Cloud platform capacity prediction method and system based on resource pool characteristics

Abstract

The invention relates to the technical field of cloud computing and big data analysis, in particular to a cloud platform capacity prediction method and system based on resource pool features, and the method comprises the steps of data collection and preprocessing, feature extraction, feature selection, prediction model construction, real-time prediction and adjustment, and model optimization and updating. The method has the beneficial effects that by analyzing historical data of a resource pool, extracting key features, combining algorithms such as time sequence analysis and machine learning, and combing and analyzing related researches, accurate prediction of resource demands of a cloud platform in a period of time in the future is realized. The prediction accuracy is improved, and the prediction error is reduced; the method is suitable for cloud platforms of different types and scales, and has high universality. According to the method, effective decision support can be provided for a cloud platform operator, resource optimization configuration is realized, the service quality of the cloud platform is improved, the operation and maintenance efficiency of the cloud platform is improved, and safe and stable operation of the cloud platform is guaranteed while the operation and maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical fields of cloud computing and big data analysis, and particularly to a cloud platform capacity prediction method and system based on resource pool characteristics. Background Art

[0002] With the popularization and development of cloud computing technology, cloud platforms have become an important way for enterprises and service providers to obtain IT resources. Cloud platforms provide flexible and scalable computing, storage, and network resources to meet the diverse needs of users, and users can dynamically obtain and release resources according to their needs. However, with the popularization of cloud services and the continuous growth of user demands, cloud platform resource management faces many challenges, one of which is how to accurately predict resource demands for effective capacity planning.

[0003] The existing cloud platform capacity prediction methods mainly include the following:

[0004] 1. Prediction based on historical trends: This method predicts future resource demands by analyzing historical resource usage data. Although it is simple and easy to implement, this method often ignores the impact of sudden events and seasonal changes on resource demands, resulting in low prediction accuracy.

[0005] 2. Time series analysis: Time series analysis methods such as ARIMA (Autoregressive Integrated Moving Average Model) are used to predict resource demands. These methods can consider the periodicity and trend of time series, but they perform poorly in dealing with non-linear complex systems and abnormal data.

[0006] 3. Machine learning methods: Some studies attempt to use machine learning algorithms such as Support Vector Machine (SVM), Random Forest, and Neural Network to predict cloud platform capacity. These methods improve the prediction accuracy to a certain extent, but they require a large amount of labeled data and a complex model training process.

[0007] 4. Prediction based on performance metrics: This method predicts resource demands by monitoring the performance metrics of the cloud platform (such as CPU utilization, memory usage, etc.). However, it may not be able to accurately capture the long-term trends and potential changes in resource demands.

[0008] The existing prediction methods have the following problems:

[0009] The prediction accuracy needs to be improved, especially in the face of complex and changing cloud environments.

[0010] Lack of real-time performance and inability to quickly respond to dynamic changes in resource demands.

[0011] The prediction models are usually relatively complex, with high computational costs and are not easy to implement in large-scale cloud platforms. Summary of the Invention

[0012] The object of the present invention is to provide a cloud platform capacity prediction method and system based on resource pool characteristics, so as to solve the problems of low prediction accuracy, poor real-time performance, high model complexity and difficulty in large-scale implementation mentioned in the above background technology.

[0013] To achieve the above object, the present invention provides the following technical solution: A cloud platform capacity prediction method based on resource pool characteristics, including data collection and preprocessing steps, specifically:

[0014] Automatically collect historical usage data of the resource pool from the monitoring system of the cloud platform. The historical usage data includes key metrics such as CPU utilization, memory usage, disk I / O, network traffic, and the number of virtual machines. The time granularity of data collection is set according to actual needs;

[0015] Clean the collected data. By setting reasonable thresholds and using statistical methods, identify and eliminate outliers caused by system failures and data transmission errors;

[0016] Normalize the cleaned data. Adopt common normalization methods such as min-max normalization and Z-score normalization to eliminate the differences in data ranges and dimensions of different metrics.

[0017] Preferably, it further includes a feature extraction step, specifically:

[0018] Perform time feature extraction, analyze the time attributes of the data, and extract features such as date, week, season, holiday, and timestamp to capture the periodic patterns in resource usage;

[0019] Perform statistical feature extraction, calculate the basic statistics of resource usage data, including mean, median, standard deviation, skewness, and kurtosis, to describe the basic situation and distribution characteristics of resource usage;

[0020] Perform trend feature extraction, analyze the long-term trend and short-term fluctuations of resource demand by calculating the growth rate, moving average, and exponential smoothing indicators of resource usage.

[0021] Preferably, it further includes a feature selection step, specifically:

[0022] Perform correlation analysis by calculating the Pearson correlation coefficient or Spearman rank correlation coefficient between features to identify features highly correlated with resource demand;

[0023] Apply the principal component analysis (PCA) method to reduce the dimensionality of the features, extract the principal components that can explain most of the data variation, reduce the number of features, and retain the most important information at the same time;

[0024] Use model-based feature selection methods, such as decision tree-based feature selection, or use the Recursive Feature Elimination (RFE) method to further screen out the features that are most valuable for prediction.

[0025] Preferably, it further includes a prediction model construction step, specifically:

[0026] According to the characteristics of the features and data, select appropriate machine learning algorithms, including Random Forest, Gradient Boosting Machine (GBM), Support Vector Regression (SVR), and Long Short-Term Memory Network (LSTM);

[0027] Use the feature data and the corresponding resource requirements as inputs to train the selected machine learning model. By adjusting the model parameters and using cross-validation techniques, optimize the model performance;

[0028] Considering the short-term and long-term changes in resource requirements, construct multiple models to predict resource requirements at different time scales respectively.

[0029] Preferably, it further includes real-time prediction and adjustment, model optimization and update steps, specifically:

[0030] Real-time prediction and adjustment step: Input the real-time collected resource pool feature data into the trained prediction model to obtain the real-time resource requirement prediction result; According to the prediction result and the actual resource usage situation, dynamically adjust the resource allocation strategy, such as automatically expanding or shrinking virtual machine instances, adjusting the load balancing; Set the warning threshold for resource requirements. When the prediction result exceeds the threshold, the system automatically issues a warning and starts the resource adjustment process;

[0031] Model optimization and update step: Regularly use the latest data to evaluate the performance of the prediction model. The evaluation metrics include prediction accuracy, Mean Squared Error (MSE), and Root Mean Squared Error (RMSE); When it is detected that the model performance deteriorates or the resource requirement pattern changes, use the latest data to retrain the model to maintain the prediction ability of the model; Develop the frequency and conditions for model updates to ensure that the model can always adapt to the latest changes in the resource requirements of the cloud platform.

[0032] A prediction system for a cloud platform capacity prediction method based on resource pool features, including a data collection and preprocessing module, which is used for:

[0033] Automatically collect the historical usage data of the resource pool from the monitoring system of the cloud platform. The historical usage data covers key metrics such as CPU utilization, memory usage, disk I / O, network traffic, and the number of virtual machines, and the time granularity of data collection is set to minutes, hours, or days according to actual needs;

[0034] Clean the collected data. By setting reasonable thresholds and using statistical methods, identify and eliminate outliers caused by system failures, data transmission errors, etc., to ensure data consistency and reliability;

[0035] Normalize the cleaned data. Use common methods such as min-max normalization and Z-score normalization to eliminate differences in data range and dimension among different indicators.

[0036] Preferably, it further includes a feature extraction module for:

[0037] Extract time features, analyze the time attributes of the data, and extract features such as date, week, season, holiday, timestamp to capture periodic patterns in resource usage;

[0038] Conduct statistical feature extraction, calculate basic statistics of resource usage data, including mean, median, standard deviation, skewness, kurtosis, to describe the basic situation and distribution characteristics of resource usage;

[0039] Implement trend feature extraction. By calculating the growth rate, moving average, exponential smoothing index of resource usage, analyze the long-term trend and short-term fluctuations of resource demand.

[0040] Preferably, it further includes a feature selection module for:

[0041] Conduct correlation analysis by calculating the Pearson correlation coefficient or Spearman rank correlation coefficient between features, and identify features highly correlated with resource demand;

[0042] Use the principal component analysis (PCA) method to reduce the dimensionality of features, extract the principal components that can explain most of the data variation, and retain key information while reducing the number of features;

[0043] Adopt model-based feature selection methods, such as feature selection based on decision trees, or use the recursive feature elimination (RFE) method to further screen out the features most valuable for prediction.

[0044] Preferably, it further includes a prediction model construction module for:

[0045] According to the characteristics of features and data, select appropriate machine learning algorithms, such as random forest, gradient boosting machine (GBM), support vector regression (SVR), long short-term memory network (LSTM);

[0046] Take the feature data and the corresponding resource demand as inputs, train the selected machine learning model, and optimize the model performance by adjusting model parameters and using techniques such as cross-validation;

[0047] Considering the short-term and long-term changes in resource requirements, multiple models are constructed to predict resource requirements at different time scales.

[0048] Preferably, it further includes a real-time prediction and adjustment module and a model optimization and update module;

[0049] The real-time prediction and adjustment module is used for: inputting the feature data of the resource pool collected in real time into the trained prediction model to obtain the real-time resource requirement prediction result; dynamically adjusting the resource allocation strategy according to the prediction result and the actual resource usage situation, such as automatically expanding or reducing virtual machine instances, adjusting load balancing, etc.; setting the warning threshold of resource requirements, and when the prediction result exceeds the threshold, the system automatically issues a warning and starts the resource adjustment process;

[0050] The model optimization and update module is used for: regularly evaluating the performance of the prediction model using the latest data, and the evaluation metrics include prediction accuracy, mean squared error MSE, and root mean squared error RMSE; when it is detected that the model performance deteriorates or the resource requirement pattern changes, retraining the model using the latest data to maintain the prediction ability of the model; formulating the frequency and conditions of model update to ensure that the model can always adapt to the latest changes in the resource requirements of the cloud platform.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] The cloud platform capacity prediction method and system based on resource pool features proposed by the present invention analyze the historical data of the resource pool, extract key features, and combine algorithms such as time series analysis and machine learning. By sorting out and analyzing relevant research, it realizes the accurate prediction of resource requirements in the cloud platform for a period of time in the future. The present invention has the following advantages: improving the prediction accuracy and reducing the prediction error; strong real-time performance, capable of quickly responding to changes in resource requirements; applicable to different types and scales of cloud platforms, with high versatility. This method can provide effective decision-making support for cloud platform operators, realize the optimal allocation of resources, improve the service quality of the cloud platform, help improve the operation and maintenance efficiency of the cloud platform, and ensure the safe and stable operation of the cloud platform while reducing the operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is the data collection and preprocessing flowchart of the present invention;

[0054] Figure 2 It is the feature extraction flowchart of the present invention;

[0055] Figure 3 It is the feature selection flowchart of the present invention;

[0056] Figure 4 It is the prediction model construction flowchart of the present invention;

[0057] Figure 5 This is the flowchart for real-time prediction and adjustment of the present invention;

[0058] Figure 6 This is the flowchart for model optimization and update of the present invention. Detailed implementation manners

[0059] In order to clearly and completely describe the objectives, technical solutions of the present invention and make the advantages clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, rather than all of the embodiments, and are only used to explain the embodiments of the present invention, not to limit the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0060] Embodiment 1. Please refer to Figures 1 to 6 , the present invention provides a technical solution: a cloud platform capacity prediction method based on resource pool characteristics, including the following steps:

[0061] 1. Data collection and preprocessing:

[0062] Data collection: The system will automatically collect historical usage data of the resource pool from the monitoring system of the cloud platform, including but not limited to key metrics such as CPU utilization, memory usage, disk I / O, network traffic, and the number of virtual machines. The time granularity of data collection can be set according to actual needs, such as every minute, every hour, or daily.

[0063] Data cleaning: Among the collected data, there may be outliers caused by system failures, data transmission errors, etc. By setting reasonable thresholds and using statistical methods, these outliers are identified and removed to ensure data consistency and reliability.

[0064] Data normalization: Since the data ranges and dimensions of different metrics may vary, in order to facilitate subsequent analysis and modeling, the data needs to be normalized. Common normalization methods include min-max normalization and Z-score normalization, etc.

[0065] 2. Feature extraction:

[0066] Time feature extraction: Analyze the time attributes of the data and extract features such as date, week, season, holiday, timestamp, etc. to capture the periodic patterns in resource usage.

[0067] Statistical feature extraction: Calculate the basic statistics of the resource usage data, such as mean, median, standard deviation, skewness, kurtosis, etc., to describe the basic situation and distribution characteristics of resource usage.

[0068] Trend Feature Extraction: Analyze the long-term trend and short-term fluctuations of resource demand by calculating indicators such as the growth rate of resource usage, moving average, and exponential smoothing.

[0069] 3. Feature Selection:

[0070] Correlation Analysis: Identify features highly correlated with resource demand by calculating the Pearson correlation coefficient or Spearman rank correlation coefficient between features.

[0071] Principal Component Analysis (PCA): Apply the PCA method to reduce the dimensionality of features, extract the principal components that can explain most of the data variation, reduce the number of features, and retain the most important information.

[0072] Feature Selection Algorithm: Use model-based feature selection methods, such as decision tree-based feature selection, or methods like Recursive Feature Elimination (RFE), to further screen out the features most valuable for prediction.

[0073] 4. Prediction Model Construction:

[0074] Machine Learning Algorithm Selection: Select appropriate machine learning algorithms according to the characteristics of features and data, such as Random Forest, Gradient Boosting Machine (GBM), Support Vector Regression (SVR), Long Short-Term Memory Network (LSTM), etc.

[0075] Model Training: Use the feature data and corresponding resource demand as input to train the selected machine learning model. Optimize the model performance by adjusting model parameters and using techniques such as cross-validation.

[0076] Multi-Time Scale Prediction: Considering the short-term and long-term changes in resource demand, multiple models can be constructed to predict resource demand at different time scales (such as hours, days, weeks, months).

[0077] 5. Real-Time Prediction and Adjustment:

[0078] Real-Time Data Input: Input the real-time collected resource pool feature data into the trained prediction model to obtain real-time resource demand prediction results.

[0079] Resource Allocation Strategy Adjustment: Dynamically adjust the resource allocation strategy according to the prediction results and actual resource usage conditions, such as automatically expanding or reducing virtual machine instances, adjusting load balancing, etc.

[0080] Early Warning Mechanism: Set the early warning threshold for resource demand. When the prediction result exceeds the threshold, the system automatically issues an early warning and initiates the resource adjustment process.

[0081] 6. Model Optimization and Update:

[0082] Model performance evaluation: Regularly evaluate the performance of the prediction model using the latest data, including metrics such as prediction accuracy, mean squared error (MSE), root mean squared error (RMSE), etc.

[0083] Model retraining: When a decline in model performance or a change in the resource demand pattern is detected, retrain the model using the latest data to maintain the model's predictive ability.

[0084] Model update strategy: Define the frequency and conditions for model updates to ensure that the model can always adapt to the latest changes in the resource requirements of the cloud platform.

[0085] The relationship model established by deeply mining the characteristics of the resource pool and using machine learning algorithms can effectively predict the capacity changes of the cloud platform, providing strong support for cloud platform management. This method has been applied in the cloud platform environment and achieved good results. In the future, we will continue to optimize and improve the prediction model to further enhance the accuracy and practicality of the prediction.

[0086] Example 2: Based on Example 1, a prediction system for the cloud platform capacity prediction method based on resource pool characteristics is proposed, including a data collection and preprocessing module, which is used for: automatically collecting the historical usage data of the resource pool from the monitoring system of the cloud platform. The historical usage data covers key metrics such as CPU utilization, memory usage, disk I / O, network traffic, and the number of virtual machines, and the time granularity of data collection is set to minutes, hours, or days according to actual needs; cleaning the collected data, by setting reasonable thresholds and using statistical methods, identifying and removing outliers caused by system failures, data transmission errors and other factors to ensure the consistency and reliability of the data; normalizing the cleaned data, using common methods such as min-max standardization and Z-score standardization to eliminate the differences in data range and dimension among different metrics.

[0087] It also includes a feature extraction module, which is used for: performing time feature extraction, analyzing the time attributes of the data, and extracting features such as date, week, season, holiday, and timestamp to capture the periodic patterns in resource usage; carrying out statistical feature extraction, calculating the basic statistics of the resource usage data, including mean, median, standard deviation, skewness, and kurtosis, to describe the basic situation and distribution characteristics of resource usage; implementing trend feature extraction, by calculating the growth rate, moving average, and exponential smoothing indicators of the resource usage volume, analyzing the long-term trends and short-term fluctuations of resource demand.

[0088] It also includes a feature selection module, which is used to: conduct correlation analysis by calculating the Pearson correlation coefficient or Spearman rank correlation coefficient between features to identify features highly correlated with resource requirements; use the principal component analysis (PCA) method to reduce the dimensionality of features, extract the principal components that can explain most of the data variation, and retain key information while reducing the number of features; adopt a model-based feature selection method, such as feature selection based on decision trees, or use the recursive feature elimination (RFE) method to further screen out the features most valuable for prediction.

[0089] It also includes a prediction model construction module, which is used to: select a suitable machine learning algorithm according to the characteristics of features and data, such as random forest, gradient boosting machine (GBM), support vector regression (SVR), long short-term memory network (LSTM); use the feature data and the corresponding resource requirements as inputs to train the selected machine learning model, and optimize the model performance by adjusting model parameters and using techniques such as cross-validation; considering the short-term and long-term changes in resource requirements, construct multiple models to predict resource requirements at different time scales.

[0090] It also includes a real-time prediction and adjustment module and a model optimization and update module; the real-time prediction and adjustment module is used to: input the real-time collected resource pool feature data into the trained prediction model to obtain the real-time resource requirement prediction result; dynamically adjust the resource allocation strategy according to the prediction result and the actual resource usage situation, such as automatically expanding or reducing virtual machine instances, adjusting the load balance, etc.; set the warning threshold for resource requirements, and when the prediction result exceeds the threshold, the system automatically issues a warning and starts the resource adjustment process; the model optimization and update module is used to: regularly evaluate the performance of the prediction model using the latest data, and the evaluation metrics include prediction accuracy, mean squared error (MSE), root mean squared error (RMSE); when it is detected that the model performance deteriorates or the resource requirement pattern changes, retrain the model using the latest data to maintain the prediction ability of the model; formulate the frequency and conditions for model update to ensure that the model can always adapt to the latest changes in the resource requirements of the cloud platform.

[0091] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A cloud platform capacity prediction method based on resource pool characteristics, characterized in that: It includes data collection and preprocessing steps, specifically: Automatically collect historical usage data of the resource pool from the monitoring system of the cloud platform. The historical usage data includes key metrics such as CPU utilization, memory usage, disk I / O, network traffic, and the number of virtual machines. The time granularity of data collection is set according to actual requirements; Clean the collected data. By setting reasonable thresholds and using statistical methods, identify and remove outliers caused by system failures, data transmission errors, etc.; Perform normalization on the cleaned data, using common normalization methods such as min-max normalization and Z-score normalization to eliminate the differences in data ranges and dimensions of different metrics.

2. The cloud platform capacity prediction method based on resource pool characteristics according to claim 1, wherein: It also includes a feature extraction step, specifically: Perform time feature extraction, analyze the time attributes of the data, and extract features such as date, week, season, holiday, and timestamp to capture the periodic patterns in resource usage; Perform statistical feature extraction, calculate the basic statistics of resource usage data, including mean, median, standard deviation, skewness, and kurtosis, to describe the basic situation and distribution characteristics of resource usage; Perform trend feature extraction, analyze the long-term trend and short-term fluctuations of resource demand by calculating the growth rate, moving average, and exponential smoothing index of resource usage.

3. The cloud platform capacity prediction method based on resource pool characteristics according to claim 2, characterized in that: It also includes a feature selection step, specifically: Perform correlation analysis by calculating the Pearson correlation coefficient or Spearman rank correlation coefficient between features to identify features highly correlated with resource demand; Apply the principal component analysis (PCA) method to reduce the dimensionality of the features, extract the principal components that can explain most of the data variation, reduce the number of features, and at the same time retain the most important information; Use model-based feature selection methods, such as decision tree-based feature selection, or use the recursive feature elimination (RFE) method to further screen out the features most valuable for prediction.

4. A cloud platform capacity prediction method based on resource pool characteristics according to claim 3, characterized in that: It also includes a prediction model construction step, specifically: According to the characteristics of the features and data, select appropriate machine learning algorithms, including random forest, gradient boosting machine (GBM), support vector regression (SVR), and long short-term memory network (LSTM); Use the feature data and the corresponding resource demand as input to train the selected machine learning model. By adjusting the model parameters and using cross-validation techniques, optimize the model performance; Considering the short-term and long-term changes in resource demand, construct multiple models to predict the resource demand at different time scales respectively.

5. A cloud platform capacity prediction method based on resource pool characteristics according to claim 4, characterized in that: It also includes real-time prediction and adjustment, model optimization and update steps, specifically: Real-time prediction and adjustment step: Input the real-time collected resource pool feature data into the trained prediction model to obtain the real-time resource demand prediction result; According to the prediction result and the actual resource usage situation, dynamically adjust the resource allocation strategy, such as automatically expanding or shrinking virtual machine instances, adjusting the load balance; Set the warning threshold for resource demand. When the prediction result exceeds the threshold, the system automatically issues a warning and initiates the resource adjustment process; Model Optimization and Update Steps: Regularly evaluate the performance of the prediction model using the latest data. The evaluation metrics include prediction accuracy, mean squared error (MSE), and root mean squared error (RMSE). When it is detected that the model performance degrades or the resource demand pattern changes, retrain the model using the latest data to maintain the model's prediction ability. Define the frequency and conditions for model updates to ensure that the model can always adapt to the latest changes in the resource demands of the cloud platform.

6. A prediction system for the cloud platform capacity prediction method based on resource pool characteristics according to claim 5, characterized in that: It includes a data collection and preprocessing module for: Automatically collect historical usage data of the resource pool from the monitoring system of the cloud platform. The historical usage data covers key metrics such as CPU utilization, memory usage, disk I / O, network traffic, and the number of virtual machines. The time granularity of data collection is set to minutes, hours, or daily according to actual requirements. Clean the collected data. By setting reasonable thresholds and using statistical methods, identify and remove outliers caused by system failures and data transmission errors to ensure data consistency and reliability. Normalize the cleaned data using common methods such as min-max normalization and Z-score normalization to eliminate differences in data range and dimension among different metrics.

7. The prediction system according to claim 6, wherein: It also includes a feature extraction module for: Perform time feature extraction, analyze the time attributes of the data, and extract features such as date, week, season, holiday, and timestamp to capture periodic patterns in resource usage. Conduct statistical feature extraction, calculate basic statistics of the resource usage data, including mean, median, standard deviation, skewness, and kurtosis, to describe the basic situation and distribution characteristics of resource usage. Implement trend feature extraction, analyze the long-term trends and short-term fluctuations of resource demands by calculating growth rates, moving averages, and exponential smoothing metrics of resource usage.

8. A prediction system according to claim 7, characterized in that: It also includes a feature selection module for: Conduct correlation analysis by calculating Pearson correlation coefficients or Spearman rank correlation coefficients between features to identify features highly correlated with resource demands. Use the principal component analysis (PCA) method to reduce the dimensionality of the features, extract the principal components that can explain most of the data variation, and retain key information while reducing the number of features. Adopt model-based feature selection methods, such as decision tree-based feature selection or recursive feature elimination (RFE) method, to further screen out the features most valuable for prediction.

9. A prediction system according to claim 8, characterized in that: It also includes a prediction model construction module for: Select appropriate machine learning algorithms according to the characteristics of the features and data, such as random forest, gradient boosting machine (GBM), support vector regression (SVR), and long short-term memory network (LSTM). Use the feature data and corresponding resource demands as inputs to train the selected machine learning model, and optimize the model performance by adjusting model parameters and using techniques such as cross-validation. Considering the short-term and long-term changes in resource demands, construct multiple models to predict resource demands at different time scales.

10. A prediction system according to claim 9, wherein: It also includes a real-time prediction and adjustment module and a model optimization and update module. The real-time prediction and adjustment module is used for: inputting the feature data of the resource pool collected in real time into the trained prediction model to obtain the real-time resource demand prediction result; dynamically adjusting the resource allocation strategy according to the prediction result and the actual resource usage situation, such as automatically expanding or reducing virtual machine instances, adjusting load balancing, etc.; setting the warning threshold for resource demand, and when the prediction result exceeds the threshold, the system automatically issues a warning and starts the resource adjustment process; The model optimization and update module is used for: regularly evaluating the performance of the prediction model using the latest data, and the evaluation metrics include prediction accuracy, mean squared error MSE, and root mean squared error RMSE; when it is detected that the model performance deteriorates or the resource demand pattern changes, retraining the model using the latest data to maintain the prediction ability of the model; formulating the frequency and conditions for model update to ensure that the model can always adapt to the latest changes in the resource demand of the cloud platform.

Images