Computer board card control method and device, electronic equipment and storage medium

Abstract

The invention discloses a computer board card control method and device, electronic equipment and a storage medium, and belongs to the technical field of BMC management. The method is applied to a BMC controller, the BMC controller controls a computer board card, and the method comprises the following steps: acquiring running state information of the computer board card; extracting target state information from the operation state information according to dependency among data in the operation state information; the target state information is input to an energy consumption control model, prediction control parameters output by the energy consumption control model are obtained, the energy consumption control model is obtained based on a plurality of training samples, and the training samples comprise sample state information and control parameter labels corresponding to the sample state information; and configuring the computer board card based on the predictive control parameter. According to the method, the parameters of the computer board card can be dynamically adjusted to quickly respond to load changes.

Description

Technical Field

[0001] This application belongs to the technical field of BMC management, and particularly relates to a control method, device, electronic device, and storage medium for computer boards. Background Art

[0002] With the rapid development of information technology, the scale of computer hardware devices, especially servers and data centers, has been continuously expanding, and the energy efficiency and stability of the system have become the core concerns. Especially in large-scale computing and high-performance computing environments, how to optimize hardware control parameters to reduce energy consumption, improve system performance and reliability is an important challenge faced by modern information technology.

[0003] In servers and data centers, computer boards usually operate under changing loads and working conditions. The Baseboard Management Controller (BMC) is used for board management of servers. When the board has no workload, it controls the board to enter the low-power mode or sleep mode, and the system will turn off some unnecessary hardware units or reduce their activity frequencies, thereby reducing the overall power consumption. Different application scenarios and workloads require computer hardware to dynamically adjust its operating parameters to ensure that it can meet the computing performance requirements while reducing energy consumption and heat dissipation. How to achieve the above goals through precise control methods is an urgent problem to be solved in computer hardware design and energy efficiency optimization.

[0004] Currently, the management of computer boards based on BMC has the defect of slow response speed, inability to adjust the system state in time, and may lead to too high power consumption when the instantaneous load surges. Summary of the Invention

[0005] This application aims to solve at least one of the technical problems existing in the prior art. For this purpose, this application provides a control method, device, electronic device, and storage medium for computer boards, which can dynamically adjust the parameters of computer boards to quickly respond to load changes.

[0006] In the first aspect, this application provides a control method for computer boards, and the method includes:

[0007] Applied to the BMC controller, the BMC controller controls the computer board, and the method includes:

[0008] Obtain the operating status information of the computer board;

[0009] Extract target status information from the operating status information according to the dependencies between the data in the operating status information;

[0010] Input the target status information into the energy consumption control model to obtain the predicted control parameters output by the energy consumption control model. The energy consumption control model is based on multiple training samples, and the training samples include sample status information and control parameter labels corresponding to the sample status information.

[0011] Configure the computer board based on the predicted control parameters.

[0012] According to an embodiment of the present application, extracting target status information from the operating status information according to the dependencies between data in the operating status information includes:

[0013] Select the operating status information according to the correlation between data or data combinations in the operating status information to determine the first status information.

[0014] Filter out the second status information according to the importance of each data or data combination in the first status information for the energy consumption control of the computer board.

[0015] Perform principal component analysis on the data in the second status information, and perform iterative traversal screening of recursive feature elimination. According to the performance impact of the data or data combinations obtained in each round of screening on the energy consumption control model, adjust the screening criteria until the adjustment degree of the screening criteria is less than the standard threshold, and use the obtained data and data combinations as the target status information.

[0016] According to an embodiment of the present application, the energy consumption control model includes a shared layer, a first specific task layer, and a second specific task layer connected in sequence.

[0017] Before inputting the target status information into the energy consumption control model, the method further includes:

[0018] Obtain multiple training samples, where the training samples include sample status information and corresponding control parameter labels.

[0019] Input the multiple training samples into the energy consumption control model in sequence to pre-train the shared layer and the specific task layer.

[0020] Perform multi-task training on the energy consumption control model through multiple training samples to obtain the trained energy consumption control model.

[0021] According to an embodiment of the present application, the shared layer is used to extract general features from the sample status information in the training samples.

[0022] The first specific task layer includes multiple component modules, which correspond one-to-one to the hardware components of the computer board. The component modules are used to generate hardware control parameters corresponding to the hardware components according to the general features;

[0023] The second specific task layer includes multiple task modules, which correspond one-to-one to the control tasks of the computer board. The task modules are used to optimize and adjust the hardware control parameters according to the control tasks to generate the predictive control parameters.

[0024] According to an embodiment of the present application, the computer board includes a main board and at least one external board extended from the main board. Before obtaining the operation status information of the computer board, the method further includes:

[0025] Obtaining first status information of the main board and the external board;

[0026] Based on the first status information, determining the basic priority of each task of the main board and the external board, and the board heat dissipation capacity and board heat dissipation cycle of the main board and the external board;

[0027] Based on the basic priority of the tasks, the board heat dissipation capacity and the board heat dissipation cycle, performing task and heat dissipation system allocation for the main board and the external board.

[0028] According to an embodiment of the present application, performing task allocation for the main board and the external board based on the basic priority of the tasks, the board heat dissipation capacity and the board heat dissipation cycle includes:

[0029] Based on the basic priority of the tasks and the board heat dissipation capacity, determining the allocation priority of each task on the main board and the external board;

[0030] Based on the workload of each task and the heat dissipation cycle, determining the actual load of the task on the main board and the external board;

[0031] Based on the allocation priority and the actual load, determining the allocation of the task.

[0032] According to an embodiment of the present application, the allocation priority is:

[0033]

[0034] where F ij is the allocation priority of task i on board j; A i is the adjustment priority of task i; C j (t) is the board heat dissipation capacity of board j at time t; Qj The resource utilization rate of board card j;

[0035] The actual load is:

[0036]

[0037] where L ij is the actual load of task i on board card j; W i is the resource occupancy of task i; H j is the heat dissipation cycle of board card j;

[0038] Determining the allocation of the task based on the allocation priority and the actual load includes:

[0039] Constructing a task scheduling model:

[0040]

[0041] The objective function O of the task scheduling model is:

[0042]

[0043] where P i is the basic priority of task i; C j (t) is the board card heat dissipation capacity of board card j at time t; L ij is the resource occupancy of task i on board card j; ∈ is a constant used to prevent division by zero errors; τ is a heat dissipation resource adjustment coefficient; R ij is the heat dissipation requirement of task i on board card j; is the resource occupancy and temperature balance adjustment coefficient; T j is the current temperature of board card j; δ j is the role adjustment factor of board card j;

[0044] Based on the objective function, solve the task scheduling model to obtain the optimal allocation plan of the task and execute it.

[0045] In a second aspect, the present application provides a control device for a computer board card, which is applied to a BMC controller, and the BMC controller controls the computer board card. The device includes:

[0046] An acquisition module, configured to acquire the operating status information of the computer board card;

[0047] A first processing module, configured to extract target status information from the operating status information according to the dependency between data in the operating status information;

[0048] A second processing module, configured to input the target status information into an energy consumption control model to obtain predicted control parameters output by the energy consumption control model. The energy consumption control model is obtained based on a plurality of training samples, and the training samples include sample status information and control parameter labels corresponding to the sample status information;

[0049] A third processing module, configured to configure the computer board based on the predicted control parameters

[0050] In a third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the control method of the computer board as described in the first aspect above is implemented.

[0051] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the control method of the computer board as described in the first aspect above is implemented.

[0052] In a fifth aspect, the present application provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor, and the processor is configured to run a program or an instruction to implement the control method of the computer board as described in the first aspect.

[0053] In a sixth aspect, the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, the control method of the computer board as described in the first aspect above is implemented.

[0054] The additional aspects and advantages of the present application will be partly given in the following description, partly will become obvious from the following description, or will be understood through the practice of the present application.

[0055] The control method, device, electronic device, and storage medium of a computer board provided by the present application have the following beneficial effects compared with the prior art:

[0056] (1) By using the energy consumption control model, predicting the optimal control parameters according to the operating status of the computer board, and adjusting the control parameters in real time, it can automatically adapt to different workloads and environmental conditions, avoid excessive energy consumption and inefficient operation in traditional hardware configurations, reduce power consumption without affecting system performance, reduce energy waste, effectively optimize the energy efficiency of the computer board, effectively avoid overheating problems caused by excessive energy consumption, avoid hardware failures caused by excessive power consumption, improve the reliability and stability of the long-term operation of the system, and can dynamically adjust the parameters of the computer board to quickly respond to load changes while ensuring performance requirements, and is applicable to large-scale data centers or high-performance computing environments.

[0057] (2) By removing the obviously redundant features, re-evaluating and selecting the most important features, and finally further refining and optimizing the feature set, redundant data is gradually removed, and a data set with high prediction ability is dynamically selected from a large amount of operating state information as the target state information, reducing overfitting caused by the introduction of redundant features or increasing the complexity of the model, effectively improving the prediction accuracy and calculation efficiency of the energy consumption control model, and enabling more optimized energy efficiency control in different operating environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, where:

[0059] Figure 1 is a schematic flowchart of a control method for a computer board card provided by an embodiment of the present application;

[0060] Figure 2 is a schematic structural diagram of a control device for a computer board card provided by an embodiment of the present application;

[0061] Figure 3 is a schematic structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] The technical solutions in the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application belong to the scope of protection of the present application.

[0063] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such data may be interchanged under appropriate circumstances so that the embodiments of the present application can be implemented in an order different from those illustrated or described herein, and the objects distinguished by "first", "second", etc. generally belong to the same category, and the number of objects is not limited. For example, the first object may be one or multiple. In addition, "and / or" in the specification and claims means at least one of the connected objects, and the character " / " generally indicates an "or" relationship between the associated objects before and after.

[0064] The control method for a computer board card, the control device for a computer board card, the electronic device, and the readable storage medium provided by the embodiments of the present application will be described in detail below with reference to the drawings through specific embodiments and their application scenarios.

[0065] Among them, the control method of the computer board can be applied to a terminal, and can be specifically executed by the hardware or software in the terminal.

[0066] The terminal includes, but is not limited to, portable communication devices such as mobile phones or tablet computers having a touch-sensitive surface (e.g., a touch screen display and / or a touchpad). It should also be understood that in some embodiments, the terminal may not be a portable communication device, but a desktop computer having a touch-sensitive surface (e.g., a touch screen display and / or a touchpad).

[0067] In the following various embodiments, terminals including a display and a touch-sensitive surface are described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, a mouse, and a joystick.

[0068] The control method of the computer board provided in the embodiments of the present application may be executed by an electronic device or a functional module or functional entity in the electronic device that can implement the control method of the computer board. The electronic devices mentioned in the embodiments of the present application include, but are not limited to, mobile phones, tablet computers, computers, cameras, and wearable devices, etc. Hereinafter, the control method of the computer board provided in the embodiments of the present application will be described by taking an electronic device as the execution subject.

[0069] As Figure 1 shown, the control method of the computer board includes:

[0070] In step 110, obtain the running state information of the computer board.

[0071] It can be understood that BMC has a powerful hardware monitoring function, including real-time monitoring of key parameters such as temperature, voltage, fan speed, power consumption, memory usage, and CPU load. BMC can collect this running information through protocols such as IPMI (Intelligent Platform Management Interface) and use it for further analysis and processing.

[0072] In this step, configure sensors in the BMC controller or obtain the real-time running state information of the computer board through the system API interface. The running state information includes, but is not limited to, CPU load, memory usage, network traffic, hard disk status, temperature, power consumption, voltage, etc.

[0073] Multiple hardware components are provided on the computer board. For example, these hardware components include, but are not limited to, CPU, GPU, memory, hard disk, fan, and network card.

[0074] In step 120, according to the dependencies between the data in the operation status information, the target status information is extracted from the operation status information.

[0075] In this step, the obtained operation status information is subjected to data cleaning, standardization, and normalization processing. And to ensure the quality of the data, noise and outliers are removed, and it is converted into a format suitable for model training.

[0076] By calculating the correlations between the various data in the operation status information, redundant information and information that makes a greater contribution to prediction are judged to dynamically select the best feature subset as the target status information, avoiding the impact of redundant features on the model performance, and at the same time improving the prediction accuracy of the model.

[0077] The target status information is data related to the energy consumption control of the computer board, including the load level of the computer board (such as CPU load, memory occupancy, etc.), the temperature of the device, the power consumption level of the device, and the task scheduling status (for example, whether there are high-priority tasks that need to be executed immediately).

[0078] In step 130, the target status information is input into the energy consumption control model to obtain the predicted control parameters output by the energy consumption control model. The energy consumption control model is obtained based on multiple training samples, and the training samples include sample status information and the control parameter labels corresponding to the sample status information.

[0079] In this step, the target status information is used as input data and passed to the shared layer of the energy consumption control model. The shared layer extracts high-level global feature representations from the input data through, such as convolutional layers, fully connected layers, etc., to provide general features for the subsequent specific task layers. The first specific task layer is used to process the features of specific tasks and output predicted control parameters (such as control values for power consumption, CPU frequency, dynamic voltage, memory energy saving, etc.). The second specific task layer is used to further adjust and optimize the energy consumption control based on the first specific task layer and output the final control parameter labels.

[0080] The energy consumption control model is a deep learning model constructed based on multi-task learning (MTL), including a shared layer and two specific task layers.

[0081] The shared layer is used to extract features, and the specific task layers are used to optimize for specific tasks.

[0082] In step 140, based on the predicted control parameters, the computer board is configured.

[0083] In this step, the BMC adjusts the overall hardware state of the computer board according to the predictive control parameters output by the energy consumption control model, realizes the fine configuration of the computer board, and dynamically adjusts the hardware configurations such as power management, temperature control, and frequency regulation of the computer board, so as to reduce energy consumption and improve performance, and ensure the optimal energy efficiency of the computer board during operation.

[0084] For example, by adjusting the power management system, fan control, and Dynamic Voltage and Frequency Scaling (DVFS), etc., the frequency and voltage of the CPU are dynamically adjusted, and the working modes of the memory and storage are controlled, etc., so as to reduce power consumption and ensure performance requirements.

[0085] In addition, the BMC controller can also dynamically adjust the control parameters according to the real-time feedback to ensure that the board card is continuously in the optimal energy efficiency state.

[0086] According to the control method of the computer board provided by this application, by using the energy consumption control model, predicting the optimal control parameters according to the operating state of the computer board, and adjusting the control parameters in real time, it can automatically adapt to different workloads and environmental conditions, avoid excessive energy consumption and inefficient operation in traditional hardware configurations, reduce power consumption without affecting system performance, reduce energy waste, effectively optimize the energy efficiency of the computer board, effectively avoid overheating problems caused by excessive energy consumption, avoid hardware failures caused by excessive power consumption, improve the reliability and stability of the long-term operation of the system, and can dynamically adjust the parameters of the computer board to quickly respond to load changes while ensuring performance requirements, and is applicable to large-scale data centers or high-performance computing environments.

[0087] In some embodiments, according to the dependencies between the data in the operating state information, extracting target state information from the operating state information includes:

[0088] Selecting the operating state information according to the correlation between the data or data combinations in the operating state information to determine the first state information;

[0089] Filtering out the second state information according to the importance of each data or data combination in the first state information to the energy consumption control of the computer board;

[0090] Performing principal component analysis on the data in the second state information, and performing iterative traversal screening of recursive feature elimination. According to the performance impact of the data or data combinations obtained in each round of screening on the energy consumption control model, adjusting the screening criteria until the adjustment degree of the screening criteria is less than the standard threshold, and taking the obtained data and data combinations as the target state information.

[0091] Among them, the correlation between data can be characterized as mutual information, which can capture the complex associations between data and is used to measure the non-linear relationship between two data items.

[0092] In actual execution, first, calculate the mutual information between all data in the operating status information through a regression task, set a screening threshold for preliminary screening. If the mutual information between two data items is greater than this threshold, then these two data items provide redundant information, and one of them can be deleted, retaining the other more important data item. Moreover, according to the correlation between data combinations, identify data groups with high correlation through clustering and screening, and use the screened data and data combinations as the first state information.

[0093] Among them, data combinations can be determined through feature crossing or decision trees, etc.

[0094] For example, combine the voltage and temperature in the operating status information as a data combination, or generate the Cartesian product of the board type and hardware version number in the operating status information as a new data combination.

[0095] Decision trees can capture the interaction between data. For example, the voltage > 12V and temperature > 50°C in the decision tree path can be regarded as a data combination.

[0096] The Wrapper Method takes the performance of the energy consumption control model as the evaluation criterion for data subsets, and selects the optimal data combination based on the performance of the energy consumption control model:

[0097] It can be understood that if the mutual information between two data items is large, it means that they provide similar information, and redundant data can be deleted.

[0098] After preliminary screening, use a random forest classifier to further evaluate the contribution of each data item or data combination in the first state information to the prediction ability of the energy consumption control model, measure the importance of the data, and thus further screen out the top N data items with the most predictive ability as the second state information.

[0099] The random forest classifier is trained based on sample data with importance labels. During the training process of the random forest classifier, multiple decision trees are generated, and each tree is split based on different data. The tree model will automatically calculate the information gain of each data item at the split node, and according to the importance score of the data, select data with higher importance and remove data with lower importance.

[0100] The second state information is further optimized through Principal Component Analysis (PCA) to remove some redundant data with low variance, retain the main components, reduce the data dimension. After PCA optimization, the efficiency of Recursive Feature Elimination (RFE) is improved.

[0101] Based on the data after PCA dimensionality reduction, RFE is performed. By analyzing the performance of the energy consumption control model after each round of recursive elimination, the threshold corresponding to the screening criterion is adjusted. When the performance of the energy consumption control model improves, the selection criterion for features is increased, and when the performance stops improving, the criterion is decreased to allow more features to be selected for optimization until the adjustment degree of the screening criterion is less than the standard threshold. The obtained data and data combinations are used as the target state information.

[0102] In this embodiment, by removing significantly redundant features, then evaluating and selecting the most important features, and finally further refining and optimizing the feature set, redundant data is gradually removed, and a data set with high predictive ability is dynamically selected from a large amount of operating state information as the target state information, reducing overfitting caused by the introduction of redundant features or increasing the complexity of the model, effectively improving the prediction accuracy and calculation efficiency of the energy consumption control model, and enabling more optimized energy efficiency control in different operating environments.

[0103] In some embodiments, the energy consumption control model includes a shared layer, a first specific task layer, and a second specific task layer connected in sequence;

[0104] Before inputting the target state information into the energy consumption control model, the method further includes:

[0105] Obtain a plurality of training samples, where the training samples include sample state information and corresponding control parameter labels;

[0106] Input the plurality of training samples into the energy consumption control model in sequence to pre-train the shared layer and the specific task layer;

[0107] Perform multi-task training on the energy consumption control model through a plurality of training samples to obtain the trained energy consumption control model.

[0108] It can be understood that when the energy consumption control model predicts the control parameters of the computer board, in the presence of multiple related tasks, through the combined training of pre-training and multi-task training, overfitting of the model can be reduced and the overall performance of the model can be improved.

[0109] The shared layer is used to learn the part of the feature representation shared by all tasks and can extract universal features that are useful for different tasks from the target state information of the computer board.

[0110] For example, the general features are local features or long-term dependencies in time series data. For the energy consumption control model, the shared layer may learn global features that affect the energy consumption of computer boards. For example, the state laws of computer boards under operating states such as temperature and workload.

[0111] The control parameter labels include energy consumption parameter labels and heat dissipation parameter labels.

[0112] In actual execution, multiple training samples are collected, including the state information of each sample and the corresponding control parameter labels.

[0113] Pre-train the shared layer and the first specific task layer through the sample state information in the training samples, enabling the energy consumption control model to first learn basic features and specific patterns of the prediction task. The shared layer learns general features, and the first specific task layer learns the mapping relationship of energy consumption prediction, providing a reasonable parameter initialization starting point for the energy consumption control model and avoiding training instability caused by random initialization.

[0114] Perform multi-task training on the energy consumption control model through the sample state information, energy consumption parameter labels, and heat dissipation parameter labels in the training samples. Jointly train the shared layer, the first specific task layer, and the second specific task layer to improve the generalization ability of the model through the correlation between tasks.

[0115] The shared layer learns the common features of multiple tasks, such as the correlation between energy consumption control and heat dissipation control.

[0116] The first specific task layer and the second specific task layer focus on their respective tasks, such as the energy consumption prediction task and the heat dissipation optimization task, etc. The tasks enhance each other's feature expressions through the shared layer.

[0117] The first specific task layer is the main task layer, which is used to perform preliminary processing on general features, perform in-depth combination and transformation on data according to task requirements, and generate preliminary control parameters related to energy consumption control.

[0118] The second specific task layer is the auxiliary task layer, which is used to introduce more complex logics or rules on the basis of the first specific task layer, share underlying features that are related to both board card state and energy consumption and heat dissipation, and further optimize and refine control parameters to adapt to different operating scenarios and task requirements. For example, the second specific task layer combines a multi-agent reinforcement learning system to dynamically adjust control parameters according to real-time data.

[0119] In this embodiment, by pre-training and multi-task training the energy consumption control model, the training efficiency and generalization ability of the energy consumption control model are improved, the configuration of the computer board is automatically optimized to achieve dynamic adjustment of the energy efficiency of the board, reduce the consumption of computing resources, and at the same time improve the control accuracy. Through multi-task learning, the model can handle complex and dynamic application scenarios, thereby optimizing energy consumption control.

[0120] In some embodiments, the shared layer is used to extract general features from the sample state information in the training samples;

[0121] The first specific task layer includes a plurality of component modules, which correspond one-to-one to the hardware components of the computer board, and the component modules are used to generate hardware control parameters corresponding to the hardware components according to the general features;

[0122] The second specific task layer includes a plurality of task modules, which correspond one-to-one to the control tasks of the computer board, and the task modules are used to optimize and adjust the hardware control parameters according to the control tasks to generate the predicted control parameters.

[0123] For example, the component modules include two or more of a CPU control module, a GPU control module, a memory control module, a storage device control module, a network control module, and a power management module, and the task modules include two or more of a temperature management module, a hardware health module, a system overload protection module, an energy efficiency recovery module, and a fault recovery and backup module.

[0124] The first specific task layer is used for the overall energy efficiency optimization of the computer board. The hardware control parameters are used to adjust the energy consumption and optimize the performance of each hardware component, relying on the general features extracted by the shared layer and converting these features into high-level features related to energy consumption prediction and optimization tasks.

[0125] Among them, the CPU control module is used to dynamically adjust the working frequency and the number of cores of the CPU to adapt to the current computing load, reduce unnecessary energy consumption, increase the frequency under high load, and reduce the frequency or enable the low-power state under low load; the GPU control module is used to manage the power consumption of the graphics processing unit, and adjust the GPU frequency and voltage according to the requirements of graphics-intensive tasks and light tasks to optimize power consumption; the memory control module is used to adjust the working frequency of the memory or enable the low-power mode to reduce the energy consumption during standby, ensuring that the memory module operates efficiently without wasting power; the storage device control module is used to control the power consumption of the hard disk or solid-state drive. By managing the read / write state of the hard disk or placing the hard disk in the low-power mode when inactive, it optimizes energy consumption; the network module control is used to manage the power consumption of the network interface card (NIC), adjust the data transfer rate, network interface enable state, etc., to ensure the energy efficiency of the network module; the power management module is responsible for the overall power management and scheduling, adjusts the efficiency and stability of the power supply to ensure optimal energy efficiency, and manages the power consumption requirements of different hardware.

[0126] The second specific task layer is used to execute more refined control tasks. For example, the stability, security of computer boards, and energy efficiency adjustment under specific circumstances, achieving the purposes of hardware protection, system stability, fault prevention and optimization, ensuring that the hardware can operate while ensuring safety, and saving energy to the greatest extent.

[0127] Among them, the temperature management module is used to manage the temperatures of hardware components such as the CPU, GPU, memory, and storage device, and start a cooling strategy or adjust the hardware working state (such as reducing the frequency or enabling the energy-saving mode) when the temperature is too high to prevent the hardware from overheating; the hardware health module is used to manage the health status of each hardware component, such as voltage fluctuations, temperature changes, workload, etc., detect abnormalities and adjust control parameters to ensure that the hardware is not overloaded or damaged.

[0128] When the system overload protection module detects that the computer board is in an overloaded state, such as the CPU or GPU having too high a load or exceeding the power consumption standard, it adjusts control parameters, such as adjusting the hardware power, or temporarily turning off some devices to avoid system crashes and hardware damage.

[0129] The energy efficiency recovery module is used to recover and manage the waste heat or other energy in the device, convert it into available energy for other components to use, and maximize the energy efficiency of the system.

[0130] The fault recovery and backup module is responsible for starting backup hardware or adopting a redundancy mechanism when a hardware fault or abnormality occurs, to reduce system downtime and protect the continuous operation of the hardware, and avoid damage caused by problems such as overload.

[0131] In this embodiment, through the collaborative work of the first specific task layer and the second specific task layer, it is possible to ensure that the computer board can achieve the best energy efficiency level under different tasks and conditions.

[0132] In some embodiments, the computer board includes a main board and at least one external board extended from the main board. Before obtaining the operating state information of the computer board, the method further includes:

[0133] Obtain the first state information of the main board and the external board;

[0134] Based on the first state information, determine the basic priority of each task of the main board and the external board, as well as the board heat dissipation capacity and board heat dissipation cycle of the main board and the external board;

[0135] Based on the basic priority of the tasks, the board heat dissipation capacity and the board heat dissipation cycle, allocate tasks to the main board and the external board.

[0136] The first state information includes the environmental state, the heat dissipation state, and the load state. The environmental state includes the temperature, humidity, air flow condition, etc. of the environment where the main board and the external board are located.

[0137] The heat dissipation state includes the heat dissipation situation of the main board and the external board, such as the temperature of each component (CPU, GPU, memory, fan, etc.), the fan speed, the efficiency of the heat sink, and the temperature control situation of the overall system.

[0138] The load state includes the current computing load state of the main board, including CPU and GPU usage rates, memory usage, network bandwidth occupancy, hard disk read and write conditions, etc., as well as the tasks being executed.

[0139] The load can be migrated through dynamic load balancing algorithms (such as weighted round-robin, least load first, temperature-based load migration, etc.). For example, on a computer board with a higher temperature, reduce its compute-intensive tasks and allocate them to the external board or other boards.

[0140] For high-load tasks, load adjustment should not only consider the instantaneous load, but also combine the board heat dissipation capacity and heat dissipation cycle. For example, when the heat dissipation system of a board has a response delay, the load should not increase too much in a short period of time, otherwise the heat dissipation capacity will not be able to keep up in time, resulting in a sharp rise in temperature. Load adjustment needs to predict temperature changes in advance and adjust the load in a timely manner according to the heat dissipation speed to avoid excessive instantaneous load.

[0141] In some embodiments, allocating tasks to the main board and the external board based on the basic priority of the tasks, the board heat dissipation capacity, and the board heat dissipation cycle includes:

[0142] Determine the allocation priority of each task on the main board card and the external board card based on the basic priority of the task and the heat dissipation capacity of the board card;

[0143] Determine the actual load of the task on the main board card and the external board card based on the workload of each task and the heat dissipation cycle;

[0144] Determine the allocation of the task based on the allocation priority and the actual load;

[0145] In some embodiments, the allocation priority is:

[0146]

[0147] where F ij is the allocation priority of task i on board card j; A i is the adjusted priority of task i, determined based on the basic priority; C j (t) is the heat dissipation capacity of board card j at time t; Q j is the resource utilization rate of board card j;

[0148] The actual load is:

[0149]

[0150] where L ij is the actual load of task i on board card j; W i is the amount of resources occupied by task i; H j is the heat dissipation cycle of board card j;

[0151] The determining the allocation of the task based on the allocation priority and the actual load includes:

[0152] Construct a task scheduling model:

[0153]

[0154] The objective function O of the task scheduling model is:

[0155]

[0156] where P i is the basic priority of task i; C j (t) is the heat dissipation capacity of board card j at time t; L ij is the amount of resources occupied (load) by task i on board card j; ∈ is a constant used to prevent division by zero errors; τ is the heat dissipation resource adjustment coefficient; R ij is the heat dissipation requirement of task i on board card j; is the occupancy (load) of resources and the temperature balance adjustment coefficient; T j is the current temperature of board j; δ j is the role adjustment factor of board j;

[0157] Based on the objective function, solve the task scheduling model to obtain the optimal allocation plan for the task and execute it.

[0158] Among them, by maximizing the difference between the priority of task allocation and the matching of the heat dissipation system, it is ensured that each task can be executed efficiently and the board is prevented from being overloaded.

[0159] The heat dissipation resources can also be allocated according to the heat dissipation requirements corresponding to the optimal allocation plan.

[0160] The basic priority P of task i i is:

[0161]

[0162] where P i is the basic priority of task i; is the load power of task i; is the current temperature of the board for executing task i; T target is the target temperature; is the type weight of task i; w1, w2, and w3 are all weight coefficients.

[0163] The adjusted priority is:

[0164]

[0165] where A i is the adjusted priority of task i; P i is the basic priority of task i; L ij is the occupancy of resources of task i on board j; L max,j is the maximum load value of board j; C j is the current board heat dissipation capacity of board j; C max,j is the maximum heat dissipation capacity of board j.; T j is the current temperature of board j; T max,j is the maximum temperature that board j can withstand; β j is the priority adjustment factor of board j;

[0166] The main board and the expansion board are set to different values. For example, the main board is set to a higher priority adjustment factor, and the expansion board is set to a lower priority adjustment factor.

[0167] The heat dissipation requirement of the task is:

[0168]

[0169] Among them, R ij is the heat dissipation requirement of task i on board j; L ij is the resource occupancy (load) of task i on board j; P i is the basic priority of task i; L max,j is the maximum load value of board j; ξ j is the heat dissipation requirement adjustment factor of board j;

[0170] The dynamic impact of the heat dissipation capacity of the board is as follows:

[0171]

[0172] Among them, C j (t) is the board heat dissipation capacity of board j at time t; T max,j is the maximum temperature that board j can withstand; T j (t) is the temperature of board j at time t; T0(t) is the ambient temperature at time t; S 1,j (t) is the fan efficiency of board j at time t; S 2,j (t) is the radiator efficiency of board j at time t; θ j is the dynamic heat dissipation adjustment factor of board j;

[0173]

[0174] Among them, ΔT j is the total temperature burden of board j; L ij is the resource occupancy (load) of task i on board j; T j is the current temperature of board j; ρ j is the load balancing adjustment factor;

[0175] The objective function is used to optimize task allocation and heat dissipation resource allocation to ensure that the system achieves a balance between temperature and load.

[0176] The priority adjustment factor β j , heat dissipation requirement adjustment factor ξ j , dynamic heat dissipation adjustment factor θ j , load balancing adjustment factor ρ j , and role adjustment factor δ j of the main board and the expansion board are all set to different values. For example, the main board is set to a higher adjustment factor, while the expansion board is set to a lower adjustment factor.

[0177] In this embodiment, the main board card has a higher priority, stronger heat dissipation ability, and higher load processing ability, while the expansion board card plays an auxiliary role. Through such adjustment, the system can allocate resources more reasonably and maintain efficient and stable operation.

[0178] It can be understood that different types of tasks are configured with different weights. For example, a high-priority task type can have a higher weight value.

[0179] In the embodiment of the present application, the execution subject of the control method of the computer board card can be the control device of the computer board card. In the embodiment of the present application, taking the control device of the computer board card executing the control method of the computer board card as an example, the control device of the computer board card provided by the embodiment of the present application is described.

[0180] The embodiment of the present application further provides a control device for a computer board card, which is applied to a BMC controller, and the BMC controller controls the computer board card.

[0181] As Figure 2 shown, the control device of the computer board card includes:

[0182] An acquisition module 210, configured to acquire the operation status information of the computer board card;

[0183] A first processing module 220, configured to extract target status information from the operation status information according to the dependence between data in the operation status information;

[0184] A second processing module 230, configured to input the target status information into an energy consumption control model to obtain predicted control parameters output by the energy consumption control model. The energy consumption control model is based on multiple training samples, and the training samples include sample status information and control parameter labels corresponding to the sample status information;

[0185] A third processing module 240, configured to configure the computer board card based on the predicted control parameters.

[0186] According to the control device of the computer board card provided by the embodiment of the present application, by using the energy consumption control model, predicting the optimal control parameters according to the operation status of the computer board card, and adjusting the control parameters in real time, it can automatically adapt to different workloads and environmental conditions, avoid excessive energy consumption and inefficient operation in traditional hardware configurations, reduce power consumption without affecting system performance, reduce energy waste, effectively optimize the energy efficiency of the computer board card, effectively avoid overheating problems caused by excessive energy consumption, avoid hardware failures caused by excessive power consumption, improve the reliability and stability of the long-term operation of the system, be able to dynamically adjust the parameters of the computer board card to quickly respond to load changes, and at the same time ensure performance requirements, and is applicable to large-scale data centers or high-performance computing environments.

[0187] The control device of the computer board in the embodiments of the present application can be an electronic device or a component in an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices other than terminals. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a handheld computer, a Mobile Internet Device (MID), a robot, a wearable device, an Ultra-Mobile Personal Computer (UMPC), a server, etc., and the embodiments of the present application do not make specific limitations.

[0188] The control device of the computer board in the embodiments of the present application can be a device with an operating system. The operating system can be the Android operating system, the iOS operating system, or other possible operating systems, and the embodiments of the present application do not make specific limitations.

[0189] The control device of the computer board provided in the embodiments of the present application can implement each process implemented by the control method embodiment of the computer board in the above embodiments. To avoid repetition, it will not be elaborated here.

[0190] In some embodiments, as Figure 3 shown, the embodiments of the present application also provide an electronic device 300, including a processor 301, a memory 302, and a computer program stored on the memory 302 and executable on the processor 301. When the program is executed by the processor 301, it implements each process of the control method embodiment of the above computer board and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0191] It should be noted that the electronic devices in the embodiments of the present application include the above-mentioned mobile electronic devices and non-mobile electronic devices.

[0192] The embodiments of the present application also provide a non-transitory computer-readable storage medium. A computer program is stored on the non-transitory computer-readable storage medium. When the computer program is executed by a processor, it implements each process of the control method embodiment of the above computer board and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0193] Among them, the processor is the processor in the electronic device in the above embodiments. The readable storage medium includes a computer-readable storage medium, such as a computer Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disc, etc.

[0194] The embodiment of the present application also provides a computer program product, including a computer program, which when executed by a processor implements the control method of the above computer board.

[0195] Wherein, the processor is the processor in the electronic device in the above embodiment. The readable storage medium includes computer-readable storage media such as computer read-only memory ROM, random access memory RAM, magnetic disk or optical disc, etc.

[0196] The embodiment of the present application further provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement each process of the control method embodiment of the above computer board, and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0197] It should be understood that the chip mentioned in the embodiment of the present application can also be referred to as a system-on-chip, system chip, chip system or system-on-chip, etc.

[0198] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in a reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0199] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a computer software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the control method of the computer board in each embodiment of the present application.

[0200] In the description of the present application, "the first feature" and "the second feature" may include one or more of such features.

[0201] In the description of the present application, the meaning of "a plurality of" is two or more.

[0202] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative rather than restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them fall within the protection scope of the present application.

[0203] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0204] Although the embodiments of the present application have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and purposes of the present application. The scope of the present application is defined by the claims and their equivalents.

Claims

1. A control method for a computer board, characterized in that Applied to a BMC controller for controlling a computer board of the BMC controller, the method includes: Obtaining the operating status information of the computer board; Extracting target status information from the operating status information according to the dependencies between the data in the operating status information; Inputting the target status information into an energy consumption control model to obtain predicted control parameters output by the energy consumption control model. The energy consumption control model is obtained based on a plurality of training samples, and the training samples include sample status information and control parameter labels corresponding to the sample status information; Configuring the computer board based on the predicted control parameters.

2. The control method of the computer board according to claim 1, wherein Extracting target status information from the operating status information according to the dependencies between the data in the operating status information includes: Selecting the operating status information according to the correlation between the data or data combinations in the operating status information to determine the first status information; Filtering out the second status information according to the importance of each data or data combination in the first status information for the energy consumption control of the computer board; Performing principal component analysis on the data in the second status information and performing iterative traversal screening of recursive feature elimination. According to the influence of the data or data combinations obtained in each round of screening on the performance of the energy consumption control model, adjusting the screening criteria until the adjustment degree of the screening criteria is less than the standard threshold, and taking the obtained data and data combinations as the target status information.

3. The control method of the computer board according to claim 1, wherein The energy consumption control model includes a shared layer, a first specific task layer, and a second specific task layer connected in sequence; Before inputting the target status information into the energy consumption control model, the method further includes: Obtaining a plurality of training samples, where the training samples include sample status information and corresponding control parameter labels; Sequentially inputting the plurality of training samples into the energy consumption control model to pre-train the shared layer and the specific task layer; Performing multi-task training on the energy consumption control model through a plurality of training samples to obtain the trained energy consumption control model.

4. The control method of the computer board according to claim 3, characterized in that The shared layer is used to extract general features from the sample status information in the training samples; The first specific task layer includes a plurality of component modules, and the component modules correspond one-to-one with the hardware components of the computer board. The component modules are used to generate hardware control parameters corresponding to the hardware components according to the general features; The second specific task layer includes a plurality of task modules, and the task modules correspond one-to-one with the control tasks of the computer board. The task modules are used to optimize and adjust the hardware control parameters according to the control tasks to generate the predicted control parameters.

5. The control method of the computer board according to claim 3, characterized in that, The computer board includes a main board and at least one external board extended by the main board. Before obtaining the operating status information of the computer board, the method further includes: Obtaining the first status information of the main board and the external board; Based on the first status information, determining the basic priority of each task of the main board and the external board, and the board heat dissipation capacity and board heat dissipation cycle of the main board and the external board; Based on the basic priority of the task, the heat dissipation capacity of the board card, and the heat dissipation cycle of the board card, allocate the tasks and the heat dissipation system for the main board card and the external board card.

6. The control method of the computer board according to claim 5, characterized in that, Based on the basic priority of the task, the heat dissipation capacity of the board card, and the heat dissipation cycle of the board card, allocate the tasks for the main board card and the external board card, including: Based on the basic priority of the task and the heat dissipation capacity of the board card, determine the allocation priority of each task on the main board card and the external board card; Based on the workload of each task and the heat dissipation cycle, determine the actual load of the task on the main board card and the external board card; Based on the allocation priority and the actual load, determine the allocation of the task.

7. The control method of the computer board according to claim 6, characterized in that, The allocation priority is: Among them, F ij is the allocation priority of task i on board j; A i is the adjusted priority of task i, determined based on the basic priority; C j (t) is the heat dissipation capacity of board j at time t; Q j is the resource utilization rate of board j; The actual load is: Among them, L ij is the actual load of task i on board j; W i is the resource occupancy of task i; H j is the heat dissipation cycle of board j; Based on the allocation priority and the actual load, determine the allocation of the task, including: Construct a task scheduling model: The objective function O of the task scheduling model is: Among them, P i is the basic priority of task i; C j (t) is the heat dissipation capacity of board j at time t; L ij is the occupancy of the resources of task i on board j; ∈ is a constant used to prevent division-by-zero errors; τ is the heat dissipation resource adjustment coefficient; R ij is the heat dissipation requirement of task i on board j; is the occupancy of resources and the temperature balance adjustment coefficient; T j is the current temperature of board j; δ j is the role adjustment factor of board j; Based on the objective function, solve the task scheduling model to obtain the optimal allocation scheme of the task and execute it.

8. A control device for a computer board, characterized in that, Applied to the BMC controller, the BMC controller controls the computer board card, and the device includes: An acquisition module, configured to acquire the operation status information of the computer board card; A first processing module, configured to extract target status information from the operation status information according to the dependency between data in the operation status information; A second processing module, configured to input the target status information into an energy consumption control model to obtain predicted control parameters output by the energy consumption control model. The energy consumption control model is obtained based on a plurality of training samples, and the training samples include sample status information and control parameter labels corresponding to the sample status information; A third processing module, configured to configure the computer board card based on the predicted control parameters.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the control method of the computer board card according to any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the control method of the computer board card according to any one of claims 1-7.

Images