Configuration method of operating system kernel, model training method and related product

Abstract

The invention provides an operating system kernel configuration method, a model training method and a related product, and relates to the technical field of computer networks. According to the method, the obtained hardware configuration information of the computing node and the operation characteristics of the computing node in the operation task state are input into the kernel parameter configuration model, and it can be ensured that the output parameter configuration value is matched with the hardware resource of the computing node and is the parameter configuration value for the specific operation task. Moreover, according to the method, the operating system kernel of the computing node is generated through the parameter configuration value output by the kernel parameter configuration model, and the generated operating system kernel is deployed on the computing node, so that automatic dynamic optimization of the operating kernel of the computing node can be realized, the workload caused by manual configuration and manual optimization is reduced, and the operating efficiency of the computing node is improved. The adaptive capacity of the operating system kernel to hardware resources and operation task requirements is improved, so that the configuration rate and accuracy of the operating system kernel are improved.

Description

Technical Field

[0001] The present application relates to the field of computer network technology, and in particular to an operating system kernel configuration method, a model training method, and related products. Background Art

[0002] High Performance Computing (HPC) clusters are usually composed of multiple computing nodes to handle large-scale and complex operation tasks. The operating system (OS) kernel of the computing node, as a bridge connecting hardware and applications, directly affects the operation efficiency of the entire HPC cluster. However, the currently widely used OS kernels mostly adopt a general design, which undoubtedly limits the full performance of the HPC cluster.

[0003] In the related technology, the OS kernel can be customized for specific HPC cluster environments and application requirements through manual configuration and manual optimization. However, this manual configuration and manual optimization method is not only complex, time-consuming and labor-intensive, but also highly dependent on the professional skills and experience of the operator. It is very easy to cause configuration errors due to human factors, affecting the speed and accuracy of kernel configuration. Summary of the invention

[0004] In order to solve the above problems, the embodiments of the present application provide a configuration method of an operating system kernel, a model training method and related products, which can improve the speed and accuracy of kernel configuration.

[0005] In the first aspect, the embodiment of the present application provides a configuration method for an operating system kernel, the method comprising: obtaining hardware configuration information of a computing node, and operating characteristics of the computing node in a running task state; inputting the hardware configuration information and operating characteristics into a kernel parameter configuration model to obtain parameter configuration values ​​of the operating system kernel of the computing node; generating the operating system kernel of the computing node according to the parameter configuration values; and deploying the generated operating system kernel to the computing node. Thus, the method can ensure that the output parameter configuration values ​​match the hardware resources of the computing node and are parameter configuration values ​​for specific running tasks by inputting the obtained hardware configuration information of the computing node and the operating characteristics of the computing node in a running task state into the kernel parameter configuration model. In addition, the method generates the operating system kernel of the computing node through the parameter configuration values ​​output by the kernel parameter configuration model, and deploys the generated operating system kernel to the computing node, which can realize automatic dynamic optimization of the operating kernel of the computing node, while reducing the workload caused by manual configuration and manual optimization, and improving the adaptability of the operating system kernel to hardware resources and running task requirements, thereby improving the configuration rate and accuracy of the operating system kernel.

[0006] In some specific implementations, the kernel parameter configuration model includes a parameter update sub-model, and the hardware configuration information and operating characteristics are input into the kernel parameter configuration model to obtain the parameter configuration value of the operating system kernel of the computing node, including: determining the parameter items to be updated of the operating system kernel of the computing node by inputting the hardware configuration information and operating characteristics into the parameter update sub-model of the kernel parameter configuration model, and determining the parameter configuration values ​​for the parameter items to be updated.

[0007] In some specific implementations, the kernel parameter configuration model includes a clustering sub-model and multiple parameter update sub-models. By inputting hardware configuration information and operating characteristics into the update sub-model of the kernel parameter configuration model, the parameter items to be updated of the operating system kernel of the computing node are determined, including: by inputting hardware configuration information and operating characteristics into the clustering sub-model of the kernel parameter configuration model, the grouping results corresponding to the hardware configuration information and operating characteristics are determined; by inputting hardware configuration information and operating characteristics into the parameter update sub-model corresponding to the grouping results, the parameter items to be updated of the operating system kernel of the computing node are determined.

[0008] In some specific implementations, the method also includes: if the operating characteristics of the computing node in the running task state after the operating system kernel is deployed are better than the operating characteristics of the computing node in the running task state before the operating system kernel is deployed, then updating the model parameters of the kernel parameter configuration model according to the hardware configuration information, operating characteristics and parameter configuration values ​​corresponding to the operating system kernel.

[0009] In some specific implementations, an operating system kernel of a computing node is generated based on parameter configuration values, including: determining unupdated parameter items of the operating system kernel based on original parameter configuration values ​​of the operating system kernel and parameter configuration values ​​output by a kernel parameter configuration model; generating the operating system kernel of the computing node based on original parameter configuration values ​​of the unupdated parameter items and parameter configuration values ​​output by the kernel parameter configuration model.

[0010] In some specific implementations, the generated operating system kernel is deployed to the computing node, including: when the running task is not running, or when the running feature characterizes that the real-time load of the running task is lower than a preset load threshold, the generated operating system kernel is deployed to the computing node.

[0011] In some specific implementations, hardware configuration information of a computing node and operating characteristics of the computing node when the computing node is in a running task state are obtained, including: obtaining hardware configuration information of multiple computing nodes in a computing cluster and operating characteristics of multiple computing nodes in a running task state; inputting the hardware configuration information and operating characteristics into a kernel parameter configuration model to obtain parameter configuration values ​​of the operating system kernel of the computing node, including: determining grouping results corresponding to multiple computing nodes respectively by inputting the hardware configuration information and operating characteristics of multiple computing nodes into a clustering sub-model of the kernel parameter configuration model; determining parameter items to be updated of the operating system kernels of multiple computing nodes by inputting the hardware configuration information and operating characteristics of multiple computing nodes into parameter update sub-models corresponding to the grouping results corresponding to multiple computing nodes respectively, and determining parameter configuration values ​​for the parameter items to be updated; generating the operating system kernel of the computing node according to the parameter configuration values, including: generating corresponding operating system kernels for multiple computing nodes respectively according to the parameter configuration values ​​of the multiple computing nodes; deploying the generated operating system kernel to the computing nodes, including: deploying the corresponding operating system kernels generated for the multiple computing nodes to the corresponding multiple computing nodes respectively.

[0012] In the second aspect, the embodiment of the present application discloses a model training method, which includes: obtaining a sample data set, the sample data set including hardware configuration information of multiple computing nodes in a historical time period, operating characteristics of multiple computing nodes in a running task state in a historical time period, and parameter configuration values ​​of the operating system kernel of multiple computing nodes in a historical time period; training a machine learning model according to the sample data set to obtain a kernel parameter configuration model. Thus, the method uses the hardware configuration information of multiple computing nodes in a historical time period, the operating characteristics of multiple computing nodes in a running task state in a historical time period, and the parameter configuration values ​​of the operating system kernel of multiple computing nodes in a historical time period as sample data sets to train a machine learning model, which can ensure that the parameter configuration values ​​output in the actual application of the trained kernel parameter configuration model match the hardware resources of the computing node and are parameter configuration values ​​for specific running tasks, thereby reducing the workload caused by manual configuration and manual optimization, while improving the operating system kernel's adaptability to hardware resources and running task requirements, thereby improving the configuration rate and accuracy of the operating system kernel.

[0013] In some specific implementations, obtaining a sample data set includes: obtaining hardware configuration information of multiple computing nodes within a historical time period, and operating characteristics of multiple computing nodes in a running task state within the historical time period; clustering multiple computing nodes according to the hardware configuration information of multiple computing nodes within the historical time period and the operating characteristics of multiple computing nodes in a running task state within the historical time period to obtain multiple grouping results, wherein the similarity of the hardware configuration information of the computing nodes in each grouping result is higher than a first similarity threshold, and the similarity of the operating characteristics of the computing nodes in each grouping result is higher than a second similarity threshold; for each grouping result, determining a parameter item to be updated and determining a parameter configuration value for the parameter item to be updated.

[0014] In the third aspect, the embodiment of the present application provides a configuration device for an operating system kernel, the device comprising: an information acquisition module, a parameter acquisition module, a kernel update module and a kernel deployment module; the information acquisition module is used to obtain the hardware configuration information of the computing node and the operating characteristics of the computing node when it is in the running task state; the parameter acquisition module is used to input the hardware configuration information and the operating characteristics into the kernel parameter configuration model to obtain the parameter configuration value of the operating system kernel of the computing node; the kernel update module is used to generate the operating system kernel of the computing node according to the parameter configuration value; the kernel deployment module is used to deploy the generated operating system kernel to the computing node. Thus, the device can ensure that the output parameter configuration value matches the hardware resources of the computing node and is a parameter configuration value for a specific running task by inputting the acquired hardware configuration information of the computing node and the operating characteristics of the computing node when it is in the running task state into the kernel parameter configuration model. In addition, the device generates an operating system kernel of a computing node through the parameter configuration values ​​output by a kernel parameter configuration model, and deploys the generated operating system kernel to the computing node, thereby realizing automatic dynamic optimization of the operating kernel of the computing node. While reducing the workload caused by manual configuration and manual optimization, the device improves the operating system kernel's adaptability to hardware resources and running task requirements, thereby improving the configuration rate and accuracy of the operating system kernel.

[0015] In the fourth aspect, the embodiment of the present application discloses a model training device, which includes: an acquisition module and a training module; the acquisition module is used to acquire a sample data set, the sample data set includes the hardware configuration information of multiple computing nodes in the historical time period, the operation characteristics of multiple computing nodes in the running task state in the historical time period, and the parameter configuration values ​​of the operating system kernel of multiple computing nodes in the historical time period; the training module is used to train the machine learning model according to the sample data set to obtain the kernel parameter configuration model. Thus, the device trains the machine learning model by using the hardware configuration information of multiple computing nodes in the historical time period, the operation characteristics of multiple computing nodes in the running task state in the historical time period, and the parameter configuration values ​​of the operating system kernel of multiple computing nodes in the historical time period as the sample data set, so as to ensure that the parameter configuration values ​​output in the actual application of the kernel parameter configuration model obtained by training are matched with the hardware resources of the computing node, and are parameter configuration values ​​for specific running tasks, thereby reducing the workload caused by manual configuration and manual optimization, while improving the adaptability of the operating system kernel to the hardware resources and running task requirements, thereby improving the configuration rate and accuracy of the operating system kernel.

[0016] In a fifth aspect, an embodiment of the present application provides an electronic device, the electronic device comprising: a memory for storing a computer program or computer instructions; a processor for executing the computer program or computer instructions stored in the memory, so that the electronic device executes the method of the first aspect or the second aspect.

[0017] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, characterized in that a computer program or instructions are stored in the computer-readable storage medium, and when the computer program or instructions are executed, the method of the first aspect or the second aspect is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A scenario architecture diagram of a method for configuring an operating system kernel provided in an embodiment of the present application;

[0019] Figure 2 A flowchart of a method for configuring an operating system kernel provided in an embodiment of the present application;

[0020] Figure 3 A flowchart of another method for configuring an operating system kernel provided in an embodiment of the present application;

[0021] Figure 4 A flowchart of a method for training a kernel parameter configuration model provided in an embodiment of the present application;

[0022] Figure 5 A flowchart of a model training method provided in an embodiment of the present application;

[0023] Figure 6 A schematic diagram of a configuration device for an operating system kernel provided in an embodiment of the present application;

[0024] Figure 7 A schematic diagram of a model training device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] The terms "first" and "second" in the embodiments of the present application are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.

[0026] First, the technical terms involved in the embodiments of the present application are explained:

[0027] A high performance computing cluster consists of tens to thousands of computing nodes (e.g., multiple computers) interconnected via a high-speed network. Each computing node has high-performance computing capabilities. These computing nodes work together to solve complex operational tasks.

[0028] The operating system (OS) kernel is the core of the computing node. It is responsible for managing the hardware resources of the computing node, providing the program running environment corresponding to the running tasks, and serving as a bridge between user programs and hardware devices.

[0029] As mentioned above, the OS kernel can be customized for specific HPC cluster environments and application requirements through manual configuration and manual optimization. However, this manual configuration and manual optimization method is not only complex, time-consuming and labor-intensive, but also highly dependent on the professional skills and experience of the operator. It is very easy to cause configuration errors due to human factors, affecting the speed and accuracy of kernel configuration.

[0030] In view of this, an embodiment of the present application provides a configuration method for an operating system kernel, a model training method and related products. The configuration method for the operating system kernel includes: obtaining hardware configuration information of a computing node, and operating characteristics of the computing node when it is in a running task state; inputting the hardware configuration information and operating characteristics into a kernel parameter configuration model to obtain parameter configuration values ​​of the operating system kernel of the computing node; generating an operating system kernel of the computing node according to the parameter configuration values; and deploying the generated operating system kernel to the computing node.

[0031] Therefore, the configuration method of the operating system kernel obtains the hardware configuration information of the computing node, obtains the operating characteristics of the computing node when it is in the running task state, and inputs the hardware configuration information and the operating characteristics into the kernel parameter configuration model, so as to ensure that the output parameter configuration value matches the hardware resources of the computing node and is the parameter configuration value for a specific running task.

[0032] Furthermore, by updating the operating system kernel of the computing node through the parameter configuration values ​​output by the kernel parameter configuration model and deploying the updated operating system kernel to the computing node, automatic and dynamic optimization of the operating kernel of the computing node can be achieved. While reducing the workload caused by traditional manual configuration and manual optimization, the operating system kernel's adaptability to hardware resources and running task requirements is improved, thereby improving the speed and accuracy of kernel configuration.

[0033] Furthermore, since the parameter configuration values ​​are dynamically generated based on the hardware configuration information of the computing node and the operating characteristics of the computing node in the running task state, they can better adapt to changes in hardware resources and running task requirements. This adaptability ensures that the computing node can maintain efficient operation under different loads and task requirements. In addition, the updated operating system kernel can better utilize the hardware resources of the computing node, thereby improving computing efficiency while reducing the risk of computing node crashes or performance degradation due to insufficient hardware resources.

[0034] Next, the execution subject of the configuration method of the operating system kernel provided in the embodiment of the present application is specifically introduced:

[0035] The execution subject of the configuration method of the operating system kernel provided in the embodiment of the present application can be a computer device with data processing capabilities, and the computer device can specifically be a terminal device or a server. As an example, the terminal device can specifically include but is not limited to mobile phones, desktop computers, tablet computers, notebook computers, PDAs, intelligent voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. In addition, the server can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. In addition, the configuration method of the operating system kernel provided in the embodiment of the present application can also be executed collaboratively by the terminal device and the server. Among them, the terminal and the server can be directly or indirectly connected by wired or wireless communication, and this application is not limited here. Therefore, in the embodiment of the present application, the implementation subject of the technical solution of the present application is not limited.

[0036] See also Figure 1 , which is a scenario architecture diagram of a configuration method of an operating system kernel provided in an embodiment of the present application, including a computer device 100. The computer device 100 can be a terminal device or a server in the above-mentioned various forms. Specifically, the following embodiment is exemplified by taking the computer device 100 as server A.

[0037] Server A is used to obtain the hardware configuration information of the computing node and the operating characteristics of the computing node when it is in the running task state. It is understandable that obtaining the hardware configuration information is crucial to understanding the physical capabilities of the computing node, because the hardware configuration information determines the type and scale of tasks that the computing node can handle. Obtaining the operating characteristics helps to understand the actual pressure of the running tasks on the computing node, as well as the bottlenecks or optimization points that the computing node may encounter when processing these running tasks.

[0038] Server A is also used to input the hardware configuration information and operation characteristics into the kernel parameter configuration model to obtain the parameter configuration value of the operating system kernel of the computing node. It can be understood that inputting the hardware configuration information and operation characteristics into the pre-trained kernel parameter configuration model can ensure that the parameter configuration value output by the kernel parameter configuration model not only matches the hardware resources of the computing node, but also is a parameter configuration value that optimizes the operating system kernel of the computing node for a specific operation task, that is, the parameter configuration value is specifically designed to maximize the computing performance of the computing node when processing a specific operation task under specific hardware resources.

[0039] Server A is also used to update the operating system kernel of the computing node according to the generated parameter configuration value, which usually involves modifying the kernel configuration file, adjusting the kernel startup parameters, or sending configuration update instructions to the kernel through a specific management interface.

[0040] Server A is also used to deploy the updated operating system kernel to the computing node. This means that the new operating system kernel will take effect soon, and the computing node will run according to the new operating system kernel. It can be understood that by updating the operating system kernel of the computing node through the parameter configuration value output by the kernel parameter configuration model and deploying the updated operating system kernel to the computing node, it is possible to achieve automatic and dynamic optimization of the operating kernel of the computing node, while reducing the workload caused by manual configuration and manual optimization, and improving the operating system kernel's adaptability to hardware resources and running task requirements, thereby improving the speed and accuracy of kernel configuration.

[0041] Next, taking server A as the execution subject, the configuration method of the operating system kernel provided by the embodiment of the present application is specifically introduced:

[0042] See also Figure 2, which is a flow chart of a configuration method of an operating system kernel provided by an embodiment of the present application. Figure 2 The configuration method of the operating system kernel shown includes the following steps:

[0043] S201: Acquire hardware configuration information of a computing node and operating characteristics of the computing node in a running task state.

[0044] The purpose of obtaining the hardware configuration information of the computing node is to understand the physical capabilities of the computing node, including the CPU model and quantity, memory size, storage device type (such as SSD or HDD) and its capacity, network interface type and speed, etc. In some specific implementations, system commands (such as lscpu command, free-m command, lsblk command and ethtool command, etc.) can be used to query the hardware configuration information of the computing node, or the hardware configuration information can be obtained by reading the operating system configuration file of the computing node, or a special hardware detection tool (such as dmidecode, lshw) can be used to obtain the hardware configuration information, etc. It is understandable that the hardware configuration information determines the type and scale of tasks that the computing node can handle. For example, a computing node with a hardware configuration of a high-performance CPU and sufficient memory is more suitable for running computing-intensive tasks, and a computing node with a hardware configuration of fast storage and network interfaces is more suitable for running I / O-intensive tasks.

[0045] The purpose of obtaining the operation characteristics of the computing node in the running task state is to understand the actual pressure of the running task on the computing node, as well as the bottlenecks or optimization points that the computing node may encounter when processing the running task, including CPU usage, memory usage, cache usage, network bandwidth, hard disk IO and read and write rates, etc. In some specific implementations, the operation characteristics of the computing node in the running task state can be observed in real time through monitoring tools set on the computing node (such as top command, htop command, vmstat command and iostat command, etc.), or log analysis tools (such as awk, sed, grep, etc.) can be used to parse historical data in log files to extract the operation characteristics of the computing node in the running task state. It can be understood that obtaining the operation characteristics helps to identify potential bottlenecks of the computing node, such as CPU overload, insufficient memory, disk I / O bottleneck, etc., and this information is crucial for generating parameter configuration values ​​optimized for specific running tasks.

[0046] Therefore, by understanding the actual requirements of running tasks and the physical capabilities of computing nodes, more reasonable and effective parameter configuration values ​​can be generated, thereby improving the efficiency of kernel configuration.

[0047] S202: Input the hardware configuration information and the operation characteristics into the kernel parameter configuration model to obtain the parameter configuration value of the operating system kernel of the computing node.

[0048] The kernel parameter configuration model is a pre-trained machine learning model. Hardware configuration information and operating characteristics can be input into the kernel parameter configuration model so that the kernel parameter configuration model outputs parameter configuration values ​​of the operating system kernel according to the input hardware configuration information and operating characteristics.

[0049] The parameter configuration values ​​output by the kernel parameter configuration model are generated based on the physical capabilities of the computing node and the requirements of the specific running tasks. Therefore, the parameter configuration values ​​help to maximize the hardware performance of the computing node while minimizing the resource consumption and potential performance bottlenecks of the computing node. Exemplarily, the parameter configuration values ​​can be parameter configuration values ​​for the settings of the kernel scheduler (such as the type of scheduler, scheduling policy, thread priority, etc.), or parameter configuration values ​​for the memory management policy (such as memory allocation policy, page replacement algorithm, memory compression and deduplication, etc.), or parameter configuration values ​​for the network stack configuration (such as optimization of network interfaces, management and control of network traffic, etc.), or parameter configuration values ​​for the file system (such as the type of file system, mount options, I / O scheduler, etc.), etc. For specific parameter configuration values, this application does not limit them.

[0050] S203: Generate an operating system kernel of the computing node according to the parameter configuration value.

[0051] It should be noted that before generating the operating system kernel of the computing node according to the parameter configuration values, a pre-verification may be performed to ensure that the obtained parameter configuration values ​​will not cause instability or performance degradation of the operating system.

[0052] In some specific implementations, a test environment (such as a virtual machine, a spare physical server, etc.) similar to the hardware configuration of the computing node can be first created, and the parameter configuration values ​​obtained from the kernel parameter configuration model can be simulated in the test environment. Subsequently, performance monitoring tools (such as top, htop, vmstat, iostat, etc.) are used to evaluate the operating characteristics (such as key indicators such as execution speed, throughput, and response time). If the operating characteristics are better (such as increased execution speed, increased throughput, and decreased response time), the operating system kernel of the computing node is generated according to the parameter configuration values.

[0053] In another specific implementation, you can also refer to the official documentation of the operating system and hardware to ensure that the parameter configuration values ​​are compatible and within a safe range. If the parameter configuration values ​​are compatible and within a safe range, then generate the operating system kernel of the computing node based on the parameter configuration values.

[0054] It should also be noted that before generating the operating system kernel of the computing node according to the parameter configuration values, the original parameter configuration values ​​of the current operating system kernel can also be backed up so that it can be rolled back to the original state when a problem occurs, thereby avoiding the risk of operating system instability or performance degradation.

[0055] It should also be noted that the specific method for generating the kernel configuration will vary depending on the type and version of the operating system. In one example, a new operating system kernel can be generated directly based on the parameter configuration values ​​output by the kernel parameter configuration model. The advantage of this method is that the functions and features of the kernel can be fully customized to meet specific needs. In another example, the unupdated parameter items of the operating system kernel can also be determined first based on the original parameter configuration values ​​of the operating system kernel and the parameter configuration values ​​output by the kernel parameter configuration model, and then the operating system kernel of the computing node is generated based on the original parameter configuration values ​​of the unupdated parameter items and the parameter configuration values ​​output by the kernel parameter configuration model. This method can be updated on the basis of retaining the existing kernel, reducing the workload, avoiding unnecessary changes, and ensuring the stability of the system. This application is not limited to this.

[0056] S204: Deploy the generated operating system kernel to the computing node.

[0057] The generated operating system kernel file can be uploaded to the specified location of the computing node through network transmission methods such as Secure copy (SCP) and Secure File Transfer Protocol (SFTP) or local copying to complete the deployment of the operating system kernel.

[0058] It should be noted that choosing the right time is critical when deploying a new kernel. This is because restarting a compute node to load a new operating system kernel will interrupt all running tasks on the compute node. In order to minimize the impact on operating system availability and performance, deployment is usually performed during the following time periods: First, choose to deploy during the time period between running tasks, that is, the time period when no running tasks are being executed, so as to ensure that there is no interference with the ongoing running tasks during the deployment process. This usually occurs during non-working hours such as nighttime or weekends, when the operating system load is low and user activity is low. Second, choose to deploy during the off-peak period of the HPC cluster operation, such as when the real-time load of the running tasks is below the preset load threshold. This can be determined by monitoring the load and resource usage of the HPC cluster. Deploying when the cluster load is low can minimize the impact on the overall performance of the cluster.

[0059] In summary, the embodiment of the present application discloses a configuration method for an operating system kernel, which can ensure that the output parameter configuration value matches the hardware resources of the computing node and is a parameter configuration value for a specific running task by inputting the acquired hardware configuration information of the computing node and the running characteristics of the computing node in the running task state into the kernel parameter configuration model. In addition, the method generates the operating system kernel of the computing node through the parameter configuration value output by the kernel parameter configuration model, and deploys the generated operating system kernel to the computing node, which can realize automatic dynamic optimization of the operating kernel of the computing node, while reducing the workload caused by manual configuration and manual optimization, and improving the operating system kernel's adaptability to hardware resources and running task requirements, thereby improving the configuration rate and accuracy of the operating system kernel.

[0060] See also Figure 3 , which is a flow chart of another method for configuring an operating system kernel provided by an embodiment of the present application. Figure 3 The configuration method of the operating system kernel shown includes the following steps:

[0061] S301: Obtain hardware configuration information of multiple computing nodes in an HPC cluster, and operating characteristics of the multiple computing nodes in a running task state.

[0062] It can be understood that by obtaining the hardware configuration information of multiple computing nodes in the computing cluster, as well as the operating characteristics of multiple computing nodes in the running task state, it can be ensured that the subsequently generated operating system kernel not only matches the hardware resources of the computing nodes, but is also an operating system kernel for specific running tasks, thereby improving the operating system kernel's adaptability to hardware resources and running task requirements, and further improving kernel configuration and accuracy.

[0063] S302: Inputting hardware configuration information and operation characteristics of multiple computing nodes into a kernel parameter configuration model to obtain parameter configuration values ​​of operating system kernels of the multiple computing nodes.

[0064] The kernel parameter configuration model is a pre-trained machine learning model. By inputting the hardware configuration information and operating characteristics of multiple computing nodes into the kernel parameter configuration model, the kernel parameter configuration model can output parameter configuration values ​​of the operating system kernels of multiple computing nodes based on the input hardware configuration information and operating characteristics.

[0065] It should be noted that the kernel parameter configuration model may include a clustering sub-model and multiple parameter update sub-models. The purpose of the clustering sub-model is to cluster multiple computing nodes into multiple grouping results with similar characteristics based on hardware configuration information and operating characteristics, so as to provide more personalized operating system kernel parameter setting suggestions for different hardware configuration information and operating characteristics. The purpose of the parameter update sub-model is to determine which parameters need to be adjusted based on specific hardware configuration information and operating characteristics, and to provide recommended parameter configuration values ​​for these parameters.

[0066] Specifically, first, by inputting the hardware configuration information and operation characteristics of multiple computing nodes into the clustering sub-model of the kernel parameter configuration model, the clustering sub-model will classify the multiple computing nodes according to a preset algorithm (such as K-means, DBSCAN, etc.), thereby determining the grouping results corresponding to the multiple computing nodes. Subsequently, by inputting the hardware configuration information and operation characteristics of the multiple computing nodes into the parameter update sub-model corresponding to the grouping results, the parameter items to be updated of the operating system kernel of the multiple computing nodes are determined, and the parameter configuration values ​​are determined for the parameter items to be updated, so as to achieve the purpose of optimizing performance.

[0067] Next, a training method for a kernel parameter configuration model provided by an embodiment of the present application is introduced. Figure 4 , which is a flow chart of a training method for a kernel parameter configuration model provided by an embodiment of the present application. Specifically, the training method for the kernel parameter configuration model can be shown in the following steps:

[0068] A1: Obtain hardware configuration information of multiple computing nodes in a historical time period, and operating characteristics of multiple computing nodes in a running task state in a historical time period.

[0069] A2: According to the hardware configuration information of the multiple computing nodes in the historical time period and the running characteristics of the multiple computing nodes in the running task state in the historical time period, the multiple computing nodes are clustered to obtain multiple grouping results.

[0070] The purpose of cluster analysis (CA) is to find out which computing nodes in the HPC cluster have similar characteristics (ie, similar hardware configuration information and similar operating characteristics) during a historical period.

[0071] In the configuration method of the operating system kernel provided in the embodiment of the present application, computing nodes with similar hardware configuration information and operating characteristics can be clustered by statistical methods to obtain several grouping results.

[0072] Specifically, first, the collected hardware configuration information and operating characteristics are preprocessed. In one example, outliers, missing values, etc. in the hardware configuration information and operating characteristics can be removed to ensure the integrity and accuracy of the hardware configuration information and operating characteristics. In another example, the hardware configuration information and operating characteristics can also be standardized to remove the influence of different feature dimensions, so that the hardware configuration information between different computing nodes and the operating characteristics between different computing nodes are more comparable.

[0073] Subsequently, the similarity of the preprocessed hardware configuration information and the operating characteristics between different computing nodes is calculated. Exemplarily, the similarity of the preprocessed hardware configuration information between different computing nodes can be compared by calculating cosine similarity, Euclidean distance, etc., and multiple computing nodes with similarities higher than a first similarity threshold are preliminarily clustered, and then the similarity of the preprocessed operating characteristics between the preliminarily clustered computing nodes is compared, and multiple computing nodes with similarities higher than a second similarity threshold are secondary clustered, thereby obtaining the final grouping result.

[0074] Finally, select a suitable clustering algorithm (such as K-means, hierarchical clustering, DBSCAN, etc.) to cluster the computing nodes according to the calculated similarity. These clustering algorithms can automatically classify computing nodes with high similarity into one category and obtain several grouping results.

[0075] It should be noted that after obtaining several grouping results, the grouping results can also be comprehensively evaluated to ensure that the grouping results meet expectations. In one example, the compactness of each grouping result can be evaluated, that is, the similarity of the computing nodes in the same grouping result. If the similarity of the hardware configuration information of the computing nodes in the same grouping result is higher than the first similarity threshold, and the similarity of the operating characteristics of the computing nodes in the same grouping result is higher than the second similarity threshold, the subsequent A3 step is performed. In another example, the separability of each grouping result can be evaluated, that is, the similarity of the computing nodes between different grouping results. If the similarity between the hardware configuration information corresponding to the multiple grouping results is lower than the third similarity threshold, and the similarity between the operating characteristics corresponding to the multiple grouping results is lower than the fourth similarity threshold, the subsequent A3 step is performed. For the specific evaluation method, this application is not limited. Thus, after cluster analysis, all computing nodes in the HPC cluster will be divided into several grouping results with similar characteristics. The computing nodes in each grouping result have high similarity in hardware configuration and operating characteristics, while the computing nodes between different grouping results have greater differences.

[0076] It is understandable that by clustering computing nodes with similar characteristics, the machine learning model can more easily learn the characteristics of the computing nodes in each grouping result during the subsequent model training process, thereby helping to improve the learning efficiency and accuracy of the model, and also providing strong support for customized optimization of HPC clusters.

[0077] A3: For each grouping result, the parameter items to be updated are determined through a decision tree model or a random forest model.

[0078] Determining the parameter items to be updated refers to identifying the factors affecting the performance bottleneck. The decision tree (DT) model is used to analyze the hardware configuration information and operating characteristics of the internal computing nodes of the grouping results. By recursively selecting the optimal features for division, the decision tree model can gradually build a model that can accurately predict performance bottlenecks. Specifically, each internal computing node of the decision tree model represents a test of a feature attribute, each branch represents the output of the feature attribute in a certain value range, and each leaf node corresponds to a performance bottleneck category or specific value. During the training process, the decision tree model will continuously try to find the features that can best distinguish different performance states, thereby building a decision path that can accurately reflect the performance bottleneck.

[0079] The Random Forest (RF) model is used to further verify and refine the prediction results of the decision tree. Specifically, the Random Forest model randomly constructs multiple decision trees and combines the prediction results of each decision tree by voting or averaging. This method can significantly reduce the risk of overfitting that may occur in a single decision tree model, while improving the generalization ability of unknown data, thereby more accurately identifying the main factors affecting performance bottlenecks.

[0080] A4: For each grouping result, determine a parameter configuration value for the parameter item to be updated.

[0081] Parameter configuration values ​​are used to optimize the running characteristics of running tasks, such as improving execution speed, throughput, and reducing response time.

[0082] In some specific implementations, a more suitable parameter configuration value of the operating system kernel can be generated for each grouping result directly based on the parameter items to be updated of each grouping result (such as hardware resources, software configuration, algorithm efficiency, etc.). In one example, the parameter item to be updated corresponding to grouping result 1 may be a memory management parameter item, then for this grouping result, the parameter configuration value generated may include increasing the size of the memory page, optimizing the memory defragmentation algorithm, or enabling a more efficient memory compression technology, etc. In another example, the parameter item to be updated corresponding to grouping result 2 may be a CPU scheduling parameter item. Then for this grouping result, the parameter configuration value generated may include optimizing the CPU scheduling strategy, such as adjusting the priority scheduling algorithm of the process, enabling multi-core parallel processing, or optimizing the CPU power management strategy, etc.

[0083] A5: According to the hardware configuration information of multiple computing nodes in the historical time period, the operating characteristics of multiple computing nodes in the running task state in the historical time period, and the parameter configuration values ​​of the operating system kernels of multiple computing nodes in the historical time period, a machine learning model is trained to obtain a kernel parameter configuration model.

[0084] Next, an HPC cluster includes 1000 computing nodes, of which 500 computing nodes mainly run memory-intensive tasks (type A) and 500 computing nodes mainly run CPU-intensive tasks (type B) as an example for explanation:

[0085] First, the hardware configuration information (such as CPU model, memory size, etc.) and operation characteristics (such as execution speed, response time, throughput, etc.) of all computing nodes are collected. Secondly, computing nodes with similar hardware configuration information and operation characteristics are clustered, that is, type A and type B are distinguished, and several grouping results of type A and several grouping results of type B are obtained. The purpose of this is to group nodes with similar characteristics for more efficient management and optimization. Then, for several grouping results of type A and several grouping results of type B, the parameter items to be updated and the parameter configuration values ​​for the parameter items to be updated are determined by decision tree model or random forest model respectively. These parameter configuration values ​​are intended to optimize the performance of computing nodes to adapt to their specific hardware and operation characteristics. Finally, according to the hardware configuration information, operation characteristics and generated parameter configuration values ​​of type A and type B nodes, machine learning models are trained respectively. For type A nodes, kernel parameter configuration model A is obtained by training; for type B nodes, kernel parameter configuration model B is obtained by training. Kernel parameter configuration model A and kernel parameter configuration model B can be used in the future to predict and optimize computing nodes newly added to the HPC cluster, or dynamically adjust kernel configuration according to changes in current computing nodes to improve overall performance and resource utilization.

[0086] S303: Generate corresponding operating system kernels for the multiple computing nodes respectively according to the parameter configuration values ​​of the multiple computing nodes.

[0087] According to the parameter configuration values ​​of the multiple computing nodes, corresponding operating system kernels are generated for the multiple computing nodes respectively, and it is ensured that the generated operating system kernel can provide the best performance for the specific hardware configuration information and operation characteristics.

[0088] It should be noted that the specific method for generating the kernel configuration may vary depending on the type and version of the operating system. In one example, a new operating system kernel may be generated directly based on the parameter configuration values ​​output by the kernel parameter configuration model. The advantage of this method is that the functions and features of the kernel can be fully customized to meet specific needs. In another example, the unupdated parameter items of the operating system kernel may also be determined first based on the original parameter configuration values ​​of the operating system kernel and the parameter configuration values ​​output by the kernel parameter configuration model, and then the operating system kernel of the computing node may be generated based on the original parameter configuration values ​​of the unupdated parameter items and the parameter configuration values ​​output by the kernel parameter configuration model. This method can be updated on the basis of retaining the existing kernel, reducing the workload, avoiding unnecessary changes, and ensuring the stability of the system. This application is not limited to this.

[0089] S304: Deploy the generated multiple operating system kernels to the corresponding multiple computing nodes, and restart the multiple computing nodes to obtain multiple computing nodes after the operating system kernels are deployed.

[0090] In some specific implementations, it is necessary to deploy multiple generated operating system kernels to the corresponding multiple computing nodes through deployment methods such as network file transmission, installation using automated scripts or management tools. In addition, since the operating system kernel is the core of the OS, any changes to the operating system kernel need to be made effective by restarting the operating system. This means that after deploying a new operating system kernel, the computing node needs to be restarted so that the new operating system kernel can take over the control of the operating system and start running, that is, to obtain the computing node after the operating system kernel is deployed.

[0091] It should be noted that choosing the right time is crucial when deploying a new operating system kernel. This is because restarting a compute node will cause all running tasks on the compute node to be interrupted. Therefore, such operations are usually performed during task gaps (i.e., a period of time when no tasks are running) or during low-peak hours of cluster operation (i.e., a period of time when the real-time load of running tasks is lower than the preset load threshold) to minimize the impact on OS availability and performance.

[0092] S305: Monitor the operation characteristics of each computing node of the HPC cluster after the operating system kernel is deployed.

[0093] Monitoring the operation characteristics of each computing node after the operating system kernel is deployed in the HPC cluster includes monitoring the execution speed, response time, and throughput of each computing node after the operating system kernel is deployed.

[0094] S306: Update the model parameters of the kernel parameter configuration model according to the operation characteristics of each computing node after the operating system kernel is deployed.

[0095] First, the operation characteristics of each computing node after the operating system kernel is deployed are deeply analyzed. If the computing node before the operating system kernel is deployed and the computing node after the operating system kernel is deployed run the same operation task, the operation characteristics of the computing node after the operating system kernel is deployed in the operation task state are better than the operation characteristics of the computing node before the operating system kernel is deployed in the operation task state (that is, the execution speed corresponding to the computing node after the operating system kernel is deployed is greater than the execution speed corresponding to the computing node before the operating system kernel is deployed, or the throughput corresponding to the computing node after the operating system kernel is deployed is greater than the throughput corresponding to the computing node before the operating system kernel is deployed, or the response time corresponding to the computing node after the operating system kernel is deployed is less than the response time corresponding to the computing node before the operating system kernel is deployed, any one or more of the following is satisfied), then according to the hardware configuration information, operation characteristics and parameter configuration values ​​corresponding to the operating system kernel, the model parameters of the kernel parameter configuration model are updated, so that the kernel parameter configuration model can better learn which parameters can bring about the improvement of operation performance.

[0096] It is understandable that the configuration method of the traditional operating system kernel is usually a one-time setting when the OS is installed or configured. This method may be able to provide good performance in the initial stage, but it is difficult to adapt to the dynamic changes in the operating environment and task requirements. With the upgrade of hardware, the update of software versions, and changes in workloads, these static settings may gradually become no longer applicable, and may even become a performance bottleneck. In contrast, the configuration method of the operating system kernel provided in this application can be dynamically adjusted according to the hardware configuration information and operating characteristics of the HPC cluster. By monitoring the changes in the hardware configuration information and operating characteristics of the HPC cluster in real time, performance problems can be discovered and resolved in a timely manner, avoiding the emergence of performance bottlenecks, and enabling the operating system to better adapt to various application scenarios.

[0097] Furthermore, the configuration method of the operating system kernel provided by the present application can also quickly identify the physical capabilities and performance requirements of new computing nodes when they join the HPC cluster, and automatically adjust the OS kernel configuration according to the physical capabilities and performance requirements of the new computing nodes to ensure that the new computing nodes can be integrated into the cluster efficiently and stably. This means that new computing nodes can be put into use immediately without going through tedious manual configuration and testing processes, thereby greatly reducing the time cost of cluster expansion.

[0098] In summary, the present application provides a configuration method for an operating system kernel, whereby the configuration method for the operating system kernel can ensure that the output parameter configuration value matches the hardware resources of the computing node and is a parameter configuration value for a specific running task by obtaining the hardware configuration information of the computing node and obtaining the running characteristics of the computing node in the running task state, and inputting the hardware configuration information and the running characteristics into the kernel parameter configuration model. In addition, the operating system kernel of the computing node is updated by the parameter configuration value output by the kernel parameter configuration model, and the updated operating system kernel is deployed on the computing node, so that the automatic and dynamic optimization of the operating kernel of the computing node can be realized, while reducing the workload caused by traditional manual configuration and manual optimization, the operating system kernel is improved. The adaptability of the hardware resources and running task requirements of the operating system kernel is improved, thereby improving the speed and accuracy of the kernel configuration. Further, since the parameter configuration value is dynamically generated according to the hardware configuration information of the computing node and the running characteristics of the computing node in the running task state, it can better adapt to the changes in hardware resources and running task requirements, and this adaptability ensures that the computing node can maintain efficient operation under different loads and task requirements. In addition, applying the updated operating system kernel can better utilize the hardware resources of the computing nodes, thereby improving computing efficiency while reducing the risk of computing node crashes or performance degradation due to insufficient hardware resources.

[0099] Next, a model training method provided by an embodiment of the present application is introduced. Figure 5 , which is a flow chart of a model training method provided in an embodiment of the present application. Specifically, the model training method can be shown in the following steps:

[0100] S501: Obtain a sample data set, where the sample data set includes hardware configuration information of multiple computing nodes in a historical time period, operating characteristics of multiple computing nodes in a running task state in a historical time period, and parameter configuration values ​​of operating system kernels of multiple computing nodes in a historical time period.

[0101] In some specific implementations, obtaining a sample data set includes: obtaining hardware configuration information of multiple computing nodes within a historical time period, and operating characteristics of multiple computing nodes in a running task state within the historical time period; clustering multiple computing nodes according to the hardware configuration information of multiple computing nodes within the historical time period and the operating characteristics of multiple computing nodes in a running task state within the historical time period to obtain multiple grouping results, wherein the similarity of the hardware configuration information of the computing nodes in each grouping result is higher than a first similarity threshold, and the similarity of the operating characteristics of the computing nodes in each grouping result is higher than a second similarity threshold; for each grouping result, determining a parameter item to be updated and determining a parameter configuration value for the parameter item to be updated.

[0102] S502: Train a machine learning model according to the sample data set to obtain a kernel parameter configuration model.

[0103] In summary, an embodiment of the present application provides a model training method, which trains a machine learning model by using the hardware configuration information of multiple computing nodes in a historical time period, the operating characteristics of multiple computing nodes in a running task state in a historical time period, and the parameter configuration values ​​of the operating system kernel of multiple computing nodes in a historical time period as a sample data set. The method can ensure that the parameter configuration values ​​output in the actual application of the trained kernel parameter configuration model match the hardware resources of the computing node and are parameter configuration values ​​for specific running tasks, thereby reducing the workload caused by manual configuration and manual optimization while improving the adaptability of the operating system kernel to hardware resources and running task requirements, thereby improving the configuration rate and accuracy of the operating system kernel.

[0104] Based on the configuration method of the operating system kernel provided in the above embodiment, the present application also provides a configuration device of the operating system kernel. The configuration device of the operating system kernel provided in the embodiment of the present application is specifically introduced as follows:

[0105] See also Figure 6 , which is a schematic diagram of a configuration device for an operating system kernel provided by an embodiment of the present application. The configuration device 600 for the operating system kernel includes: an information acquisition module 601, a parameter acquisition module 602, a kernel update module 603 and a kernel deployment module 604.

[0106] The information acquisition module 601 is used to acquire the hardware configuration information of the computing node and the operation characteristics of the computing node when it is in the running task state;

[0107] The parameter acquisition module 602 is used to input the hardware configuration information and the operation characteristics into the kernel parameter configuration model to obtain the parameter configuration value of the operating system kernel of the computing node;

[0108] The kernel update module 603 is used to generate the operating system kernel of the computing node according to the parameter configuration value;

[0109] The kernel deployment module 604 is used to deploy the generated operating system kernel to the computing node.

[0110] In some specific implementations, the kernel parameter configuration model includes a parameter update sub-model, and the parameter acquisition module 602 is specifically used to: determine the parameter items to be updated of the operating system kernel of the computing node by inputting hardware configuration information and operating characteristics into the parameter update sub-model of the kernel parameter configuration model, and determine the parameter configuration values ​​for the parameter items to be updated.

[0111] In some specific implementations, the kernel parameter configuration model includes a clustering sub-model and multiple parameter update sub-models, and the parameter acquisition module 602 is specifically used to: determine the grouping results corresponding to the hardware configuration information and operating characteristics by inputting the hardware configuration information and operating characteristics into the clustering sub-model of the kernel parameter configuration model; determine the parameter items to be updated of the operating system kernel of the computing node by inputting the hardware configuration information and operating characteristics into the parameter update sub-model corresponding to the grouping results.

[0112] In some specific implementations, the configuration device 600 of the operating system kernel also includes: a parameter update module; the parameter update module is used to: if the operating characteristics of the computing node in the running task state after the operating system kernel is deployed are better than the operating characteristics of the computing node in the running task state before the operating system kernel is deployed, then update the model parameters of the kernel parameter configuration model according to the hardware configuration information, operating characteristics and parameter configuration values ​​corresponding to the operating system kernel.

[0113] In some specific implementations, the kernel update module 603 is specifically used to: determine the non-updated parameter items of the operating system kernel based on the original parameter configuration values ​​of the operating system kernel and the parameter configuration values ​​output by the kernel parameter configuration model; generate the operating system kernel of the computing node based on the original parameter configuration values ​​of the non-updated parameter items and the parameter configuration values ​​output by the kernel parameter configuration model.

[0114] In some specific implementations, the kernel deployment module 604 is specifically used to deploy the generated operating system kernel to the computing node when the running task is not running, or when the running characteristics characterize that the real-time load of the running task is lower than a preset load threshold.

[0115] In some specific implementations, the information acquisition module 601 is specifically used to: obtain the hardware configuration information of multiple computing nodes in the computing cluster, and the operating characteristics of the multiple computing nodes in the running task state; the parameter acquisition module 602 is specifically used to: determine the grouping results corresponding to the multiple computing nodes by inputting the hardware configuration information and operating characteristics of the multiple computing nodes into the clustering sub-model of the kernel parameter configuration model; determine the parameter items to be updated of the operating system kernel of the multiple computing nodes by inputting the hardware configuration information and operating characteristics of the multiple computing nodes into the parameter update sub-model corresponding to the grouping results corresponding to the multiple computing nodes, and determine the parameter configuration values ​​for the parameter items to be updated; the kernel update module 603 is specifically used to: generate corresponding operating system kernels for the multiple computing nodes according to the parameter configuration values ​​of the multiple computing nodes; the kernel deployment module 604 is specifically used to: deploy the corresponding operating system kernels generated for the multiple computing nodes to the corresponding multiple computing nodes.

[0116] In summary, the present application provides a configuration device for an operating system kernel, whereby the configuration device for the operating system kernel can ensure that the output parameter configuration value matches the hardware resources of the computing node and is a parameter configuration value for a specific running task by obtaining the hardware configuration information of the computing node and obtaining the running characteristics of the computing node in the running task state, and inputting the hardware configuration information and the running characteristics into the kernel parameter configuration model. In addition, the operating system kernel of the computing node is updated by the parameter configuration value output by the kernel parameter configuration model, and the updated operating system kernel is deployed on the computing node, so that the automatic and dynamic optimization of the operating kernel of the computing node can be realized, while reducing the workload caused by traditional manual configuration and manual optimization, the operating system kernel is improved. The adaptability of the hardware resources and running task requirements of the operating system kernel is improved, thereby improving the speed and accuracy of the kernel configuration. Further, since the parameter configuration value is dynamically generated according to the hardware configuration information of the computing node and the running characteristics of the computing node in the running task state, it can better adapt to the changes in hardware resources and running task requirements, and this adaptability ensures that the computing node can maintain efficient operation under different loads and task requirements. In addition, applying the updated operating system kernel can better utilize the hardware resources of the computing nodes, thereby improving computing efficiency while reducing the risk of computing node crashes or performance degradation due to insufficient hardware resources.

[0117] Based on the model training method provided in the above embodiment, the present application also provides a model training device accordingly. The model training device provided in the embodiment of the present application is specifically introduced below:

[0118] See also Figure 7, which is a schematic diagram of a model training device provided in an embodiment of the present application. The model training device 700 includes: an acquisition module 701 and a training module 702.

[0119] The acquisition module 701 is used to acquire a sample data set, which includes hardware configuration information of multiple computing nodes in a historical time period, operation characteristics of multiple computing nodes in a running task state in a historical time period, and parameter configuration values ​​of operating system kernels of multiple computing nodes in a historical time period;

[0120] The training module 702 is used to train the machine learning model according to the sample data set to obtain the kernel parameter configuration model.

[0121] In some specific implementations, the acquisition module 701 is specifically used to: obtain the hardware configuration information of multiple computing nodes in a historical time period, and the operating characteristics of multiple computing nodes in a running task state in a historical time period; cluster the multiple computing nodes according to the hardware configuration information of the multiple computing nodes in the historical time period and the operating characteristics of the multiple computing nodes in the running task state in the historical time period to obtain multiple grouping results, wherein the similarity of the hardware configuration information of the computing nodes in each grouping result is higher than a first similarity threshold, and the similarity of the operating characteristics of the computing nodes in each grouping result is higher than a second similarity threshold; for each grouping result, determine the parameter item to be updated and determine the parameter configuration value for the parameter item to be updated.

[0122] In summary, an embodiment of the present application provides a model training device, which trains a machine learning model by using the hardware configuration information of multiple computing nodes in a historical time period, the operating characteristics of multiple computing nodes in a running task state in a historical time period, and the parameter configuration values ​​of the operating system kernel of multiple computing nodes in a historical time period as a sample data set. The device can ensure that the parameter configuration values ​​output in the actual application of the trained kernel parameter configuration model match the hardware resources of the computing node and are parameter configuration values ​​for specific running tasks, thereby reducing the workload caused by manual configuration and manual optimization while improving the adaptability of the operating system kernel to hardware resources and running task requirements, thereby improving the configuration rate and accuracy of the operating system kernel.

[0123] An embodiment of the present application also provides an electronic device, which includes: a memory for storing computer programs or computer instructions; a processor for executing the computer programs or computer instructions stored in the memory, so that the electronic device executes the configuration method or model training method of the operating system kernel in any of the above embodiments.

[0124] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program or instruction is stored. When the computer program or instruction is executed, the configuration method of the operating system kernel or the model training method in any of the above embodiments is implemented.

[0125] An embodiment of the present application also provides a computer program product. When the computer program product runs on a server, the server implements the configuration method or model training method of the operating system kernel in any of the above embodiments.

[0126] It should be understood that in the embodiments of the present application, "at least one (item)" refers to one or more, and "plurality" refers to two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can represent: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0127] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0128] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0129] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for configuring an operating system kernel, characterized in that: The method comprises: Obtaining hardware configuration information of a computing node and operating characteristics of the computing node when the computing node is in an operating task state; Inputting the hardware configuration information and the operation characteristics into a kernel parameter configuration model to obtain parameter configuration values ​​of the operating system kernel of the computing node; Generating an operating system kernel of the computing node according to the parameter configuration value; The generated operating system kernel is deployed to the computing node.

2. The method according to claim 1, characterized in that The kernel parameter configuration model includes a parameter update sub-model, and the inputting of the hardware configuration information and the operation characteristics into the kernel parameter configuration model to obtain the parameter configuration value of the operating system kernel of the computing node includes: By inputting the hardware configuration information and the operation characteristics into the parameter update sub-model of the kernel parameter configuration model, the parameter items to be updated of the operating system kernel of the computing node are determined, and the parameter configuration values ​​are determined for the parameter items to be updated.

3. The method according to claim 2, characterized in that The kernel parameter configuration model includes a clustering sub-model and a plurality of parameter update sub-models, and the step of inputting the hardware configuration information and the operation characteristics into the update sub-model of the kernel parameter configuration model to determine the parameter items to be updated of the operating system kernel of the computing node includes: Determine a grouping result corresponding to the hardware configuration information and the operating characteristics by inputting the hardware configuration information and the operating characteristics into a clustering sub-model of the kernel parameter configuration model; The parameter items to be updated of the operating system kernel of the computing node are determined by inputting the hardware configuration information and the operation characteristics into a parameter update sub-model corresponding to the grouping result.

4. The method according to claim 1, characterized in that: The method further comprises: If the operating characteristics of the computing node in the running task state after the operating system kernel is deployed are better than the operating characteristics of the computing node in the running task state before the operating system kernel is deployed, the model parameters of the kernel parameter configuration model are updated according to the hardware configuration information, operating characteristics and parameter configuration values ​​corresponding to the operating system kernel.

5. The method according to claim 1, characterized in that Generating the operating system kernel of the computing node according to the parameter configuration value includes: Determining the unupdated parameter items of the operating system kernel according to the original parameter configuration values ​​of the operating system kernel and the parameter configuration values ​​output by the kernel parameter configuration model; The operating system kernel of the computing node is generated according to the original parameter configuration value of the unupdated parameter item and the parameter configuration value output by the kernel parameter configuration model.

6. The method according to claim 1, characterized in that The step of deploying the generated operating system kernel to the computing node includes: When the running task is not running, or when the running characteristic indicates that the real-time load of the running task is lower than a preset load threshold, the generated operating system kernel is deployed to the computing node.

7. The method according to claim 1, characterized in that The obtaining of the hardware configuration information of the computing node and the operation characteristics of the computing node in the running task state includes: Obtaining hardware configuration information of multiple computing nodes in a computing cluster, and operating characteristics of the multiple computing nodes in a running task state; The step of inputting the hardware configuration information and the operation characteristics into a kernel parameter configuration model to obtain parameter configuration values ​​of the operating system kernel of the computing node includes: Determine the grouping results respectively corresponding to the plurality of computing nodes by inputting the hardware configuration information and the operation characteristics of the plurality of computing nodes into the clustering sub-model of the kernel parameter configuration model; Determine the parameter items to be updated of the operating system kernels of the multiple computing nodes by respectively inputting the hardware configuration information and the operation characteristics of the multiple computing nodes into the parameter update sub-models corresponding to the grouping results respectively corresponding to the multiple computing nodes, and determine the parameter configuration values ​​for the parameter items to be updated; Generating the operating system kernel of the computing node according to the parameter configuration value includes: Generating corresponding operating system kernels for the multiple computing nodes respectively according to the parameter configuration values ​​of the multiple computing nodes; Deploying the generated operating system kernel to the computing node includes: The operating system kernels generated for the multiple computing nodes are deployed to the corresponding multiple computing nodes respectively.

8. A model training method, characterized in that: The method comprises: Acquire a sample data set, the sample data set including hardware configuration information of multiple computing nodes in a historical time period, operation characteristics of the multiple computing nodes in a running task state in the historical time period, and parameter configuration values ​​of operating system kernels of the multiple computing nodes in the historical time period; A machine learning model is trained according to the sample data set to obtain a kernel parameter configuration model.

9. The method according to claim 8, characterized in that The obtaining of the sample data set comprises: Acquire hardware configuration information of multiple computing nodes in a historical time period, and operation characteristics of the multiple computing nodes in a running task state in the historical time period; Clustering the multiple computing nodes according to the hardware configuration information of the multiple computing nodes in the historical time period and the operating characteristics of the multiple computing nodes in the running task state in the historical time period to obtain multiple grouping results, wherein the similarity of the hardware configuration information of the computing nodes in each of the grouping results is higher than a first similarity threshold, and the similarity of the operating characteristics of the computing nodes in each of the grouping results is higher than a second similarity threshold; For each of the grouping results, a parameter item to be updated is determined, and a parameter configuration value is determined for the parameter item to be updated.

10. An electronic device, characterized in that: The electronic device comprises: Memory for storing computer programs or computer instructions; A processor, configured to execute a computer program or computer instruction stored in the memory, so that the electronic device executes the method as claimed in any one of claims 1 to 7, or claims 8 to 9.

Images