature regulation module starts the heat exchanger adjustment operation. Specifically, the system may achieve a lower coolant outlet temperature by increasing the heat exchanger area through which the coolant flows or extending the flow time.

[0083] In some embodiments, the distributed control unit can also work in conjunction with the server task management module. For example, when a high-load task is predicted to start, the control unit can increase the coolant flow and lower the target temperature in advance to avoid the problem of hardware temperature rising too fast due to delayed temperature control response.

[0084] As a possible implementation method, the distributed control unit can also set up independent control loops for different hardware components. For example, the coolant flow of the GPU and target temperature Cooling parameters can be set independently from the CPU and This partitioned control approach enables more granular cooling management.

[0085] Generally speaking, the execution of distributed control units has a high response speed. For example, from receiving the control parameters to completing the adjustment of the cooling pump and heat exchanger, the time delay of the entire process can be controlled within 50 milliseconds, which is sufficient to meet the needs of dynamic temperature control.

[0086] In some embodiments, the distributed control unit may also integrate a status monitoring module for real-time monitoring of the working status of the cooling pump and the heat exchanger. For example, when an abnormal cooling pump flow rate or a decrease in heat exchanger efficiency is detected, the control unit will issue an alarm signal and try to automatically adjust the relevant parameters to restore normal operation.

[0087] As an option, the control unit can also dynamically adjust cooling parameters according to the ambient temperature. For example, when the ambient temperature is low, the system will reduce the coolant flow rate and target temperature to save energy; when the ambient temperature is high, the system will appropriately increase the cooling intensity to compensate for the lack of heat dissipation capacity.

[0088] S5, fault detection and circuit switching: real-time detection of the operating status of the cooling circuit, and automatic switching to the backup cooling circuit when a cooling circuit fails; Fault detection and circuit switching are important links to ensure the stable operation of the cooling system. When a cooling circuit fails, the system needs to detect the fault in real time and quickly switch to the backup cooling circuit to ensure the normal flow of coolant and avoid abnormal temperature rise of the server due to cooling interruption. This process requires not only accurate detection, but also fast switching operations to minimize the impact of the fault on server operation.

[0089] Generally, the operating status of the cooling circuit is monitored in real time through embedded flow sensors and temperature sensors. The flow sensor is used to detect the flow rate of the coolant. When the flow rate is lower than the preset threshold, the system preliminarily determines whether there is a fault in the cooling circuit. Subsequently, combined with the data from the temperature sensor, it is further verified whether the circuit has failed. As an option, the system can also monitor the operating status and pressure data of the cooling pump to determine the fault condition in multiple dimensions.

[0090] Specifically, after the fault detection is triggered, the system closes the main cooling circuit through the electronically controlled valve and opens the backup cooling circuit at the same time to achieve a rapid switching of the cooling function. During the switching process, the system will continuously monitor the operating status of the backup circuit to ensure that the backup circuit can work normally.

[0091] In this embodiment, the fault detection and circuit switching module includes a flow sensor, a temperature sensor, an electronically controlled valve, and a fault processing logic unit. The module can collect the flow rate and temperature data of the cooling circuit in real time, analyze the operating status, and automatically start the circuit switching operation when a fault is detected.

[0092] Generally, the flow sensor is installed at the inlet of the main cooling circuit to detect the real-time flow rate of the coolant. When the flow rate falls below the preset threshold The system will trigger the fault detection logic. The temperature sensor is installed at the outlet of the coolant to monitor the outlet temperature. In the event of a fault, the outlet temperature usually rises significantly. The system uses a combination of flow rate and temperature to avoid false alarms caused by a single sensor error.

[0093] In some embodiments, the fault judgment formula can be expressed as:

[0094] in: Indicates the fault status (1 for fault, 0 for normal); The flow rate for the main cooling circuit; is the flow rate threshold; = − Indicates the temperature difference of the coolant; is the temperature difference threshold, which is used to determine whether the cooling efficiency has dropped significantly; as an option, The rate of change of the temperature difference can also be used as an auxiliary indicator. When the rate of change of the temperature difference is too fast, the system will accelerate the start of the fault detection process.

[0095] In this embodiment, when the fault detection logic determines that the main cooling circuit has a fault, the system immediately starts the circuit switching operation. The core device of the switching is the electric control valve, and its opening and closing state is controlled by the system.

[0096] Specifically, the system completes the switching operation through the following steps: First, close the inlet valve of the main cooling circuit and open the inlet valve of the backup cooling circuit; then, monitor the flow rate of the backup circuit through the flow sensor. , ensure that the coolant flow returns to normal; finally, monitor the outlet temperature of the backup circuit , verify the cooling effect of the backup circuit.

[0097] In one possible implementation, the system records the failure time of the primary circuit and the startup time of the backup circuit to evaluate the efficiency of the switching process. Generally, the total delay of the circuit switching does not exceed 100 milliseconds.

[0098] In some embodiments, in order to further improve the stability of circuit switching, the system can preheat the backup circuit before starting it. For example, when the backup circuit is in standby mode for a long time, its coolant temperature may be high. The system can precool the backup circuit by circulating low-temperature coolant before a fault occurs to ensure that the backup circuit can provide a good cooling effect immediately after starting.

[0099] As an option, the system can also dynamically adjust the flow rate of the backup loop to adapt to different cooling requirements. For example, when the server task load is low, the flow rate of the backup loop It can be set to a lower value to save energy, while when the load is higher, the flow rate of the backup circuit is automatically increased to increase cooling capacity.

[0100] In some embodiments, the fault detection and circuit switching module can work in conjunction with the server's temperature control decision module. For example, when a cooling circuit failure is detected and the backup circuit cannot be started normally, the system can issue a frequency reduction instruction to the server to temporarily reduce the hardware load to slow down the temperature rise and buy time for maintenance.

[0101] In addition, the system can also be combined with pressure sensors to conduct more comprehensive monitoring of the cooling circuit. For example, when the pressure of the main circuit drops to an abnormal level, the system will determine that there may be a leakage problem and issue a corresponding alarm signal.

[0102] S6. Offline operation support: In the server offline scenario, data collection, temperature prediction, regulation decision and execution control are completed through the locally deployed temperature control module.

[0103] The offline operation support module is one of the core designs to ensure the reliability and independence of the system. In some special scenarios, the server is prone to go offline due to network interruption or external control system failure. In order to ensure the continuous operation of the temperature control function in such cases, the present invention implements the full process of data collection, temperature prediction, regulation and control, and execution control through a locally deployed temperature control module. The local temperature control module can operate completely independently of the external control system to ensure that the hardware temperature of the server in the offline state is always kept within a safe range.

[0104] Generally speaking, the offline operation support module integrates the core functions of the data acquisition module, temperature prediction module, regulation and decision module, and execution control module. These functions are optimized at the hardware and algorithm levels to meet the real-time requirements and operate efficiently in offline scenarios. Specifically, the module adopts an embedded architecture, and all calculations and control logic are completed locally without relying on external servers or cloud systems.

[0105] In this embodiment, the local temperature control module integrates the full-process functions of data collection, temperature prediction, regulation and control decisions, and execution control, and can autonomously complete temperature control operations after the server enters the offline state.

[0106] In some embodiments, the data collection part is implemented by an embedded sensor module. The module can obtain hardware temperature data and task load data in real time, and store the collection results in local memory as input data for subsequent prediction and regulation. Generally, the sampling frequency is set to once per second by default. In one possible implementation, the collected data is stored in the form of a time series, and historical data is cleaned up regularly to free up storage space.

[0107] Specifically, the temperature prediction part runs based on the built-in prediction algorithm. The prediction model uses a trained long short-term memory network (LSTM) to analyze the temperature change trend over a period of time in the future using the current hardware temperature and task load data. As an option, the input of the prediction model includes: Hardware temperature data at the current moment Represents hardware components Real-time temperature; Task load data at the current moment ,in Represents task load indicators .

[0108] Model output is future time interval The temperature rise or drop within In general, The value range is 10 seconds to 60 seconds, and the specific value is adjusted according to the operating characteristics of the server.

[0109] In the control decision part, the local temperature control module generates control parameters according to the predicted temperature change trend, including coolant flow and target temperature The core logic of the control decision is consistent with the online mode, mainly based on the safe temperature range of the server. Perform dynamic calculations. For example, when a large temperature rise is predicted, the system will appropriately increase the coolant flow rate and lower the coolant target temperature to enhance the cooling effect.

[0110] In a possible implementation, the calculation formula of the control parameter is as follows:

[0111]

[0112] in: Indicates coolant flow rate; Indicates the coolant target temperature; , and is the adjustment factor related to the cooling system characteristics; Indicates the predicted temperature rise; Indicates the current hardware temperature.

[0113] The execution control part completes the adjustment of the coolant flow and the target temperature through the distributed control unit. In some embodiments, the local temperature control module is connected through I 2 C or CAN bus communicates with distributed control units to achieve fast and low-power control command transmission. Specifically, the cooling pump regulation module dynamically adjusts the pump speed according to the calculated flow parameters, while the temperature regulation module achieves target temperature control by changing the heat exchanger parameters.

[0114] In some embodiments, the local temperature control module can be combined with the server's task scheduling system to optimize the temperature control strategy in the offline state. For example, when the server task load is low, the system will reduce the coolant flow to save energy. When a high-load task is detected, the system will immediately increase the cooling intensity to prevent the temperature from rising rapidly.

[0115] As an option, the local temperature control module can also locally deploy fault detection and circuit switching functions. When a cooling circuit fails, the module can automatically switch to the backup circuit without waiting for external instructions. This design further improves system stability in offline scenarios.

[0116] In another possible implementation, the module can also dynamically adjust the sampling frequency and calculation frequency to balance real-time performance and energy consumption. For example, when the server is in a low-load state, the system will reduce the sampling frequency and prediction frequency, thereby extending the service life of the local temperature control module.

[0117] In some embodiments, the local temperature control module can combine the data of the external environment sensor to further optimize the temperature control strategy. For example, when the ambient temperature is detected to be high, the system will increase the coolant flow and lower the target temperature in advance to compensate for the decrease in heat dissipation capacity. In addition, the module can also combine the coolant pressure data to adjust the control strategy in real time to ensure the normal operation of the cooling circuit.

[0118] As an option, the module can also support remote fault logging. After the server is back online, the module will upload the fault records during offline operation to an external system for subsequent analysis and maintenance.

[0119] The liquid-cooled server described below and the liquid-cooled server intelligent temperature control method based on local software monitoring described above can refer to each other.

[0120] Please see attached Figure 2The present invention also provides a liquid cooling server, comprising: Embedded real-time monitoring module, used to collect server hardware temperature data and task load data; Local intelligent control module, used to predict temperature change trends and generate control parameters based on hardware temperature data and task load data; Distributed control unit for independently performing the regulation of coolant flow and target temperature; A cooling circuit module, comprising a main cooling circuit and a backup cooling circuit, wherein when a failure occurs in the main cooling circuit, the backup cooling circuit is switched to through an electronically controlled valve; The local offline temperature control module is used to implement temperature control operations in offline scenarios.

[0121] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, which will not be repeated here.

[0122] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent temperature control method for a liquid cooling server based on local software monitoring, characterized in that: The following steps are involved: Data collection: The hardware temperature data and task load data of the liquid-cooled server are collected through the embedded real-time monitoring module; Temperature prediction: Based on the hardware temperature data and task load data, a prediction algorithm is used to analyze the temperature change trend of the server in the future; Control decision: generating control parameters for adjusting the coolant flow rate and the coolant target temperature according to the temperature change trend; Execution control: receiving the control parameters through a distributed control unit and independently executing the adjustment of the coolant flow rate and the target temperature; Fault detection and circuit switching: Real-time detection of the operating status of the cooling circuit, and automatic switching to the backup cooling circuit when a cooling circuit fails; Offline operation support: In the server offline scenario, data collection, temperature prediction, regulation decision-making and execution control are completed through the locally deployed temperature control module.

2. The liquid cooling server intelligent temperature control method based on local software monitoring according to claim 1 is characterized in that: In the data collection step, the following data are monitored in real time: Temperature data of key server hardware, including the temperature of CPU, GPU, and memory; Server task load information, including CPU usage, GPU queue length, and memory bandwidth usage.

3. The liquid cooling server intelligent temperature control method based on local software monitoring according to claim 1 is characterized in that: The temperature prediction step comprises the following steps: Preprocessing the hardware temperature data and task load data collected in real time, wherein the preprocessing includes noise removal, normalization processing, and feature extraction based on time series; The preprocessed data is input into a temperature prediction model, which is built based on a time series algorithm and combines historical operation data to learn the temperature rise characteristics of the server; Predicting the temperature change trend within a preset time interval in the future, the trend including the direction of temperature change and the magnitude of temperature change; The prediction results are provided to the control decision module for adjusting the coolant flow and target temperature in advance.

4. The liquid cooling server intelligent temperature control method based on local software monitoring according to claim 1 is characterized in that: In the control decision step, the control parameters of the coolant flow rate and the target temperature are calculated according to the predicted temperature change trend and the safe temperature range of the server, wherein the control parameters meet the following conditions: The coolant flow rate tends to increase as the temperature rises; The target temperature of the coolant is dynamically adjusted according to the direction of temperature change to keep the server operating temperature within the safe temperature range.

5. The liquid cooling server intelligent temperature control method based on local software monitoring according to claim 1 is characterized in that: In the execution control step, each server independently completes the adjustment of the coolant flow rate and the target temperature through a distributed control unit, and the distributed control unit includes: A cooling pump regulating module for regulating the coolant flow rate; The temperature regulating module is used to adjust the heat exchange efficiency of the coolant.

6. The liquid cooling server intelligent temperature control method based on local software monitoring according to claim 1 is characterized in that: The fault detection and circuit switching steps include: The flow rate of the coolant is monitored in real time through the flow sensor in the cooling circuit. When the flow rate is detected to be lower than the preset threshold, the system determines that the cooling circuit has a fault; Verify the fault condition by combining temperature sensor data, including detecting whether abnormal coolant flow rate causes server temperature to exceed the limit, to ensure the accuracy of fault judgment; After confirming the cooling circuit failure, sending a start command to the standby cooling circuit, the standby cooling circuit is connected to the main cooling circuit through an electronically controlled valve; Perform circuit switching operation, the electronically controlled valve closes the main cooling circuit and opens the standby cooling circuit to ensure the continuity of coolant flow; After the switch is completed, the system monitors the operating status of the backup cooling circuit in real time, including flow rate and temperature, to ensure that the backup circuit operates normally and meets the heat dissipation requirements of the server; The completion time of the loop switching operation does not exceed 100 milliseconds to avoid overheating of the server due to a malfunction.

7. The liquid cooling server intelligent temperature control method based on local software monitoring according to claim 6 is characterized in that: The execution loop switching operation comprises the following steps: When a failure occurs in the main cooling circuit, the system automatically closes the coolant flow channel of the main cooling circuit and simultaneously opens the coolant flow channel of the backup cooling circuit; The switching is achieved by an electrically controlled valve arranged on the cooling circuit, and the opening and closing actions of the electrically controlled valve are controlled by a switching instruction triggered by a flow sensor; During the switching process, the system maintains the continuity of the coolant flow rate to avoid overheating of the server due to flow interruption; After the switch is completed, the system monitors the flow rate and cooling effect of the backup circuit in real time to determine whether the backup circuit is operating normally, and further adjusts the cooling parameters based on the monitoring results; If the backup circuit fails to work properly, the system will send out an alarm signal and can choose to reduce the task load of the server to reduce the cooling demand; The backup circuit is in a standby state during normal operation and is connected to the main circuit through a shared coolant circulation pipeline to reduce system resource occupation.

8. The liquid cooling server intelligent temperature control method based on local software monitoring according to claim 1 is characterized in that: In the offline operation support step, the temperature control module independently completes the following operations through the local embedded software when the server and external communication are interrupted: Data collection; Temperature forecast; Regulatory decision-making; Execution control.

9. A liquid cooling server, using the liquid cooling server intelligent temperature control method based on local software monitoring according to any one of claims 1 to 8, characterized in that: include: Embedded real-time monitoring module, used to collect server hardware temperature data and task load data; A local intelligent control module, used to predict the temperature change trend and generate control parameters based on the hardware temperature data and task load data; Distributed control unit for independently performing the regulation of coolant flow and target temperature; A cooling circuit module, comprising a main cooling circuit and a backup cooling circuit, wherein when a failure occurs in the main cooling circuit, the backup cooling circuit is switched to through an electronically controlled valve; The local offline temperature control module is used to implement temperature control operations in offline scenarios.

10. The liquid cooling server according to claim 9, characterized in that: The cooling circuit module includes a main cooling circuit and a standby cooling circuit, and the connection and switching between the main cooling circuit and the standby cooling circuit are realized by the following devices: Flow sensors are provided at the inlet of the main cooling circuit and the standby cooling circuit to detect abnormal flow rate of the coolant; An electrically controlled valve is arranged on the main cooling circuit and the backup cooling circuit channel, and the electrically controlled valve is controlled by the fault detection module of the server, and is used to close the main circuit and open the backup circuit when a fault occurs in the main cooling circuit; The operating parameter monitoring module of the standby cooling circuit includes a flow rate sensor and a temperature sensor, which is used to monitor the flow rate and cooling effect of the standby circuit in real time and transmit the monitoring results to the control module of the server; A standby module for the backup cooling circuit, which maintains the low-power standby state of the backup circuit when the main circuit is working normally, and immediately restores normal cooling performance after the backup circuit is started; The cooling circuit fault alarm module is used to send out an alarm signal when the backup circuit fails to switch or operates abnormally, so as to prompt the user to take further maintenance measures; The backup cooling circuit can work in conjunction with the main cooling circuit to ensure the continuity and reliability of server heat dissipation by dynamically adjusting the coolant flow rate and target temperature.

Images