Intelligent operation and maintenance method and system for managing different types of computing resources

Abstract

The invention discloses an intelligent operation and maintenance method and system for managing different types of computing resources, and the method comprises the steps: building a resource prediction model through collecting the information of the computing resources in a platform, predicting a resource demand condition, setting a task priority evaluation mechanism, and carrying out the calculation. The task priority evaluation mechanism performs task priority evaluation according to the predicted resource demand condition, task priority processing adopts different processing strategies according to an evaluation result, the task priority evaluation uses latest resource utilization rate data and a prediction model, the priority of the task is evaluated again, and the task priority evaluation result is obtained. The intelligent scheduling dynamically adjusts the execution sequence of the tasks according to the task priority and the resource utilization rate data, the dynamic matching strategy dynamically adjusts the resource allocation strategy according to the resource utilization rate data and the prediction model, and dynamic adjustment and allocation of different computing resources can be efficiently and accurately achieved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an intelligent operation and maintenance method and system for managing different types of computing resources. Background Art

[0002] With the rapid development of cloud computing and big data technologies, the management of computing resources has become increasingly complex. Traditional operation and maintenance methods are unable to cope with large-scale and diversified computing resource demands, resulting in problems such as low resource utilization and long task response time.

[0003] The Chinese patent publication number: CN 118409847 A discloses a multi-task parallel processing system for digital business hall equipment terminals. The invention aims to achieve efficient task execution and high-quality user experience through task scheduling, resource management, security protection, user experience and intelligent operation and maintenance modules.

[0004] Its core technical solution is to intelligently allocate user requests through the task scheduling module, optimize resource utilization with the resource management module, ensure system security through the security protection module, provide a smooth interface and personalized recommendations through the user experience module to enhance user satisfaction, and monitor the system status in real time through the intelligent operation and maintenance module to predict and repair faults. The above content only improves the processing speed and response efficiency of the digital business hall equipment terminals, and cannot solve the dynamic adjustment and allocation between different computing resources.

[0005] The present invention proposes a solution to the above-mentioned shortcomings: establish a resource prediction model, predict resource demand conditions, and set a task priority evaluation mechanism, wherein the task priority evaluation mechanism performs task priority evaluation according to the predicted resource demand conditions, and task priority processing adopts different processing strategies according to the evaluation results. Task priority evaluation uses the latest resource utilization data and prediction model to re-evaluate the priority of the task, and intelligent scheduling dynamically adjusts the execution order of tasks according to the task priority and resource utilization data. The dynamic matching strategy dynamically adjusts the resource allocation strategy according to the resource utilization data and prediction model, thereby efficiently and accurately realizing the dynamic adjustment and allocation between different computing resources. Summary of the invention

[0006] In order to solve the above-mentioned technical problems, the present invention provides an intelligent operation and maintenance method and system for managing different types of computing resources.

[0007] The technical solution of the present invention is implemented as follows: an intelligent operation and maintenance method for managing different types of computing resources, comprising:

[0008] S1. Collect information about computing resources in the platform, including task type, resource utilization, and task response time. Based on the acquired computing resource information, use a time series prediction algorithm and establish a resource prediction model based on historical data.

[0009] S2. The resource prediction model performs feature engineering processing on the collected data, including task type, resource utilization, and task response time;

[0010] S3. According to the relationship between learning features and resource requirements, the resource prediction model is trained to obtain a fitting model, the prediction accuracy of the model is evaluated, and the autocorrelation of the residual is tested to test the validity of the fitting model;

[0011] S4. Setting up a task priority evaluation mechanism. The task priority evaluation mechanism evaluates the predicted resource demand and classifies the task levels into high priority, medium priority and low priority according to the evaluation results, and adopts different processing strategies;

[0012] S5. Perform intelligent scheduling and adjust dynamic matching strategies based on task priority and resource utilization data.

[0013] Furthermore, in the step S1, information on computing resources in the platform is collected to establish a resource prediction model:

[0014] Collect information on computing resources in the platform, including task type, resource utilization, and task response time. Task type is to collect information on different types of tasks, including: computing-intensive tasks, memory-intensive tasks, and I / O-intensive tasks; resource utilization is to collect utilization data of computing resources, including: CPU utilization, memory utilization, and disk I / O utilization; task response time is to collect task execution time, including: task start time, running time, and completion time. Establishing a resource prediction model is to use the ARIMA algorithm to establish a resource prediction model.

[0015] Furthermore, in step S2, feature engineering is performed on the collected data:

[0016] Perform outlier processing on the collected data to detect and process outliers in time series data;

[0017] Furthermore, outliers are handled as follows:

[0018] Perform outlier processing on the collected data, clean the collected data, including calculating task type data, resource utilization data and task response time data, use the z-score function to calculate task type data, resource utilization data and task response time data, set the outlier threshold comparison, when the output value is greater than or less than the outlier threshold, the output value is an outlier, and take processing methods to handle the outlier.

[0019] Furthermore, in the step S3, the resource prediction model is trained according to the relationship between the learning features and the resource requirements to obtain a fitting model, the prediction accuracy of the model is evaluated, and the autocorrelation of the residual is tested to test the validity of the fitting model:

[0020] According to the ACF and PACF diagrams, the parameters p, d, and q of the ARIMA model are selected, where P is the number of autoregressive terms, d refers to the difference stage, and q is the number of moving average terms. The resource forecasting model is trained by estimating the parameters of the ARIMA model using the maximum likelihood estimation method. The prediction accuracy of the trained ARIMA model is evaluated using the cross-validation method, and the autocorrelation of the residuals is tested using the Ljung-Box test method.

[0021] Furthermore, the task priority evaluation mechanism in step S4 performs evaluation according to the predicted resource demand situation, and according to the evaluation result, the task level is divided into high priority, medium priority and low priority, and different processing strategies are adopted:

[0022] Preset task priority evaluation mechanism:

[0023] Priority = (current resource utilization - resource demand forecast) * time importance weight, set the priority output threshold interval [a, b], and divide it into high priority, medium priority and low priority according to the interval range where the output value falls, and adopt different processing strategies for different priorities;

[0024] Furthermore, the processing strategy is:

[0025] High priority tasks: prioritize resource allocation to ensure timely completion of tasks;

[0026] Medium priority tasks: dynamically adjust the execution order based on resource conditions;

[0027] Low priority tasks: temporarily put on hold until resources are free to execute.

[0028] Furthermore, in step S5, intelligent scheduling is performed and the dynamic matching strategy is adjusted:

[0029] Intelligent Scheduling:

[0030] Task priority evaluation uses the latest resource utilization data and prediction models to re-evaluate the priorities of tasks, and dynamically adjust the execution order of tasks based on the re-evaluated task priorities and resource utilization data;

[0031] Dynamic matching strategy:

[0032] Dynamically adjust resource allocation strategies based on resource utilization data and forecasting models;

[0033] The resource allocation strategy includes:

[0034] Allocate idle resources to high-priority tasks;

[0035] Pre-allocate some resources to upcoming high-priority tasks;

[0036] Dynamically adjust resource allocation ratio based on resource utilization.

[0037] The system solution of the present invention is implemented as follows: an intelligent operation and maintenance system for managing different types of computing resources, comprising:

[0038] A data collection module, connected to the model building module, is used to collect information about computing resources in the platform and transmit the collected information to the model building module;

[0039] The model building module is connected to the data collection module and the model evaluation module, and is used to receive the information collected by the data collection module, establish a resource prediction model based on the collected data using a time series prediction algorithm, perform feature engineering on the collected data, and transmit the identified data to the model evaluation module;

[0040] The model evaluation module is connected to the model construction module and the task priority evaluation module, and is used to receive the data recognized by the model construction module, train the resource prediction model based on the relationship between the learning features and the resource requirements, evaluate the prediction accuracy of the model, and test the autocorrelation of the residual to test the validity of the fitting model, and transmit the results to the task priority evaluation module;

[0041] The task priority evaluation module is connected with the model evaluation module and the dynamic adjustment module, and is used to receive the resource demand predicted by the model evaluation module, evaluate its task priority, calculate the task priority, and assign different processing strategies according to the priority;

[0042] The dynamic adjustment module is connected to the model evaluation module and the task priority evaluation module, and is used to receive the evaluation information of the model evaluation module and the task priority evaluation module, re-evaluate the priority of the task according to the evaluation information, and dynamically adjust the execution order of the task and the resource allocation strategy according to the task priority, resource utilization data and the prediction model.

[0043] Beneficial effects:

[0044] The present invention aims at solving the shortcoming of the prior art that dynamic adjustment and allocation between different computing resources cannot be solved. By collecting information of computing resources in a platform, a resource prediction model is established, resource demand conditions are predicted, and a task priority evaluation mechanism is set. The task priority evaluation mechanism performs task priority evaluation according to the predicted resource demand conditions. Task priority processing adopts different processing strategies according to the evaluation results. Task priority evaluation uses the latest resource utilization data and prediction models to re-evaluate the priority of tasks. Intelligent scheduling dynamically adjusts the execution order of tasks according to task priorities and resource utilization data. Dynamic matching strategies dynamically adjust resource allocation strategies according to resource utilization data and prediction models, thereby efficiently and accurately realizing dynamic adjustment and allocation between different computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A structural block diagram of an intelligent operation and maintenance method for managing different types of computing resources in an embodiment of the present invention;

[0046] Figure 2 A flowchart of an intelligent operation and maintenance method for managing different types of computing resources in an embodiment of the present invention;

[0047] Figure 3 This is a structural block diagram of an intelligent operation and maintenance system for managing different types of computing resources in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] The preferred implementation methods of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these implementation methods are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0050] It should be noted that when an element is considered to be "connected" to another element, it can be directly connected to the other element, or connected to the other element through an intermediate element. In addition, the "connection" in the following embodiments should be understood as "electrical connection", "communication connection", etc. if there is transmission of electrical signals or data between the connected objects.

[0051] When used herein, the singular forms "a", "an", and "said / the" may also include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "include / comprise" or "have" and the like specify the presence of stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the relevant listed items.

[0052] See also Figure 1-Figure 3 As shown, an intelligent operation and maintenance method for managing different types of computing resources includes:

[0053] S1. Collect information about computing resources in the platform, including task type, resource utilization, and task response time. Based on the acquired computing resource information, use a time series prediction algorithm and establish a resource prediction model based on historical data.

[0054] S2. The resource prediction model performs feature engineering processing on the collected data, including task type, resource utilization, and task response time;

[0055] S3. According to the relationship between learning features and resource requirements, the resource prediction model is trained to obtain a fitting model, the prediction accuracy of the model is evaluated, and the autocorrelation of the residual is tested to test the validity of the fitting model;

[0056] S4. Setting up a task priority evaluation mechanism. The task priority evaluation mechanism evaluates the predicted resource demand and classifies the task levels into high priority, medium priority and low priority according to the evaluation results, and adopts different processing strategies;

[0057] S5. Perform intelligent scheduling and adjust dynamic matching strategies based on task priority and resource utilization data.

[0058] like Figure 2 As shown, in step S1, information about computing resources in the platform is collected, including task type, resource utilization, and task response time. Based on the acquired information about computing resources, a time series prediction algorithm is used to establish a resource prediction model based on historical data.

[0059] Specifically, the task type, resource utilization, and task response time are: Task type is to collect information on different types of tasks, including: computing-intensive tasks, memory-intensive tasks, and I / O-intensive tasks; Resource utilization is to collect utilization data of computing resources, including: CPU utilization, memory utilization, and disk I / O utilization; Task response time is to collect the execution time of tasks, including: task start time, running time, and completion time; Establishing a resource prediction model is to use the ARIMA algorithm to establish a resource prediction model.

[0060] like Figure 2 As shown, in step S2, the resource prediction model performs feature engineering processing on the collected data, including task type, resource utilization, and task response time.

[0061] Specifically, outlier processing is performed on the collected data to detect and process outliers in time series data;

[0062] The collected data is processed for outliers and cleaned, including calculating task type data, resource utilization data, and task reaction time data. The z-score function is used to calculate the task type data, resource utilization data, and task reaction time data. The calculation formula is:

[0063]

[0064] Among them, z1 is the standardized value of the task type data, x1 is the original value of the task type data, μ1 is the mean value of the task type data, and σ1 is the standard deviation of the task type data;

[0065]

[0066] Among them, z2 is the standardized value of resource utilization data, x2 is the original value of resource utilization data, μ2 is the average value of resource utilization data, and σ2 is the standard deviation of resource utilization data;

[0067]

[0068] Among them, z3 is the standardized value of the task reaction time data, x3 is the original value of the task reaction time data, μ3 is the mean value of the task reaction time data, and σ3 is the standard deviation of the task reaction time data;

[0069] Define |Z|>3 as an outlier, that is, when Z is greater than 3 or Z is less than -3, the output value is an outlier. The outlier is processed, including deleting or replacing the outlier;

[0070] After processing, the time series sets of task type, resource utilization, and task response time are obtained as ai 、b i 、c i .

[0071] like Figure 2 As shown, in step S3, according to the relationship between learning features and resource requirements, the resource prediction model is trained to obtain a fitting model, the prediction accuracy of the model is evaluated, and the autocorrelation of the residual is tested to test the validity of the fitting model.

[0072] Specifically, according to the ACF and PACF graphs, the parameters p, d, and q of the ARIMA model are selected, where P is the number of autoregressive terms, d refers to the difference stage, and q is the number of moving average terms, which are respectively used for the time series a of task type, resource utilization, and task reaction time. i 、b i 、c i Draw a picture to see if it is a stationary time series. Perform a d-order difference operation on the non-stationary time series to transform it into a stationary time series set A. i , B i , C i , obtain the ACF graph and PACF graph of the autocorrelation coefficient of the stationary time series respectively. By observing the graph, if the ACF decays exponentially and the PACF is truncated after the pth order, then select p as the number of autoregressive terms. If the PACF decays exponentially and the ACF is truncated after the qth order, then select q as the number of moving average terms. After determining the model order, use the maximum likelihood estimation method to estimate the parameters of the ARIMA model for the parameters p, d, q, select the optimal parameters p, d, q, and use the ARMA (p, q) model for the stationary time series set A i , B i , C i After fitting, the fitted ARIMA models are:

[0073] Task type ARIMA model:

[0074]

[0075] Among them, A t is the timing prediction result of the task type, μ is a constant, α i is the coefficient of the autoregressive model AR, β i is the coefficient of the moving average model MA, ε t is the model residual at time t;

[0076] Resource Utilization ARIMA Model:

[0077]

[0078] Among them, B tis the time series prediction result of resource utilization, μ is a constant, α i is the coefficient of the autoregressive model AR, β i is the coefficient of the moving average model MA, ε i is the model residual at time t;

[0079] Task reaction time ARIMA model:

[0080]

[0081] Among them, C t is the time series prediction result of resource utilization, μ is a constant, α i is the coefficient of the autoregressive model AR, β i is the coefficient of the moving average model MA, ε i is the model residual at time t;

[0082] The prediction accuracy of the model is evaluated using the cross-validation method. The settings of the cross-validation technique are:

[0083] The prediction accuracy of the model is evaluated using the K-fold cross-validation method, with K = 10 as the default. After cleaning the data, a time series set of task type, resource utilization, and task response time is obtained. i 、b i 、c i Divide into K equal-sized subsets, repeat K times, each time use one subset as the test set, and the remaining K-1 subsets as the training set. In each iteration, use the training set to train the logistic regression model, record the number of correct samples, calculate the accuracy of each round: (the number of samples predicted correctly) / (the total number of samples), and then use the calculated accuracy of each round to find the average accuracy: (the sum of the accuracy of the samples predicted correctly in each round) / 10;

[0084] The Ljung-Box test method is used to test the autocorrelation of the residuals. The Ljung-Box test method is set as:

[0085] Calculate the residual sequence of the trained ARIMA model:

[0086]

[0087] Among them, e t is the residual sequence, y t Input data to the model, Predict results for the model;

[0088] Using the Ljung-Box test Q statistic:

[0089]

[0090] Among them, Q is the Ljung-Box statistic, which follows a chi-square distribution with degrees of freedom h-q n is the sample size of the residual sequence, h is the lag order of the residual sequence, is the autocorrelation coefficient of the k-th lag of the residual sequence;

[0091] According to the statistic formula, the autocorrelation of the residuals is tested. When the statistic Q > 0.05, there is no autocorrelation in the residuals. When the statistic Q < 0.05 and Q = 0.05, there is autocorrelation in the residuals, and the model parameters need to be readjusted.

[0092] As Figure 2 shown, in step S4, a task priority evaluation mechanism is set up. The task priority evaluation mechanism evaluates according to the predicted resource demand situation. According to the evaluation results, the task levels are divided into high priority, medium priority, and low priority, and different processing strategies are adopted.

[0093] Specifically, a preset task priority evaluation mechanism:

[0094] Priority = (Current resource utilization rate - Predicted value of resource demand) * Time importance weight. Set a priority output threshold interval [a, b], and divide it into high priority, medium priority, and low priority according to the interval range where the output value falls;

[0095] Preset the time importance weight to 1, calculate the priority, compare the output priority data with the priority output threshold interval [a, b]. When the output priority > b, it is a high-priority task. When the output priority < a, it is a medium-priority task. When the output priority is in the interval of a and b and equal to a or b value, it is a low-priority task. Different processing strategies are adopted for different priorities. The processing strategies are:

[0096] High-priority tasks: Allocate resources first to ensure the timely completion of tasks;

[0097] Medium-priority tasks: Dynamically adjust the execution order according to the resource situation;

[0098] Low-priority tasks: Temporarily shelve and wait until resources are idle before execution.

[0099] As Figure 2 shown, in step S5, according to the task priority and resource utilization rate data, intelligent scheduling is performed, and the dynamic matching strategy is adjusted.

[0100] Specifically, the intelligent scheduling process is:

[0101] The task priority evaluation uses the latest resource utilization rate data and prediction model to re-evaluate the priority of the task. According to the re-evaluated priority of the task and the resource utilization rate data, the execution order of the task is dynamically adjusted;

[0102] The dynamic matching strategy is:

[0103] Dynamically adjust resource allocation strategies based on resource utilization data and forecasting models;

[0104] The resource allocation strategy includes:

[0105] Allocate idle resources to high-priority tasks;

[0106] Pre-allocate some resources to upcoming high-priority tasks;

[0107] Dynamically adjust resource allocation ratio based on resource utilization.

[0108] like Figure 1-Figure 2 As shown, the present invention works as follows in the application scenario of solving the disadvantage that the prior art cannot solve the dynamic adjustment allocation between different computing resources:

[0109] Collect information about computing resources in the platform, including task type, resource utilization, and task response time. Based on the acquired computing resource information, use a time series prediction algorithm to build a resource prediction model based on historical data.

[0110] The resource prediction model performs feature engineering on the collected data, including task type, resource utilization, and task response time;

[0111] According to the relationship between learning features and resource requirements, the resource prediction model is trained to obtain the fitting model, the prediction accuracy of the model is evaluated, and the autocorrelation of the residual is tested to test the validity of the fitting model;

[0112] Set up a task priority assessment mechanism. The task priority assessment mechanism conducts assessments based on the predicted resource demand. Based on the assessment results, the task levels are divided into high priority, medium priority and low priority, and different processing strategies are adopted;

[0113] Intelligent scheduling is performed and dynamic matching strategies are adjusted based on task priority and resource utilization data.

[0114] See also Figure 3 As shown, an intelligent operation and maintenance system for managing different types of computing resources includes:

[0115] A data collection module, connected to the model building module, is used to collect information about computing resources in the platform and transmit the collected information to the model building module;

[0116] The model building module is connected to the data collection module and the model evaluation module, and is used to receive the information collected by the data collection module, establish a resource prediction model based on the collected data using a time series prediction algorithm, perform feature engineering on the collected data, and transmit the identified data to the model evaluation module;

[0117] The model evaluation module is connected to the model construction module and the task priority evaluation module, and is used to receive the data recognized by the model construction module, train the resource prediction model based on the relationship between the learning features and the resource requirements, evaluate the prediction accuracy of the model, and test the autocorrelation of the residual to test the validity of the fitting model, and transmit the results to the task priority evaluation module;

[0118] The task priority evaluation module is connected with the model evaluation module and the dynamic adjustment module, and is used to receive the resource demand predicted by the model evaluation module, evaluate its task priority, calculate the task priority, and assign different processing strategies according to the priority;

[0119] The dynamic adjustment module is connected to the model evaluation module and the task priority evaluation module, and is used to receive the evaluation information of the model evaluation module and the task priority evaluation module, re-evaluate the priority of the task according to the evaluation information, and dynamically adjust the execution order of the task and the resource allocation strategy according to the task priority, resource utilization data and the prediction model.

[0120] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0121] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and rules of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent operation and maintenance method for managing different types of computing resources, characterized by: S1. Collect information about computing resources in the platform, including task type, resource utilization, and task response time. Based on the acquired computing resource information, use a time series prediction algorithm and establish a resource prediction model based on historical data. S2. The resource prediction model performs feature engineering processing on the collected data, including task type, resource utilization, and task response time; S3. According to the relationship between learning features and resource requirements, the resource prediction model is trained to obtain a fitting model, the prediction accuracy of the model is evaluated, and the autocorrelation of the residual is tested to test the validity of the fitting model; S4. Setting up a task priority evaluation mechanism. The task priority evaluation mechanism evaluates the predicted resource demand and classifies the task levels into high priority, medium priority and low priority according to the evaluation results, and adopts different processing strategies; S5. Perform intelligent scheduling and adjust dynamic matching strategies based on task priority and resource utilization data.

2. The intelligent operation and maintenance method for managing different types of computing resources according to claim 1, characterized in that: In the step S1, information about computing resources in the platform is collected, including task type, resource utilization, and task response time. Based on the acquired computing resource information, a time series prediction algorithm is used to establish a resource prediction model based on historical data: Collect information on computing resources in the platform, including task type, resource utilization, and task response time. Task type is to collect information on different types of tasks, including: computing-intensive tasks, memory-intensive tasks, and I / O-intensive tasks; resource utilization is to collect utilization data of computing resources, including: CPU utilization, memory utilization, and disk I / O utilization; task response time is to collect task execution time, including: task start time, running time, and completion time. Establishing a resource prediction model is to use the ARIMA algorithm to establish a resource prediction model.

3. The intelligent operation and maintenance method for managing different types of computing resources according to claim 1, characterized in that: In step S2, the feature engineering process of the collected data is as follows: Perform outlier processing on the collected data to detect and process outliers in time series data; The outlier processing method is: Perform outlier processing on the collected data, clean the collected data, including calculating task type data, resource utilization data and task response time data, use the z-score function to calculate task type data, resource utilization data and task response time data, set the outlier threshold interval comparison, when the output value is greater than or less than the outlier threshold interval, the output value is an outlier, and take processing methods to handle the outlier.

4. The intelligent operation and maintenance method for managing different types of computing resources according to claim 1, characterized in that: In the step S3, the resource prediction model is trained according to the relationship between the learning features and the resource requirements to obtain the fitting model, the prediction accuracy of the model is evaluated, and the autocorrelation of the residual is tested to test the validity of the fitting model: According to the ACF and PACF diagrams, the parameters p, d, and q of the ARIMA model are selected, where P is the number of autoregressive terms, d refers to the difference stage, and q is the number of moving average terms. The resource forecasting model is trained by estimating the parameters of the ARIMA model using the maximum likelihood estimation method. The prediction accuracy of the trained ARIMA model is evaluated using the cross-validation method, and the autocorrelation of the residuals is tested using the Ljung-Box test method.

5. The intelligent operation and maintenance method for managing different types of computing resources according to claim 1, characterized in that: The task priority evaluation mechanism in step S4 performs evaluation according to the predicted resource demand situation. According to the evaluation result, the task level is divided into high priority, medium priority and low priority, and different processing strategies are adopted: Preset task priority evaluation mechanism: Priority = (current resource utilization - resource demand forecast) * time importance weight, set the priority output threshold interval [a, b], and divide it into high priority, medium priority and low priority according to the interval range where the output value falls, and adopt different processing strategies for different priorities; The processing strategy is: High priority tasks: prioritize resource allocation to ensure timely completion of tasks; Medium priority tasks: dynamically adjust the execution order based on resource conditions; Low priority tasks: temporarily put on hold until resources are free to execute.

6. The intelligent operation and maintenance method for managing different types of computing resources according to claim 1, characterized in that: In step S5, intelligent scheduling is performed and the dynamic matching strategy is adjusted as follows: Intelligent Scheduling: Task priority evaluation uses the latest resource utilization data and prediction models to re-evaluate the priorities of tasks, and dynamically adjust the execution order of tasks based on the re-evaluated task priorities and resource utilization data; Dynamic matching strategy: Dynamically adjust resource allocation strategies based on resource utilization data and forecasting models; The resource allocation strategy is: Allocate idle resources to high-priority tasks; Pre-allocate some resources to upcoming high-priority tasks; Dynamically adjust resource allocation ratio based on resource utilization.

7. An intelligent operation and maintenance system for managing different types of computing resources, characterized by: A data collection module, connected to the model building module, is used to collect information about computing resources in the platform and transmit the collected information to the model building module; The model building module is connected to the data collection module and the model evaluation module, and is used to receive the information collected by the data collection module, establish a resource prediction model based on the collected data using a time series prediction algorithm, perform feature engineering on the collected data, and transmit the identified data to the model evaluation module; The model evaluation module is connected to the model construction module and the task priority evaluation module, and is used to receive the data recognized by the model construction module, train the resource prediction model based on the relationship between the learning features and the resource requirements, evaluate the prediction accuracy of the model, and test the autocorrelation of the residual to test the validity of the fitting model, and transmit the results to the task priority evaluation module; The task priority evaluation module is connected with the model evaluation module and the dynamic adjustment module, and is used to receive the resource demand predicted by the model evaluation module, evaluate its task priority, calculate the task priority, and assign different processing strategies according to the priority; The dynamic adjustment module is connected to the model evaluation module and the task priority evaluation module, and is used to receive the evaluation information of the model evaluation module and the task priority evaluation module, re-evaluate the priority of the task according to the evaluation information, and dynamically adjust the execution order of the task and the resource allocation strategy according to the task priority, resource utilization data and the prediction model.

Images