Cloud platform database storage and use prediction method based on time sequence modeling

Abstract

The invention provides a cloud platform database storage use prediction method based on time sequence modeling, and belongs to the technical field of cloud computing operation and maintenance management. The method comprises the steps that historical storage use data of a target database in a cloud platform is collected, a time sequence prediction model is built based on the historical storage use data, and the time sequence prediction model is a Prophet model supporting trend term and periodic term modeling; using the time sequence prediction model to predict the storage usage amount of the target database in a future preset period, and generating a prediction result value; and comparing the prediction result value with a preset capacity threshold value, and if the prediction result value exceeds the preset capacity threshold value, generating capacity early warning information. By automatically and accurately predicting the use trend of database storage and providing early warning information, the automation and foresight of database capacity management are improved.

Description

Technical Field

[0001] This application belongs to the field of cloud computing operation and maintenance management technology, specifically involving a cloud platform database storage usage prediction method based on time series modeling. Background Technology

[0002] With the widespread application of cloud computing technology, more and more enterprises and organizations are deploying their core businesses on cloud platforms. As a critical data storage component in cloud platforms, the changes in database storage usage directly impact system stability and business continuity. Especially against the backdrop of explosive data growth and rapidly changing business needs, database storage space faces risks such as capacity exhaustion and delayed expansion, posing significant challenges to system operation and maintenance.

[0003] Currently, database capacity management typically relies on manual inspections, fixed threshold alerts, or inferences based on historical experience. These methods often fail to detect trends and potential risks in storage usage in a timely manner, easily leading to delayed capacity warnings or even storage exhaustion incidents. Furthermore, they lack the ability to accurately predict storage usage, which not only reduces resource utilization but may also result in ineffective expansion and increased operational costs. Summary of the Invention

[0004] To address at least one of the technical problems existing in the background art, this application provides a cloud platform database storage usage prediction method based on time series modeling, which improves the automation and foresight of database capacity management by automatically and accurately predicting database storage usage trends and providing early warning information.

[0005] The technical solution adopted in this application is as follows: The first aspect of this application provides a cloud platform database storage usage prediction method based on time-series modeling, including: Collect historical storage usage data of the target database in the cloud platform, and construct a time series prediction model based on the historical storage usage data. The time series prediction model is a Prophet model that supports modeling of trend and periodic items. The time-series prediction model is used to predict the storage usage of the target database within a future preset period, and a prediction result value is generated. The predicted value is compared with a preset capacity threshold. If the predicted value exceeds the preset capacity threshold, a capacity warning message is generated.

[0006] According to the cloud platform database storage usage prediction method based on time-series modeling provided in the first aspect of this application, historical storage usage data of the target database in the cloud platform is first collected, and a Prophet time-series prediction model supporting trend and periodic term modeling is constructed based on this historical data. The Prophet model is chosen because it can effectively handle trend changes, seasonal fluctuations, and holiday effects in time series, giving it a significant advantage in dealing with complex and ever-changing data patterns. The beneficial effect of this step is that, through in-depth analysis and learning of historical data, the long-term trends and periodic changes in database storage usage can be accurately captured, laying a solid foundation for subsequent accurate predictions.

[0007] Next, the existing Prophet model is used to predict storage usage over a pre-defined period, generating forecast values ​​with confidence intervals. This process considers not only historical patterns of database storage usage but also external factors that may influence storage demand (such as holidays). The core value of this step lies in providing a forward-looking view of future storage needs, helping enterprises plan resource allocation in advance and avoid service interruptions or performance degradation due to insufficient storage space. Simultaneously, the introduction of confidence intervals increases the reliability of the forecast results, allowing decision-makers to formulate appropriate strategies based on different risk preferences.

[0008] Finally, the predicted value is compared with a pre-set capacity threshold. If the predicted value exceeds the threshold, a capacity warning is immediately generated. This real-time data analysis-based warning mechanism can promptly notify the operations team to take action before potential problems occur, such as proactively scaling up or optimizing data management strategies. This approach significantly improves the proactiveness and efficiency of database operations, reduces emergency response costs caused by storage exhaustion, and helps maintain high system availability and stability.

[0009] Overall, this approach achieves effective management of cloud platform database storage through intelligent prediction and early warning, thereby improving resource utilization and service quality.

[0010] According to one embodiment of this application, the historical storage usage data of the target database in the data collection cloud platform, and the construction of a time-series prediction model based on the historical storage usage data, specifically involves: The heterogeneous database adaptation layer is used to dynamically collect the historical storage usage data. The heterogeneous database adaptation layer is suitable for unified access to relational databases and non-relational databases, and can interface with different types of database systems in the cloud platform. Missing values ​​are filled and outliers are removed from the collected historical storage and usage data; The Prophet model is initialized, and the Prophet model is trained using the preprocessed historical storage data. Future time windows are constructed and predictions are made to output prediction results containing confidence intervals.

[0011] According to one embodiment of this application, the step of using the time-series prediction model to predict the storage usage of the target database within a future preset period and generating a prediction result value specifically involves: The future time window is constructed based on the current time point and the prediction period length set by the user. The time range predicted by the Prophet model is defined based on the future time window. The Prophet model calculates the predicted storage usage at each time point and outputs the confidence interval corresponding to the predicted storage usage. Extract preset fields, including timestamps, predicted values, and upper and lower confidence intervals, from the predicted storage usage to form structured prediction result values.

[0012] According to one embodiment of this application, comparing the predicted result value with a preset capacity threshold, and generating capacity warning information if the predicted result value exceeds the preset capacity threshold, specifically involves: Obtain the maximum database capacity threshold set by the user or system, and compare the maximum database capacity threshold with the predicted value at each time point in the structured prediction result value one by one to identify the first time point when the predicted result value first exceeds the maximum database capacity threshold. Based on any future time point after the first time point, if the predicted storage usage is greater than or equal to the preset capacity threshold, then the capacity warning trigger condition is determined to be met.

[0013] According to one embodiment of this application, the method further includes: When the warning condition is triggered, a capacity warning message is generated, which includes the expected time point when the preset capacity threshold is reached, the current predicted value, the confidence interval, and the remaining available space.

[0014] According to one embodiment of this application, the method further includes: The training data is automatically updated and the Prophet model is retrained within a preset period to adapt to dynamic changes in database storage usage trends. The Prophet model was optimized based on parametric grid search and RMSE evaluation mechanism.

[0015] According to one embodiment of this application, the method further includes: connecting the early warning information to the cloud platform operation and maintenance management system to achieve linked automatic expansion and load balancing.

[0016] A second aspect of this application provides a cloud platform database storage usage prediction device based on time-series modeling, comprising: The model building module is suitable for collecting historical storage usage data of the target database in the cloud platform and building a time series prediction model based on the historical storage usage data. The time series prediction model is a Prophet model that supports the modeling of trend and periodic items. The prediction module is adapted to use the time-series prediction model to predict the storage usage of the target database within a future preset period and generate prediction result values. The numerical comparison module is adapted to compare the predicted result value with a preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, a capacity warning message is generated.

[0017] A third aspect of this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the cloud platform database storage usage prediction method based on time-series modeling as described in any of the first aspects.

[0018] This application also provides a non-volatile computer storage medium storing computer-executable instructions thereon, wherein the computer program, when executed by a processor, implements the cloud platform database storage usage prediction method based on time-series modeling as described in any of the embodiments of the first aspect above. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating the prediction method for cloud platform database storage based on time-series modeling, provided for embodiments of this application; Figure 2 A schematic diagram of the structure of a cloud platform database storage usage prediction device based on time series modeling provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0020] Figure label: 110. Model building module; 120. Prediction module; 130. Numerical comparison module; 810, Processor; 820, Communication interface; 830, Memory; 840, Communication bus. Detailed Implementation

[0021] To more clearly illustrate the overall concept of this application, a detailed explanation is provided below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below. It should be noted that, unless otherwise specified, the embodiments of this application and the features thereof can be combined with each other.

[0023] In this application, unless otherwise expressly specified and limited, the "above" or "below" of the second feature can mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples.

[0024] like Figure 1 As shown, the first aspect of this application provides a cloud platform database storage usage prediction method based on time series modeling, including: Step 100: Collect historical storage usage data of the target database in the cloud platform, and build a time series prediction model based on the historical storage usage data. The time series prediction model is the Prophet model, which supports the modeling of trend items and periodic items.

[0025] Step 200: Use a time-series forecasting model to predict the storage usage of the target database within a future preset period and generate the forecast result value.

[0026] Step 300: Compare the predicted result value with the preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, generate a capacity warning message.

[0027] In step 100, historical storage usage data of the target database in the cloud platform needs to be dynamically collected through a heterogeneous database adaptation layer. This process involves unified access to different types of databases (such as relational and non-relational databases) and interfacing with different database systems in the cloud platform to ensure comprehensive and accurate historical data. For example, in a typical scenario, data from the past few months needs to be collected from databases such as MySQL, PostgreSQL, and MongoDB. This data then undergoes preprocessing steps including missing value imputation and outlier removal to ensure data quality. After the Prophet model is initialized, it will be trained using the preprocessed data to construct a time window suitable for future predictions.

[0028] This step primarily relies on an efficient heterogeneous database adaptation layer to standardize multi-source data, ensuring data consistency and availability. Simultaneously, linear interpolation algorithms are used to fill in missing values, and the Z-score method is employed to filter out outliers, providing a high-quality data foundation for subsequent model training. This approach effectively addresses prediction bias caused by incomplete or anomaly-prone data, improving model stability and accuracy.

[0029] It not only provides support for multiple types of databases but also ensures the quality of input data through a rigorous data cleaning process. This is crucial for building reliable predictive models, as any noise or errors can lead to inaccurate predictions. Furthermore, this approach allows even databases from complex and heterogeneous environments to be efficiently incorporated into the analysis, enhancing the overall system's adaptability and flexibility.

[0030] In step 200, this step first constructs a future time window based on the current time and the prediction period length set by the user. For example, if the current date is June 26th, and the user wants to predict the situation for the next 30 days, a time window from June 26th to July 25th will be constructed. Next, based on this future time window, the time range for the Prophet model's prediction is defined, and the prediction storage usage at each time point is calculated, while the corresponding confidence interval is output.

[0031] The Prophet model can decompose trend components, periodic components, and holiday effects in time series data, thus more accurately capturing changes in database storage usage patterns. Especially when considering factors such as business growth and seasonal fluctuations, the Prophet model can provide predictions that are closer to actual needs.

[0032] By accurately predicting future storage needs, enterprises can plan resource allocation in advance, avoiding service interruptions or performance degradation due to insufficient storage space. Furthermore, the provided confidence intervals increase the reliability of the predictions, allowing decision-makers to develop corresponding strategies based on different risk appetites, further optimizing resource allocation and improving operational efficiency.

[0033] In step 300, the system compares the predicted value generated in step 200 with the preset maximum database capacity threshold one by one. For example, assuming the maximum database capacity threshold is set to 1000GB, and the predicted result shows that the storage usage will reach 1050GB on a future day, the system will identify this over-limit situation and trigger the capacity warning mechanism. Once the predicted value is detected to exceed the maximum database capacity threshold for the first time, the system will begin monitoring any future time point thereafter. When the predicted storage usage is greater than or equal to the preset capacity threshold, the system determines that the capacity warning trigger condition has been met.

[0034] The core of this step lies in timely detection of potential storage exhaustion risks by monitoring the relationship between prediction results and capacity thresholds in real time. Specifically, the system extracts key fields from the prediction results, such as timestamps, predicted values, and their upper and lower confidence intervals, and compares them with the set thresholds. This data analysis-based early warning mechanism provides the operations team with sufficient warning time before potential problems occur, allowing them to take appropriate measures to address them.

[0035] By issuing timely capacity alerts, enterprises can avoid emergency expansion or other contingency operations due to insufficient storage space, reducing operational costs and improving system stability and availability. Furthermore, combining the expected threshold time, current forecast value, confidence interval, and remaining available space in the alert information helps operations personnel make more informed and rational decisions, thereby better ensuring business continuity and data security.

[0036] According to the cloud platform database storage usage prediction method based on time-series modeling provided in the first aspect of this application, historical storage usage data of the target database in the cloud platform is first collected, and a Prophet time-series prediction model supporting trend and periodic term modeling is constructed based on this historical data. The Prophet model is chosen because it can effectively handle trend changes, seasonal fluctuations, and holiday effects in time series, giving it a significant advantage in dealing with complex and ever-changing data patterns. The beneficial effect of this step is that, through in-depth analysis and learning of historical data, the long-term trends and periodic changes in database storage usage can be accurately captured, laying a solid foundation for subsequent accurate predictions.

[0037] Next, the existing Prophet model is used to predict storage usage over a pre-defined period, generating forecast values ​​with confidence intervals. This process considers not only historical patterns of database storage usage but also external factors that may influence storage demand (such as holidays). The core value of this step lies in providing a forward-looking view of future storage needs, helping enterprises plan resource allocation in advance and avoid service interruptions or performance degradation due to insufficient storage space. Simultaneously, the introduction of confidence intervals increases the reliability of the forecast results, allowing decision-makers to formulate appropriate strategies based on different risk preferences.

[0038] Finally, the predicted value is compared with a pre-set capacity threshold. If the predicted value exceeds the threshold, a capacity warning is immediately generated. This real-time data analysis-based warning mechanism can promptly notify the operations team to take action before potential problems occur, such as proactively scaling up or optimizing data management strategies. This approach significantly improves the proactiveness and efficiency of database operations, reduces emergency response costs caused by storage exhaustion, and helps maintain high system availability and stability.

[0039] Overall, this approach achieves effective management of cloud platform database storage through intelligent prediction and early warning, thereby improving resource utilization and service quality.

[0040] In some embodiments of this application, historical storage usage data of the target database in the cloud platform is collected, and a time-series prediction model is constructed based on the historical storage usage data, specifically as follows: The heterogeneous database adaptation layer is used to dynamically collect historical storage usage data. The heterogeneous database adaptation layer is suitable for unified access to relational databases and non-relational databases, and can connect to different types of database systems in the cloud platform. Missing values ​​are filled and outliers are removed from the collected historical storage and usage data; Initialize the Prophet model, train the Prophet model using preprocessed historical data, construct future time windows and make predictions, and output prediction results containing confidence intervals.

[0041] Suppose we have an application running on a cloud platform, with its data distributed across MySQL (a relational database) and MongoDB (a non-relational database). We need to monitor the storage usage of these two databases for future capacity planning. First, through a heterogeneous database adaptation layer, we can seamlessly collect historical storage usage data from these two different types of databases over the past six months. For example, we can retrieve daily disk usage (in GB) from the MySQL database and similar information from MongoDB. This data includes key metrics such as timestamps and used storage space.

[0042] The collected data may contain missing data for certain dates or abnormally high values ​​due to system malfunctions. For missing values, linear interpolation can be used to fill in the gaps. For example, if data from May 20th is missing, but we know that data from May 19th and May 21st are 300GB and 320GB respectively, then linear interpolation can estimate the value for May 20th to be approximately 310GB. As for outliers, such as a sudden spike in storage usage to 1000GB on a particular day, far exceeding the normal range, the Z-score method can be used to identify and remove such outliers, ensuring that the data used in subsequent analyses is accurate and reliable.

[0043] Next, initialize the Prophet model and train it using the preprocessed data. During this process, considering business characteristics, it may be necessary to configure model parameters, such as enabling daily, weekly, or holiday cycles (if there are specific days that cause a surge in storage demand). After training, construct a 30-day time window and use the model to predict storage usage during this period. The prediction results include not only the daily projected storage usage (yhat) but also upper and lower confidence intervals (yhat_lower, yhat_upper), which helps to understand the uncertainty of the prediction results.

[0044] By following these steps, the operations team can anticipate database storage usage trends over the next month and identify critical time points where expansion may be necessary. For example, if forecasts indicate that storage usage will approach 90% of the maximum capacity threshold by mid-July, expansion plans can be scheduled in advance to avoid service interruptions due to storage exhaustion. Furthermore, the provision of confidence intervals provides an additional safety margin for decision-making, making resource planning more flexible and robust. This approach significantly improves the efficiency and predictability of database management and reduces the risks associated with unforeseen circumstances.

[0045] In some embodiments of this application, a time-series prediction model is used to predict the storage usage of the target database within a future preset period, generating prediction result values, specifically as follows: Construct a future time window based on the current time point and the prediction period length set by the user; The time range for prediction by the Prophet model is defined based on a future time window. The Prophet model calculates the predicted storage usage at each time point and outputs the confidence interval corresponding to the predicted storage usage. Extract preset fields, including timestamps, predicted values, and upper and lower confidence intervals, from the predicted storage usage to form structured prediction results.

[0046] Assuming the current time is June 26, 2025, a user wants to predict the storage usage of a database over the next 30 days (from June 26, 2025 to July 25, 2025). First, based on the current time and the user-defined prediction period (30 days), a future time window is constructed. This time window defines the specific time period that the Prophet model will predict, from June 26, 2025 to July 25, 2025.

[0047] Next, the Prophet model calculates the predicted storage usage for each time point and outputs the corresponding confidence interval. For example, for July 15, 2025, the model predicts storage usage of 450GB, with a 95% confidence interval of [430GB, 470GB]. This means that based on historical data and model training results, we have 95% confidence that the actual storage usage on that date will fall between 430GB and 470GB. Subsequently, key fields from these predictions are extracted, including the timestamp (e.g., 2025-07-15), the predicted value (450GB), and the upper and lower confidence intervals (430GB to 470GB), forming structured prediction results for easier subsequent analysis and early warning assessment.

[0048] This time-series modeling approach provides accurate forecasts of future storage demand, significantly improving the foresight and flexibility of resource management. By understanding storage demand trends over a future period, enterprises can rationally plan expansion initiatives, avoiding service interruptions or performance degradation due to insufficient storage space. Furthermore, the provided confidence intervals not only increase the reliability of the forecasts but also help decision-makers develop more appropriate strategies based on different risk appetites, such as taking preventative measures when approaching capacity limits.

[0049] Furthermore, because the predictions are continuously updated based on the latest data, the system can respond and adjust its predictions promptly, even in the face of rapidly changing business needs or unforeseen circumstances, ensuring it always remains in an optimal state. This approach allows the operations team to shift from reactive to proactive management, significantly improving the stability and efficiency of the database system and reducing unnecessary costs.

[0050] In some embodiments of this application, the predicted value is compared with a preset capacity threshold. If the predicted value exceeds the preset capacity threshold, a capacity warning message is generated. Specifically: Obtain the maximum database capacity threshold set by the user or system, and compare the maximum database capacity threshold with the predicted value at each time point in the structured prediction result value to identify the first time point when the predicted result value first exceeds the maximum database capacity threshold. Based on any future time point after the first time point, if the predicted storage usage is greater than or equal to the preset capacity threshold, the capacity warning trigger condition is determined to be met.

[0051] Assuming the maximum capacity threshold for the database is set to 1000GB, the system compares these predicted values ​​with this threshold one by one based on the predictions generated by the Prophet model. For example, in the predictions for the next 30 days, it was found that the predicted storage usage for July 18, 2025, reached 1020GB, marking the first time the 1000GB threshold was exceeded. The system then continues to check all time points after July 18. If the predicted storage usage is still greater than or equal to 1000GB on any subsequent date (such as July 20), the capacity warning trigger condition is met, and a corresponding capacity warning message is generated.

[0052] First, this method accurately identifies the specific time when the capacity threshold is first exceeded, enabling the operations team to plan for expansion or other contingency measures in advance, thus avoiding service interruptions or performance degradation caused by insufficient storage space. For example, knowing the specific date of exceeding the limit, the team can arrange data migration or increase storage resources in advance to ensure that business continuity and stability are not affected.

[0053] Secondly, continuous monitoring after the initial threshold is further enhanced, improving the effectiveness of the early warning mechanism. Even after the threshold is first exceeded, the system continues to issue warnings as long as future predicted values ​​remain close to or exceed the threshold, ensuring that operations personnel are always aware of potential risks and take timely action. This proactive early warning mechanism not only improves system reliability and availability but also reduces the additional costs and risks associated with emergency capacity expansion. Ultimately, through this automated capacity warning process, enterprises can manage their cloud platform database resources more efficiently, optimize resource allocation, and improve overall operational efficiency.

[0054] In some embodiments of this application, the method further includes: When the warning conditions are triggered, a capacity warning message is generated, which includes the expected time point when the preset capacity threshold is reached, the current predicted value, the confidence interval, and the remaining available space.

[0055] Assuming the maximum database capacity threshold is 1000GB, and based on the Prophet model's prediction, the database storage usage is expected to reach 1020GB on July 18, 2025, exceeding the threshold for the first time. At this point, the system will automatically generate a capacity warning message. This warning message will include the following: The expected time to reach the preset capacity threshold is, for example, July 18, 2025. Current forecast: 1020GB; Confidence interval: For example, [990GB, 1050GB], which means there is a 95% probability that the actual storage usage will fall within this range; Remaining available space: The amount of space remaining before the warning time point is reached. For example, if the current total capacity is 1000GB and 900GB has been used, then the remaining available space is 100GB.

[0056] First, this detailed early warning information provides the operations team with clear operational guidance, enabling them to accurately determine when action is needed and its urgency. By clearly specifying the exact time and predicted value of the expected overrun, the team can plan ahead for capacity expansion or other mitigation measures, avoiding service interruptions or performance degradation due to unforeseen circumstances.

[0057] Secondly, providing confidence intervals helps assess the uncertainty of forecast results, making the decision-making process more scientific and rational. For example, if the upper limit of the confidence interval is much higher than the threshold, it indicates a significant risk, and it may be necessary to prepare countermeasures earlier; conversely, if the confidence interval is close to the threshold, the plan can be flexibly adjusted according to the actual situation. In addition, information on remaining available space further helps the team understand the current resource status, facilitating more accurate resource management decisions.

[0058] Finally, this comprehensive and detailed early warning mechanism significantly improves the system's maintainability and stability. It not only enhances the ability to anticipate potential problems but also strengthens the response speed to emergencies, reducing the negative impact of events such as storage exhaustion. In this way, enterprises can optimize resource allocation and reduce operating costs while ensuring business continuity.

[0059] In some embodiments of this application, the method further includes: The training data is automatically updated and the Prophet model is retrained within a preset period to adapt to dynamic changes in database storage usage trends. The Prophet model was optimized based on parametric grid search and RMSE evaluation mechanism.

[0060] Within a preset period (e.g., daily or weekly), the system automatically adds the latest database storage usage data to the historical training dataset and removes expired data to ensure that the trends and cyclical features learned by the model remain consistent with the current situation. For example, assuming the initial training data is historical data from the past 90 days, when the model is updated weekly, the system adds data from the most recent week and deletes data from the oldest week, thus forming a new training set. Subsequently, the Prophet model is retrained using the updated data to adapt to the dynamic changes in database storage usage trends.

[0061] The mechanism operates on the principle of a sliding window, ensuring that the model always makes predictions based on the latest and most relevant data. This dynamic updating approach is particularly important for scenarios involving accelerated business growth and seasonal fluctuations. For example, if an e-commerce platform experiences a significant increase in storage during a promotional period, failure to update the model in a timely manner could lead to predictions lagging behind actual trends, affecting the accuracy of early warnings.

[0062] To further improve the model's predictive performance, the system also introduces a grid search and root mean square error (RMSE) evaluation mechanism. Specifically, each time the model is retrained, the system tries various combinations of model parameters (such as trend smoothing coefficient, periodic term strength, holiday effect on / off, etc.), then backtests using historical data, calculates the RMSE value for each set of parameters, and finally selects the set of parameters with the smallest RMSE as the optimal configuration.

[0063] For example, during a model tuning process, the system tried three different trend smoothing settings and two periodic modeling methods, for a total of six combinations. By performing predictive backtesting on data from the past 30 days and calculating the RMSE value for each combination, the system found that one combination had the lowest RMSE, indicating that this combination had the best fit under the current data distribution, and therefore applied it to a new round of model training.

[0064] This method ensures the model constantly learns the latest database storage usage trends by periodically and automatically updating training data, enhancing the model's adaptability and predictive foresight. Combined with model tuning based on parametric grid search and RMSE evaluation mechanisms, it achieves automated optimization of model hyperparameters, significantly improving prediction accuracy and stability. This dual-drive mechanism of "dynamic update + intelligent tuning" not only improves the predictive system's adaptability but also reduces manual maintenance costs, providing solid technical support for intelligent capacity management of cloud platform databases.

[0065] In some embodiments of this application, the method further includes: The early warning information is integrated into the cloud platform operation and maintenance management system to achieve coordinated automatic expansion and load balancing.

[0066] In the prediction method provided in this application, after capacity warning information is generated, the system does not merely stop at the "notification" level, but further standardizes and encapsulates the warning information and pushes it to the cloud platform's operation and maintenance management system (such as CMP or AIOps platform). After receiving the warning information, the operation and maintenance management system can automatically trigger corresponding resource scheduling actions based on preset policies.

[0067] For example, suppose the storage usage of a database instance is expected to reach the threshold of 1000GB on July 18th. After generating an alert, the system pushes it to the cloud platform's operations and maintenance (O&M) module via API. The O&M system parses the alert, identifies the impending capacity overrun risk, and automatically performs one or a combination of the following actions based on pre-configured response strategies: Automatic scaling: Call the cloud platform API to scale up the storage volume where the database is located, for example, from 1TB to 1.5TB; Load balancing: Migrate some data tables or read / write requests to other database instances to reduce the pressure on the current instance; Alarm notification + manual confirmation mechanism: Send alarm emails or text messages to maintenance personnel and prompt manual intervention decision-making at critical business time points; Resource recycling recommendation: If some historical cold data is found to have not been accessed for a long time, it is recommended to archive or migrate it to a low-cost storage space.

[0068] This process achieves a complete closed-loop workflow from prediction and early warning to strategy judgment and automatic execution, thereby improving the intelligence level of cloud platform resource management.

[0069] Integrating capacity warning information into the operation and maintenance management system breaks through the limitation of traditional predictive models that "only predict but do not act," enabling prediction results to directly drive actual resource scheduling behavior and improving the system's autonomous response capability and operation and maintenance efficiency. Especially in large-scale distributed database environments, this automated linkage mechanism can significantly reduce the frequency of manual intervention and the risk of misoperation.

[0070] Secondly, it supports dynamic elastic scaling and load optimization, which helps improve overall system performance and stability. By proactively identifying capacity bottlenecks and adjusting resource configurations in advance, it can avoid service interruptions and increased latency caused by sudden storage exhaustion, thereby ensuring business continuity and user experience.

[0071] Finally, it enhances the cloud platform's intelligent operations and maintenance capabilities (AIOps), laying the foundation for building a self-healing data center. In the future, it can further optimize linkage strategies by combining AI algorithms to achieve more refined resource scheduling and cost control, driving the cloud platform towards a "prediction-driven" operations and maintenance model.

[0072] like Figure 2 As shown, a second aspect of this application provides a cloud platform database storage usage prediction device based on time-series modeling, comprising: The model building module 110 is suitable for collecting historical storage usage data of the target database in the cloud platform and building a time series prediction model based on the historical storage usage data. The time series prediction model is a Prophet model that supports the modeling of trend items and periodic items. Prediction module 120 is adapted to use the time-series prediction model to predict the storage usage of the target database within a future preset period and generate prediction result values. The numerical comparison module 130 is adapted to compare the predicted result value with a preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, a capacity warning message is generated.

[0073] The cloud platform database storage usage prediction device based on time-series modeling provided in the second aspect of this application can realize the cloud platform database storage usage prediction method based on time-series modeling in any of the embodiments of the first aspect above. Therefore, it can achieve any of the technical effects in the cloud platform database storage usage prediction method based on time-series modeling above, which will not be elaborated here.

[0074] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the cloud platform database storage usage prediction method based on time-series modeling as described in any of the first aspects above.

[0075] Figure 3An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute the cloud platform database storage usage prediction method based on time-series modeling in any embodiment of the first aspect described above. This method includes: Step 100: Collect historical storage usage data of the target database in the cloud platform, and build a time series prediction model based on the historical storage usage data. The time series prediction model is the Prophet model, which supports the modeling of trend items and periodic items.

[0076] Step 200: Use a time-series forecasting model to predict the storage usage of the target database within a future preset period and generate the forecast result value.

[0077] Step 300: Compare the predicted result value with the preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, generate a capacity warning message.

[0078] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0079] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the cloud platform database storage usage prediction method based on time-series modeling provided by the above methods, the method comprising: Step 100: Collect historical storage usage data of the target database in the cloud platform, and build a time series prediction model based on the historical storage usage data. The time series prediction model is the Prophet model, which supports the modeling of trend items and periodic items.

[0080] Step 200: Use a time-series forecasting model to predict the storage usage of the target database within a future preset period and generate the forecast result value.

[0081] Step 300: Compare the predicted result value with the preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, generate a capacity warning message.

[0082] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the cloud platform database storage usage prediction method based on time-series modeling provided by the methods described above, the method comprising: Step 100: Collect historical storage usage data of the target database in the cloud platform, and build a time series prediction model based on the historical storage usage data. The time series prediction model is the Prophet model, which supports the modeling of trend items and periodic items.

[0083] Step 200: Use a time-series forecasting model to predict the storage usage of the target database within a future preset period and generate the forecast result value.

[0084] Step 300: Compare the predicted result value with the preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, generate a capacity warning message.

[0085] Finally, the present invention also provides a non-volatile computer storage medium storing computer-executable instructions thereon, which, when executed by a processor, implements the time-series modeling-based cloud platform database storage usage prediction method provided by the above methods, the method comprising: Step 100: Collect historical storage usage data of the target database in the cloud platform, and build a time series prediction model based on the historical storage usage data. The time series prediction model is the Prophet model, which supports the modeling of trend items and periodic items.

[0086] Step 200: Use a time-series forecasting model to predict the storage usage of the target database within a future preset period and generate the forecast result value.

[0087] Step 300: Compare the predicted result value with the preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, generate a capacity warning message.

[0088] For any parts not mentioned in this application, existing technologies may be used or referenced.

[0089] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0090] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A cloud platform database storage usage prediction method based on time-series modeling, characterized in that, include: Collect historical storage usage data of the target database in the cloud platform, and construct a time series prediction model based on the historical storage usage data. The time series prediction model is a Prophet model that supports the modeling of trend items and periodic items. The time-series prediction model is used to predict the storage usage of the target database within a future preset period, and a prediction result value is generated. The predicted value is compared with a preset capacity threshold. If the predicted value exceeds the preset capacity threshold, a capacity warning message is generated.

2. The cloud platform database storage usage prediction method based on time series modeling according to claim 1, characterized in that, The historical storage and usage data of the target database in the data collection cloud platform is used to construct a time-series prediction model based on the historical storage and usage data, specifically as follows: The heterogeneous database adaptation layer is used to dynamically collect the historical storage usage data. The heterogeneous database adaptation layer is suitable for unified access to relational databases and non-relational databases, and can interface with different types of database systems in the cloud platform. Missing values ​​are filled and outliers are removed from the collected historical storage and usage data; The Prophet model is initialized, and the Prophet model is trained using the preprocessed historical storage data. Future time windows are constructed and predictions are made to output prediction results containing confidence intervals.

3. The cloud platform database storage usage prediction method based on time series modeling according to claim 2, characterized in that, The step of using the time-series prediction model to predict the storage usage of the target database within a future preset period and generating prediction result values ​​specifically involves: The future time window is constructed based on the current time point and the prediction period length set by the user. The time range predicted by the Prophet model is defined based on the future time window. The Prophet model calculates the predicted storage usage at each time point and outputs the confidence interval corresponding to the predicted storage usage. Extract preset fields, including timestamps, predicted values, and upper and lower confidence intervals, from the predicted storage usage to form structured prediction result values.

4. The cloud platform database storage usage prediction method based on time series modeling according to claim 3, characterized in that, The step of comparing the predicted result value with a preset capacity threshold, and generating a capacity warning message if the predicted result value exceeds the preset capacity threshold, specifically involves: Obtain the maximum database capacity threshold set by the user or system, and compare the maximum database capacity threshold with the predicted value at each time point in the structured prediction result value one by one to identify the first time point when the predicted result value first exceeds the maximum database capacity threshold. Based on any future time point after the first time point, if the predicted storage usage is greater than or equal to the preset capacity threshold, then the capacity warning trigger condition is determined to be met.

5. The cloud platform database storage usage prediction method based on time series modeling according to claim 4, characterized in that, The method also includes: When the warning condition is triggered, a capacity warning message is generated, which includes the expected time point when the preset capacity threshold is reached, the current predicted value, the confidence interval, and the remaining available space.

6. The cloud platform database storage usage prediction method based on time series modeling according to any one of claims 1 to 5, characterized in that, The method also includes: The training data is automatically updated and the Prophet model is retrained within a preset period to adapt to dynamic changes in database storage usage trends. The Prophet model was optimized based on parametric grid search and RMSE evaluation mechanism.

7. The cloud platform database storage usage prediction method based on time series modeling according to any one of claims 1 to 5, characterized in that, The method also includes: The early warning information is integrated into the cloud platform operation and maintenance management system to achieve coordinated automatic expansion and load balancing.

8. A cloud platform database storage usage prediction device based on time series modeling, characterized in that, include: The model building module is suitable for collecting historical storage usage data of the target database in the cloud platform and building a time series prediction model based on the historical storage usage data. The time series prediction model is a Prophet model that supports the modeling of trend and periodic items. The prediction module is adapted to use the time-series prediction model to predict the storage usage of the target database within a future preset period and generate prediction result values. The numerical comparison module is adapted to compare the predicted result value with a preset capacity threshold. If the predicted result value exceeds the preset capacity threshold, a capacity warning message is generated.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the cloud platform database storage usage prediction method based on time-series modeling as described in any one of claims 1 to 7.

10. A non-volatile computer storage medium storing computer-executable instructions thereon, characterized in that, When the computer program is executed by the processor, it implements the cloud platform database storage usage prediction method based on time-series modeling as described in any one of claims 1 to 7.

Images