Equipment processing method and device, computer program product and electronic equipment

Abstract

The embodiment of the invention provides an equipment processing method and device, a computer program product and electronic equipment, and relates to the field of computers.The method comprises the steps that performance data of a graphic processing unit on a hardware management platform within a preset time period are obtained, feature extraction processing is conducted on the performance data, and performance features are obtained; the performance features are input into a target prediction model integrated in a hardware management platform, an equipment prediction result is output, the target prediction model is obtained through training of a preset support vector machine model, and the equipment prediction result is used for representing the equipment state of the graphic processing unit; and under the condition that the equipment state of the graphic processing unit is a fault state, determining a fault processing strategy according to the equipment prediction result, and processing the graphic processing unit according to the fault processing strategy. According to the fault processing method and device, the problem of low processing efficiency during fault processing of the equipment in the prior art is solved, and the effect of improving the fault processing efficiency is further achieved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computers, and more specifically, to a device processing method, apparatus, computer program product, and electronic device. Background Art

[0002] With the popularization of informatization in various fields of society, servers, as key devices for data processing and storage, are playing an increasingly important role. In this process, Open BMC (Open Baseboard Management Controller), as an open source management controller based on an operating system, is widely used in the management of modern data centers and servers due to its flexible and customizable features. It provides strong support for the monitoring, management and maintenance of server hardware and systems, and can remotely monitor and manage servers and data center equipment. The controller is equipped with an important computing component, GPU (Graphics Processing Unit), which plays a key role in processing graphics-intensive tasks and parallel computing tasks. However, its potential failure may cause unpredictable degradation of system performance and may even cause system crashes. This situation not only increases the risk of service interruption, but also brings expensive repair and replacement costs. Therefore, it is necessary to perform fault diagnosis and early prediction of the above components.

[0003] However, the fault detection methods of related technologies can often only detect the fault after it has occurred. After detecting a graphics processing unit failure, the system may experience service interruption or performance degradation, and emergency repair or equipment replacement is required. This not only increases maintenance costs but also causes data loss or damage.

[0004] There is a problem of low processing efficiency when troubleshooting equipment in related technologies, and no effective solution has been proposed yet. Summary of the Invention

[0005] The embodiments of the present application provide a device processing method, apparatus, computer program product, and electronic device to at least solve the problem of low processing efficiency when performing fault processing on a device in the related art.

[0006] According to one embodiment of the present application, a device processing method is provided, including: obtaining performance data of a graphics processing unit on a hardware management platform within a preset time period, performing feature extraction processing on the performance data, and obtaining performance characteristics; inputting the performance characteristics into a target prediction model integrated in the hardware management platform, and outputting a device prediction result, wherein the target prediction model is obtained by training a preset support vector machine model, and the device prediction result is used to characterize the device status of the graphics processing unit; in a case where the device prediction result characterizes that the device status of the graphics processing unit is a fault state, determining a fault handling strategy according to the device prediction result, and processing the graphics processing unit according to the fault handling strategy.

[0007] In an exemplary embodiment, the target prediction model is trained in the following manner: obtaining the maintenance log and fault record of each hardware device, filtering out maintenance data from M maintenance logs and M fault records according to each set of performance data, and obtaining M sets of maintenance data, where M is a positive integer; performing feature extraction processing on each set of maintenance data to obtain M sets of fault features, obtaining the performance fault states associated with each set of fault features, and obtaining M performance fault states; forming a model training set consisting of the M sets of fault features and the performance fault states associated with each set of fault features; using the model training set to train a preset support vector machine model to obtain a trained support vector machine model, and determining the trained support vector machine model as the target prediction model.

[0008] In an exemplary embodiment, each set of maintenance data is subjected to feature extraction processing to obtain M sets of fault features, including: for each set of maintenance data, each maintenance data is calculated to obtain N maintenance indicator data, where N is a positive integer; a performance threshold list is obtained, the N maintenance indicator data are screened using the performance threshold list to obtain Y maintenance indicator data, and a set of fault features is determined based on the Y maintenance indicator data, where Y is a positive integer.

[0009] In an exemplary embodiment, a preset support vector machine model is trained using a model training set to obtain a trained support vector machine model, including: performing data conversion processing on each group of fault features to obtain M groups of fault indicator data, wherein the data conversion processing method includes a logarithmic transformation method and an exponential transformation method; identifying the correlation state of the M groups of fault indicator data, and determining an initial kernel function according to the correlation state, wherein the initial kernel function refers to a kernel function initialized with function parameters, and the function type of the kernel function is one of the following: a linear kernel function, a polynomial kernel function, and a radial basis kernel function; obtaining an initial regularization coefficient, and configuring the preset support vector machine model according to the initial regularization coefficient and the initial kernel function to obtain an initial support vector machine model; training the initial support vector machine model using the model training set to obtain a first support vector machine model; determining a radial basis kernel function according to the first support vector machine model Lagrange multipliers, calculating the decision boundary of the Lagrange multipliers, updating the Lagrange multipliers according to the decision boundary of the Lagrange multipliers to obtain updated Lagrange multipliers; judging whether the updated Lagrange multipliers meet the convergence conditions, and when the updated Lagrange multipliers meet the convergence conditions, calculating the first support vector according to the updated Lagrange multipliers, updating the first support vector machine model according to the first support vector, and determining the updated support vector machine model as the trained support vector machine model; when the updated Lagrange multipliers do not meet the convergence conditions, performing quadratic programming processing on the updated Lagrange multipliers to obtain processed Lagrange multipliers, until the processed Lagrange multipliers meet the convergence conditions, and calculating the second support vector according to the processed Lagrange multipliers; updating the updated support vector machine model according to the second support vector to obtain the trained support vector machine model.

[0010] In an exemplary embodiment, identifying the association status of M groups of fault indicator data, and determining the initial kernel function based on the association status includes: when the association status of the M groups of fault indicator data is a linear state, determining the function type of the initial kernel function as the function type of the linear kernel function; when the association status of the M groups of fault indicator data is a curve arrangement state, determining the function type of the initial kernel function as the function type of the polynomial kernel function; when the association status of the M groups of fault indicator data is a nonlinear state, determining the function type of the initial kernel function as the function type of the radial basis kernel function.

[0011] In an exemplary embodiment, determining a fault handling strategy based on a device prediction result includes: extracting fault information from the device prediction result to obtain a predicted fault level and a fault type; obtaining a fault strategy list, and filtering out a fault handling strategy from the fault strategy list based on the predicted fault level and the fault type.

[0012] In an exemplary embodiment, the fault handling strategy includes a performance optimization strategy and an equipment replacement strategy. Processing the graphics processing unit according to the fault handling strategy includes: when the fault type associated with the equipment prediction result is a performance fault, issuing a first alarm message, and performing performance optimization processing on the graphics processing unit through the performance optimization strategy; when the fault type associated with the equipment prediction result is an equipment fault, issuing a second alarm message, and performing equipment replacement processing on the graphics processing unit through the equipment replacement strategy.

[0013] According to another embodiment of the present application, a processing device for a device is provided, including: an acquisition unit, used to acquire performance data of a graphics processing unit on a hardware management platform within a preset time period, perform feature extraction processing on the performance data, and obtain performance characteristics; an input unit, used to input the performance characteristics into a target prediction model integrated in the hardware management platform, and output a device prediction result, wherein the target prediction model is obtained by training a preset support vector machine model, and the device prediction result is used to characterize the device status of the graphics processing unit; a determination unit, used to determine a fault handling strategy according to the device prediction result when the device prediction result characterizes that the device status of the graphics processing unit is a fault state, and process the graphics processing unit according to the fault handling strategy.

[0014] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored. The computer program is configured to execute the steps of any one of the above method embodiments when run.

[0015] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any one of the above method embodiments.

[0016] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps of any one of the above method embodiments are performed.

[0017] Through this application, by obtaining the performance characteristics of the graphics processing unit on the hardware management platform within a preset time period, inputting the performance characteristics into the target prediction model integrated in the hardware management platform, outputting the device prediction results, and using the device prediction results to achieve optimization and adjustment of the graphics processing unit, it is possible to solve the problem of low processing efficiency when performing fault processing on the device in the related technology, thereby achieving the effect of improving the fault processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1This is a hardware structure block diagram of a mobile terminal for a processing method of a device according to an embodiment of the present application;

[0019] Figure 2 is a flowchart of a processing method of a device according to an embodiment of the present application;

[0020] Figure 3 is a schematic diagram of a processing method of an optional device according to an embodiment of the present application;

[0021] Figure 4 is a schematic diagram of a method for collecting performance data according to an embodiment of the present application;

[0022] Figure 5 is a schematic diagram of a processing device of an apparatus according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0024] It should be noted that the terms "first", "second", etc. in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0025] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal of a processing method of a device according to an embodiment of the present application. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown) a processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices) and a memory 104 for storing data, wherein the above mobile terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0026] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the processing method of the device in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implementing the above-mentioned method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories may be connected to the mobile terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0027] The transmission device 106 is used to receive or send data via a network. A specific example of the aforementioned network may include a wireless network provided by the mobile terminal's communications provider. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0028] In this embodiment, a processing method for a device running on the above mobile terminal or network architecture is provided. Figure 2 is a flowchart of a processing method of a device according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0029] Step S201 : acquiring performance data of a graphics processing unit on a hardware management platform within a preset time period, performing feature extraction processing on the performance data, and obtaining performance features.

[0030] Specifically, the hardware management platform can be Open BMC (Open Baseboard Management Controller). As an open source management controller based on an operating system, Open BMC can remotely monitor and manage servers and data center equipment. A graphics processing unit refers to a processor unit GPU (Graphics Processing Unit), which can process images and perform graphics-related calculations, and can also accelerate graphics rendering, scientific computing, and deep learning tasks. In order to detect faults in a graphics processing unit, the performance data of the image processing unit can be obtained in real time, and the performance data can be processed by feature extraction and other processing to obtain performance characteristics, wherein the performance characteristics may include temperature characteristics, voltage characteristics, frequency characteristics, load characteristics, etc., and may also include frequency change rate, power consumption volatility, etc.

[0031] In step S202, the performance characteristics are input into a target prediction model integrated in the hardware management platform, and a device prediction result is output. The target prediction model is trained by a preset support vector machine model, and the device prediction result is used to characterize the device status of the graphics processing unit.

[0032] Specifically, the target prediction model is trained using a support vector machine (SVM) model. The SVM is a supervised learning algorithm for classification and regression problems that finds the optimal decision boundary in feature space to maximize the margin between different classes. After obtaining performance characteristics, these characteristics are integrated into the target prediction model of the open infrastructure management controller. The model then processes the performance characteristics to generate device prediction results that represent the device status of the graphics processing unit (GPU).

[0033] Step S203 : when the device prediction result indicates that the device state of the graphics processing unit is a fault state, a fault handling strategy is determined according to the device prediction result, and the graphics processing unit is handled according to the fault handling strategy.

[0034] Specifically, after the target prediction model outputs the device prediction result, the result can be identified. If the device prediction result indicates that the graphics processing unit is operating normally, the graphics processing unit can continue to be monitored; if the device prediction result indicates that there is a fault in the operation of the graphics processing unit, the corresponding processing strategy can be determined based on the output device prediction result, that is, the fault handling strategy can be determined, and then the above-mentioned graphics processing unit can be optimized and adjusted according to the fault handling strategy to reduce the impact of the fault on the operation of the hardware management platform.

[0035] Through the above steps, by obtaining the performance characteristics of the graphics processing unit on the hardware management platform within a preset time period, the performance characteristics are input into the target prediction model integrated in the hardware management platform, and the device prediction results are output. The device prediction results are used to optimize and adjust the graphics processing unit, which solves the problem of low processing efficiency when performing fault processing on the device in related technologies, thereby achieving the effect of improving fault processing efficiency.

[0036] The execution entity of the above steps may be a server, a terminal, etc., but is not limited thereto.

[0037] Before using the target prediction model to predict the device status, the target prediction model needs to be trained first. Optionally, in the device processing method provided in the embodiment of the present application, the target prediction model is trained in the following manner: obtain the maintenance log and fault record of each hardware device, filter out maintenance data from M maintenance logs and M fault records according to each set of performance data, and obtain M groups of maintenance data, where M is a positive integer; perform feature extraction processing on each set of maintenance data to obtain M groups of fault features, obtain the performance fault state associated with each set of fault features, and obtain M performance fault states; the M groups of fault features and the performance fault state associated with each set of fault features constitute a model training set; use the model training set to train a preset support vector machine model to obtain a trained support vector machine model, and determine the trained support vector machine model as the target prediction model.

[0038] Specifically, when training the model, maintenance logs and fault records of the graphics processing unit (GPU) can be obtained from the hardware management platform where the GPU is installed. The required maintenance data can then be filtered from these logs and fault records. These logs and fault records may include the model of the device being maintained, the length of time the device has been in use, the type of component replaced, the time the fault occurred, and the type of fault. The collected maintenance data is then cleaned to remove invalid or incomplete data. Feature extraction is then performed on the cleaned data to ensure data quality and consistency. For example, time series analysis and statistical analysis can be used to identify key indicators of device performance and subsequently derive fault characteristics.

[0039] The processed fault features can be further combined with the corresponding fault states to form a training set for the model. After selecting the support vector machine as the prediction model, the model is trained based on the above training set. It should be noted that before this, appropriate kernel functions and regularization parameters can be selected to optimize model performance. Finally, the trained model is evaluated using a validation set consisting of fault features and fault states using methods such as cross-validation to avoid overfitting and ensure the generalization ability of the model. During the evaluation, the model parameters can be adjusted according to the evaluation results, and a satisfactory prediction accuracy is achieved after multiple iterative training to obtain the target prediction model. This embodiment constructs an effective prediction model and uses the constructed prediction model to predict the device status of the graphics processing unit, thereby helping maintenance personnel to identify potential problems with hardware equipment in advance, thereby improving the reliability of the equipment and reducing maintenance costs.

[0040] In order to improve the prediction accuracy of the model, optionally, in the processing method of the device provided in the embodiment of the present application, each group of maintenance data is subjected to feature extraction processing to obtain M groups of fault features, including: for each group of maintenance data, each maintenance data is calculated to obtain N maintenance indicator data, where N is a positive integer; obtaining a performance threshold list, using the performance threshold list to filter the N maintenance indicator data to obtain Y maintenance indicator data, and determining a group of fault features based on the Y maintenance indicator data, where Y is a positive integer.

[0041] Specifically, in order to improve the prediction accuracy of the model, the training set used to train the target prediction model can be optimized, that is, the fault characteristics that best meet the fault requirements can be selected. First, the maintenance data can be calculated in combination with the hardware characteristics and prior knowledge of the graphics processing unit to obtain the corresponding maintenance indicator data, where the maintenance data can include the equipment's operating data, sensor readings, performance indicators, and historical maintenance records.

[0042] After obtaining a performance threshold list including indicator thresholds for multiple fault states, the performance threshold list is used to screen the above maintenance indicator data. Through screening, multiple maintenance indicator data that meet the standards can be obtained, and then by identifying fault characteristics, potential equipment failures can be predicted in advance, so that preventive maintenance measures can be taken, reducing unexpected downtime and improving the overall efficiency and reliability of the equipment.

[0043] For example, GPU performance and health monitoring can be based on collected temperature data from hardware devices. This can be achieved by collecting temperature data from the device over a period of time and then quickly understanding the rate of temperature change through slope analysis, or by analyzing the moving average to determine the trend of temperature changes. If the temperature shows a continuous upward trend, this may indicate a problem with the cooling component or that the device is overloaded. Furthermore, the device's status can be determined by determining the correlation between temperature data and other sensor data (such as fan speed and system load). For example, if the fan speed increases but the temperature remains high, this may indicate insufficient cooling efficiency. Correlation analysis can provide a deeper understanding of the interaction between different parameters and optimize system configuration. Device monitoring can also be performed by setting temperature thresholds. When the collected temperature data exceeds the set temperature threshold, or the number of times or duration of temperature exceeding the threshold within a specific time period, an alarm can be issued, helping to promptly detect potential overheating issues and assess the stability of the system under specific conditions.

[0044] Finally, fault characteristics can be determined by calculating the temperature change rate (i.e., the rate of temperature change over time). A rapid temperature rise can indicate an impending failure, allowing for pre-planned maintenance and upgrades, reducing unplanned downtime. This embodiment filters fault characteristics and constructs a training set using these characteristics. This further improves the prediction accuracy of the target prediction model, thereby enhancing the ability to control device status and improving overall efficiency and reliability.

[0045] Lagrange multipliers can be used for model optimization. Optionally, in the processing method of the device provided in the embodiment of the present application, a preset support vector machine model is trained using a model training set, and the trained support vector machine model includes: performing data conversion processing on each group of fault features to obtain M groups of fault indicator data, wherein the data conversion processing method includes a logarithmic transformation method and an exponential transformation method; identifying the correlation state of the M groups of fault indicator data, and determining an initial kernel function according to the correlation state, wherein the initial kernel function refers to a kernel function initialized with function parameters, and the function type of the kernel function is one of the following: a linear kernel function, a polynomial kernel function, and a radial basis kernel function; obtaining an initial regularization coefficient, and configuring the preset support vector machine model according to the initial regularization coefficient and the initial kernel function to obtain an initial support vector machine model; training the initial support vector machine model using the model training set to obtain a first support vector machine model; and Determine the Lagrange multiplier according to the first support vector machine model, calculate the decision boundary of the Lagrange multiplier, update the Lagrange multiplier according to the decision boundary of the Lagrange multiplier to obtain an updated Lagrange multiplier; determine whether the updated Lagrange multiplier meets the convergence condition, and when the updated Lagrange multiplier meets the convergence condition, calculate the first support vector according to the updated Lagrange multiplier, update the first support vector machine model according to the first support vector, and determine the updated support vector machine model as the trained support vector machine model; when the updated Lagrange multiplier does not meet the convergence condition, perform quadratic programming processing on the updated Lagrange multiplier to obtain a processed Lagrange multiplier, until the processed Lagrange multiplier meets the convergence condition, and calculate the second support vector according to the processed Lagrange multiplier; update the updated support vector machine model according to the second support vector to obtain a trained support vector machine model.

[0046] Specifically, in the process of processing fault feature data, the original data must first be preprocessed and converted to improve data quality and usability. At this time, methods such as logarithmic transformation and exponential transformation can be used to transform each set of fault features to reduce data noise and nonlinear problems. Through these processes, multiple sets of fault indicator data can be obtained, providing a basis for subsequent model training and fault identification. After obtaining the fault indicator data, it is also necessary to identify its correlation state to determine the appropriate kernel function. The kernel function is a mathematical function used to map input data to a high-dimensional space to achieve linear separability. Common kernel functions include linear kernel functions, polynomial kernel functions, and radial basis kernel functions. According to the correlation state of the data, an appropriate initial kernel function can be selected and its parameters initialized to adapt to different data characteristics and fault types.

[0047] After obtaining the initial regularization coefficient used to control the model's complexity and generalization ability, the preset support vector machine model is configured according to the initial regularization coefficient and the initial kernel function to obtain the initial support vector machine model. The initial model is then trained using the model training set to optimize the model parameters and improve model performance.

[0048] During model training, it is necessary to determine the Lagrange multiplier used to describe the distribution of data points on the decision boundary. By calculating the Lagrange multiplier, it is possible to determine whether the data point is a support vector or a non-support vector. Based on the decision boundary of the Lagrange multiplier, the multiplier can be updated to optimize the model's performance. During the update of the Lagrange multiplier, it is necessary to determine whether it meets the convergence condition. If the convergence condition is met, it means that the model has converged to the optimal state. The first support vector can be calculated based on the updated Lagrange multiplier, and the first support vector machine model obtained by training the initial support vector machine model can be updated, so that the updated model has better generalization ability and performance.

[0049] If the updated Lagrange multiplier does not meet the convergence conditions, it means that the model has not yet converged to the optimal state. At this time, it is necessary to perform quadratic programming on the Lagrange multiplier to solve the optimal solution. After multiple iterations and optimizations, until the processed Lagrange multiplier meets the convergence conditions, the second support vector is calculated based on the processed Lagrange multiplier, and the updated support vector machine model is updated based on the second support vector. After multiple iterations and optimizations, the updated support vector machine model obtained will have higher accuracy and generalization ability, and the trained support vector machine model can be used for fault identification tasks, which can effectively improve the accuracy and reliability of fault detection. This embodiment obtains a target prediction model with high prediction accuracy by training the target prediction model. When the target prediction model is needed to predict the status of the equipment, it can effectively process fault feature data and identify the fault type, thereby helping staff to quickly locate the faulty equipment and further improve the operating efficiency and safety of the equipment.

[0050] The kernel function may include multiple types. Optionally, in the processing method of the device provided in the embodiment of the present application, the association status of M groups of fault indicator data is identified, and the initial kernel function is determined according to the association status, including: when the association status of the M groups of fault indicator data is a linear state, the function type of the initial kernel function is determined to be the function type of the linear kernel function; when the association status of the M groups of fault indicator data is a curve arrangement state, the function type of the initial kernel function is determined to be the function type of the polynomial kernel function; when the association status of the M groups of fault indicator data is a nonlinear state, the function type of the initial kernel function is determined to be the function type of the radial basis kernel function.

[0051] Specifically, in the fields of machine learning and pattern recognition, kernel functions are a critical component of support vector machine models. The choice of kernel function significantly impacts the performance and results of the algorithm. The appropriate kernel function type can be selected based on the varying correlations in the fault indicator data to optimize the algorithm's performance. When the fault indicator data exhibits linear correlation, this implies a linearly separable boundary between the data points. In this case, a linear kernel function can be selected, effectively addressing linearly separable problems.

[0052] If the correlation of the fault indicator data exhibits a curvilinear arrangement, meaning that there is quadratic or higher-order correlation between the data points, a polynomial kernel function can be selected to capture more complex data features. This kernel function can map the original data into a higher-dimensional space and identify nonlinearly separable boundaries. When the correlation of the fault indicator data is nonlinear, with no clear linear or polynomial relationship between the data points, the radial basis kernel function can be used as the kernel function of the support vector machine model. By introducing a scale parameter and a distance metric, this kernel function can effectively handle nonlinear relationships in high-dimensional space, enabling it to handle very complex data structures.

[0053] It should be noted that since kernel functions may be affected by factors such as data size, dimensionality, and noise, to further improve the model's generalization capabilities, combined kernel functions or customized kernel functions can be considered. This embodiment selects an appropriate kernel function based on the correlation state of the fault indicator data, and then uses the selected kernel function to construct a support vector machine model. The trained target prediction model can then be used to accurately predict the device state of the graphics processing unit, significantly improving the model's accuracy and efficiency.

[0054] Optionally, in the processing method of the device provided in the embodiment of the present application, determining the fault handling strategy based on the device prediction results includes: extracting fault information from the device prediction results to obtain the predicted fault level and fault type; obtaining a fault strategy list, and filtering out the fault handling strategy from the fault strategy list based on the predicted fault level and fault type.

[0055] Specifically, to determine an accurate fault handling strategy, fault information can be extracted from the device prediction results and used to select the most appropriate handling strategy. For example, for a high-level equipment failure, the strategy might be to immediately shut down the machine and replace it; for a low-level temperature anomaly, the strategy might be to adjust the cooling system's operating parameters. This embodiment extracts fault information from the device prediction results and quickly formulates and executes a fault handling strategy based on the fault level and type, thereby minimizing the impact of the fault on production.

[0056] In order to improve processing efficiency, optionally, in the processing method of the device provided in the embodiment of the present application, the fault handling strategy includes a performance optimization strategy and an equipment replacement strategy, and processing the graphics processing unit according to the fault handling strategy includes: when the fault type associated with the equipment prediction result is a performance fault, issuing a first alarm message, and performing performance optimization processing on the graphics processing unit through the performance optimization strategy; when the fault type associated with the equipment prediction result is an equipment fault, issuing a second alarm message, and performing equipment replacement processing on the graphics processing unit through the equipment replacement strategy.

[0057] Specifically, after the target prediction model outputs a device prediction result, if the fault type associated with the device prediction result is a performance fault, a first alarm message can be issued while simultaneously performing performance optimization operations, which can include adjusting parameters, upgrading drivers, cleaning fans, and other operations to ensure maximum device performance. If the fault type associated with the device prediction result is a device fault, the graphics processing unit can be replaced. A second alarm message can be issued while selecting an appropriate backup device for replacement based on the specific circumstances of the device fault to quickly restore the device to normal operation. This embodiment utilizes different device prediction results to determine different fault handling strategies, allowing for rapid resolution of fault issues and minimizing the impact of system failures on work and production.

[0058] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0059] The present application also provides a method for processing a device. Figure 3 is a schematic diagram of a processing method of an optional device according to an embodiment of the present application, such as Figure 3 As shown, the method includes:

[0060] When monitoring and testing a graphics processing unit, it is first necessary to collect the performance data of the graphics processing unit. Figure 4 is a schematic diagram of a method for collecting performance data according to an embodiment of the present application, such as Figure 4As shown, the system can be a hardware management platform that collects device data of the graphics processing unit from the hardware management platform. After identifying the device data, the hardware management platform is initialized and a connection is established between the platform and the graphics processing unit. To further improve the stability and accuracy of the model, the hardware management platform can collect performance data of the graphics processing unit in real time and then perform preliminary processing on it. For example, the performance data can be cleaned and transformed: missing data, outliers, and duplicates are processed, and data normalization, standardization, or logarithmic transformation are performed to obtain fault characteristics, which are then divided into training and test sets. Finally, the training and test sets are used to train the support vector machine model.

[0061] It should be noted that before training the model, it is necessary to select an appropriate kernel function and regularization coefficient. Finally, the initialized support vector machine model is trained using the training set, and performance is evaluated using the test set. After model training is complete, a target prediction model is obtained. At this time, performance data of the graphics processing unit (GPU) within a preset time period can be obtained. Feature extraction processing is performed on this performance data to obtain performance characteristics. These characteristics are input into the target prediction model, and the device prediction result is output. If the device prediction result indicates that the device status of the GPU is in a faulty state, a fault handling strategy is determined based on the device prediction result, and the GPU is handled according to the fault handling strategy.

[0062] This embodiment obtains the performance characteristics of the graphics processing unit on the hardware management platform within a preset time period, inputs the performance characteristics into a target prediction model integrated in the hardware management platform, outputs device prediction results, and uses the device prediction results to optimize and adjust the graphics processing unit, thereby achieving the effect of improving fault handling efficiency.

[0063] This embodiment also provides a processing device for a device, which is used to implement the above-mentioned embodiments and preferred embodiments. Details that have already been described will not be repeated here. As used below, the term "unit" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0064] Figure 5 is a schematic diagram of a processing device according to an embodiment of the present application, such as Figure 5 As shown, the device includes: an acquisition unit 50, an input unit 51, and a determination unit 52.

[0065] An acquisition unit 50 is configured to acquire performance data of a graphics processing unit on a hardware management platform within a preset time period, and perform feature extraction processing on the performance data to obtain performance features;

[0066] An input unit 51 is configured to input the performance characteristics into a target prediction model integrated in the hardware management platform and output a device prediction result, wherein the target prediction model is trained by a preset support vector machine model, and the device prediction result is used to characterize the device status of the graphics processing unit;

[0067] The determining unit 52 is configured to determine a fault handling strategy based on the device prediction result when the device prediction result indicates that the device state of the graphics processing unit is a fault state, and handle the graphics processing unit according to the fault handling strategy.

[0068] It should be noted that the above-mentioned units can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above-mentioned units are all located in the same processor; or the above-mentioned units are located in different processors in any combination.

[0069] The processing device of the device provided in the embodiment of the present application obtains the performance data of the graphics processing unit on the hardware management platform within a preset time period through the acquisition unit 50, performs feature extraction processing on the performance data, and obtains performance characteristics; the input unit 51 inputs the performance characteristics into the target prediction model integrated in the hardware management platform, and outputs the device prediction result, wherein the target prediction model is obtained by training a preset support vector machine model, and the device prediction result is used to characterize the device status of the graphics processing unit; the determination unit 52 determines the fault handling strategy according to the device prediction result when the device prediction result characterizes that the device status of the graphics processing unit is a fault state, and processes the graphics processing unit according to the fault handling strategy, thereby solving the problem of low processing efficiency when performing fault handling on the device in the related art, by obtaining the performance characteristics of the graphics processing unit on the hardware management platform within a preset time period, inputting the performance characteristics into the target prediction model integrated in the hardware management platform, outputting the device prediction result, and using the device prediction result to optimize and adjust the graphics processing unit, thereby achieving the effect of improving the fault handling efficiency.

[0070] Optionally, in the processing device of the equipment provided in the embodiment of the present application, the input unit 51 includes: the target prediction model is trained in the following manner: a first acquisition module, used to obtain the maintenance log and fault record of each hardware device, and filter out maintenance data from M maintenance logs and M fault records according to each group of performance data to obtain M groups of maintenance data, where M is a positive integer; a first processing module, used to perform feature extraction processing on each group of maintenance data to obtain M groups of fault features, obtain performance fault states associated with each group of fault features, and obtain M performance fault states; a composition module, used to construct a model training set from M groups of fault features and performance fault states associated with each group of fault features; a first training module, used to use the model training set to train a preset support vector machine model to obtain a trained support vector machine model, and determine the trained support vector machine model as the target prediction model.

[0071] Optionally, in the processing device of the equipment provided in the embodiment of the present application, the input unit 51 includes: a calculation module, used to calculate each maintenance data for each group of maintenance data to obtain N maintenance indicator data, where N is a positive integer; a second acquisition module, used to obtain a performance threshold list, use the performance threshold list to filter the N maintenance indicator data to obtain Y maintenance indicator data, and determine a set of fault characteristics based on the Y maintenance indicator data, where Y is a positive integer.

[0072] Optionally, in the processing device of the device provided in the embodiment of the present application, the input unit 51 includes: a second processing module, which is used to perform data conversion processing on each group of fault features to obtain M groups of fault indicator data, wherein the data conversion processing method includes a logarithmic transformation method and an exponential transformation method; an identification module, which is used to identify the association state of the M groups of fault indicator data, and determine the initial kernel function according to the association state, wherein the initial kernel function refers to the kernel function initialized by the function parameters, and the function type of the kernel function is one of the following: linear kernel function, polynomial kernel function and radial basis kernel function; a third acquisition module, which is used to obtain the initial regularization coefficient, and configure the preset support vector machine model according to the initial regularization coefficient and the initial kernel function to obtain the initial support vector machine model; a second training module, which is used to train the initial support vector machine model using the model training set to obtain the first support vector machine model; a first determination module, which is used to determine the radial kernel function according to the first support vector machine model. The invention also provides a method for training the support vector machine model by using a Lagrange multiplier, ...

[0073] Optionally, in the processing device of the equipment provided in the embodiment of the present application, the input unit 51 includes: a second determination module, used to determine the function type of the initial kernel function as the function type of the linear kernel function when the association state of the M groups of fault indicator data is a linear state; a third determination module, used to determine the function type of the initial kernel function as the function type of the polynomial kernel function when the association state of the M groups of fault indicator data is a curve arrangement state; and a fourth determination module, used to determine the function type of the initial kernel function as the function type of the radial basis kernel function when the association state of the M groups of fault indicator data is a nonlinear state.

[0074] Optionally, in the processing device of the equipment provided in the embodiment of the present application, the determination unit 52 includes: an extraction module, used to extract fault information from the equipment prediction results to obtain the predicted fault level and fault type; a fourth acquisition module, used to obtain a fault strategy list, and filter out the fault handling strategy from the fault strategy list according to the predicted fault level and fault type.

[0075] Optionally, in the processing device of the equipment provided in the embodiment of the present application, the determination unit 52 includes: a first issuing module, used to issue a first alarm message when the fault type associated with the equipment prediction result is a performance fault, and perform performance optimization processing on the graphics processing unit through a performance optimization strategy; a second issuing module, used to issue a second alarm message when the fault type associated with the equipment prediction result is an equipment fault, and perform equipment replacement processing on the graphics processing unit through an equipment replacement strategy.

[0076] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any of the above method embodiments when run.

[0077] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0078] An embodiment of the present application further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0079] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0080] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail here.

[0081] Obviously, those skilled in the art should understand that the modules or steps of the present application described above can be implemented using a general-purpose computing device, they can be concentrated on a single computing device, or distributed across a network composed of multiple computing devices, they can be implemented using program code executable by the computing device, and thus, they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be performed in a different order than herein, or they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0082] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for processing a device, It is characterized in that include: Acquire performance data of a graphics processing unit on a hardware management platform within a preset time period, perform feature extraction processing on the performance data, and obtain performance features; Inputting the performance characteristics into a target prediction model integrated in the hardware management platform, and outputting a device prediction result, wherein the target prediction model is obtained by training a preset support vector machine model, and the device prediction result is used to characterize the device state of the graphics processing unit; In the case that the device prediction result indicates that the device state of the graphics processing unit is a fault state, a fault processing strategy is determined according to the device prediction result, and the graphics processing unit is processed according to the fault processing strategy.

2. The method according to claim 1, It is characterized in that The target prediction model is trained in the following way: Obtain maintenance logs and fault records of each hardware device, and filter out maintenance data from M maintenance logs and M fault records according to each set of performance data to obtain M sets of maintenance data, where M is a positive integer; Perform feature extraction on each group of maintenance data to obtain M groups of fault features, obtain the performance fault state associated with each group of fault features, and obtain M performance fault states; The M groups of fault features and the performance fault states associated with each group of fault features constitute a model training set; The preset support vector machine model is trained using the model training set to obtain a trained support vector machine model, and the trained support vector machine model is determined as the target prediction model.

3. The method according to claim 2, It is characterized in that Each group of maintenance data is processed by feature extraction to obtain M groups of fault features including: For each group of maintenance data, each maintenance data is calculated to obtain N maintenance index data, where N is a positive integer; A performance threshold list is obtained, the N maintenance indicator data are screened using the performance threshold list to obtain Y maintenance indicator data, and a set of fault characteristics is determined based on the Y maintenance indicator data, where Y is a positive integer.

4. The method according to claim 2, It is characterized in that The preset support vector machine model is trained using the model training set, and the trained support vector machine model obtained includes: Performing data conversion processing on each group of fault features to obtain M groups of fault indicator data, wherein the data conversion processing method includes a logarithmic transformation method and an exponential transformation method; Identify the association state of the M groups of fault indicator data, and determine an initial kernel function according to the association state, wherein the initial kernel function refers to a kernel function initialized by function parameters, and the function type of the kernel function is one of the following: a linear kernel function, a polynomial kernel function, and a radial basis kernel function; Obtaining an initial regularization coefficient, and configuring the preset support vector machine model according to the initial regularization coefficient and the initial kernel function to obtain an initial support vector machine model; Using the model training set to train the initial support vector machine model to obtain a first support vector machine model; Determine a Lagrangian multiplier according to the first support vector machine model, calculate a decision boundary of the Lagrangian multiplier, and update the Lagrangian multiplier according to the decision boundary of the Lagrangian multiplier to obtain an updated Lagrangian multiplier; Determine whether the updated Lagrangian multiplier satisfies a convergence condition, and if the updated Lagrangian multiplier satisfies the convergence condition, calculate a first support vector according to the updated Lagrangian multiplier, update the first support vector machine model according to the first support vector, and determine the updated support vector machine model as the trained support vector machine model; In the case that the updated Lagrangian multiplier does not satisfy the convergence condition, performing quadratic programming processing on the updated Lagrangian multiplier to obtain a processed Lagrangian multiplier until the processed Lagrangian multiplier satisfies the convergence condition, and calculating a second support vector according to the processed Lagrangian multiplier; The updated support vector machine model is updated according to the second support vector to obtain the trained support vector machine model.

5. The method according to claim 4, It is characterized in that Identifying the association status of the M groups of fault indicator data, and determining the initial kernel function according to the association status includes: When the association state of the M groups of fault indicator data is a linear state, determining the function type of the initial kernel function to be the function type of the linear kernel function; When the association state of the M groups of fault indicator data is a curve arrangement state, determining the function type of the initial kernel function to be the function type of the polynomial kernel function; In a case where the association state of the M groups of fault indicator data is a nonlinear state, the function type of the initial kernel function is determined to be the function type of the radial basis kernel function.

6. The method according to claim 1, It is characterized in that Determining a fault handling strategy based on the equipment prediction result includes: Extracting fault information from the equipment prediction result to obtain a predicted fault level and a fault type; A fault strategy list is obtained, and the fault handling strategy is selected from the fault strategy list according to the predicted fault level and the fault type.

7. The method according to claim 6, It is characterized in that The fault handling strategy includes a performance optimization strategy and a device replacement strategy, and processing the graphics processing unit according to the fault handling strategy includes: In the case where the fault type associated with the device prediction result is a performance fault, issuing a first alarm message, and performing performance optimization processing on the graphics processing unit through the performance optimization strategy; In the case where the fault type associated with the device prediction result is a device fault, a second alarm message is issued, and a device replacement process is performed on the graphics processing unit using the device replacement strategy.

8. A processing device for an equipment, It is characterized in that include: An acquisition unit, used to acquire performance data of a graphics processing unit on a hardware management platform within a preset time period, and perform feature extraction processing on the performance data to obtain performance features; An input unit, used to input the performance characteristics into a target prediction model integrated in the hardware management platform, and output a device prediction result, wherein the target prediction model is obtained by training a preset support vector machine model, and the device prediction result is used to characterize the device state of the graphics processing unit; A determination unit is used to determine a fault handling strategy according to the device prediction result when the device prediction result indicates that the device state of the graphics processing unit is a fault state, and to process the graphics processing unit according to the fault handling strategy.

9. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the processing method of the device according to any one of claims 1 to 7 is implemented.

10. An electronic device, It is characterized in that It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the processing method of the device described in any one of claims 1 to 7.

Images