Containerized operation and maintenance orchestration engine and resource dynamic scheduling method in multi-cloud environment

Abstract

The invention relates to the technical field of cloud computing and containerization, and discloses a containerization operation and maintenance orchestration engine and a resource dynamic scheduling method in a multi-cloud environment. The method comprises the steps of collecting and integrating multi-cloud environment information, collecting service characteristics of each cloud platform, carrying out analysis and modeling of containerized application, constructing an application model according to an analysis result, formulating a resource dynamic scheduling strategy, and executing the scheduling strategy by utilizing a containerized operation and maintenance orchestration engine. The multi-resource dynamic scheduling method solves the problems that in a traditional resource dynamic scheduling prediction method, resource demand and supply prediction is not accurate, resource dynamic scheduling lacks flexibility and intelligence, and complexity of a multi-cloud environment cannot be effectively handled, and a complex change mode of resource demand is accurately captured; dynamic changes of cloud platform resource supply are effectively coped with, data are collected in real time, a prediction model is optimized through a machine learning algorithm, an accurate basis is provided for a resource dynamic scheduling strategy, and efficient and intelligent resource scheduling is achieved.

Description

Technical Field

[0001] The present invention relates to the field of cloud computing and containerization technology, and specifically to a containerized operation and maintenance orchestration engine and a dynamic resource scheduling method in a multi-cloud environment. Background Art

[0002] Traditional dynamic resource scheduling forecasting often uses simple statistical analysis methods for resource demand prediction, such as the moving average method. This method calculates the average resource usage over a period of time to predict future resource demand. This method only considers the simple average of historical data and fails to fully explore the trends and cyclical characteristics in the data. For resource supply forecasting, traditional methods rely on static resource planning information from cloud service providers and lack dynamic monitoring and analysis of real-time resource supply fluctuations. For example, cloud service providers set resource supply scales in advance based on their own hardware equipment procurement plans and estimated user demand. In actual operations, it is difficult to adjust resource supply forecasts in a timely manner based on sudden changes in resource demand or hardware failures. For overall dynamic resource scheduling, traditional technologies mainly rely on manual experience and simple rule engines. Administrators develop basic resource allocation rules based on long-term experience, such as allocating a fixed proportion of resources to critical applications during peak business hours. However, this approach cannot adapt to complex and changing business scenarios and dynamic cloud environments. Traditional simple statistical analysis methods are unable to capture the complex changing patterns of resource demand, resulting in a large deviation between the prediction results and actual demand. This may cause the application to degrade in performance due to insufficient resources during business peak periods, or idle and wasted resources during business troughs. In resource supply prediction, the traditional method of relying on static planning information cannot cope with the dynamic changes in cloud platform resource supply, such as a sudden hardware failure leading to a reduction in resource supply, or a surge in resource demand due to a large influx of new users, while the supply cannot be adjusted in time, affecting service quality. The traditional resource dynamic scheduling method that relies on manual experience and simple rule engines cannot be adjusted in a timely and intelligent manner according to real-time resource demand and supply changes. For this reason, this application now proposes a containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment. Summary of the Invention (1) Technical problems solved

[0003] In response to the shortcomings of the existing technology, the present invention provides a containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment. It has the advantages of accurately capturing the complex changing patterns of resource demand, effectively responding to the dynamic changes in resource supply of cloud platforms, collecting data in real time and optimizing the prediction model using machine learning algorithms to achieve efficient and intelligent resource scheduling. It solves the problems of inaccurate resource demand and supply predictions, lack of flexibility and intelligence in dynamic resource scheduling, and inability to effectively cope with the complexity of multi-cloud environments in traditional resource dynamic scheduling prediction methods. (2) Technical solution

[0004] To achieve the above objectives, the present invention provides the following technical solution: a containerized operation and maintenance orchestration engine and a resource dynamic scheduling method in a multi-cloud environment, comprising the following steps: S1. Collect and integrate multi-cloud environment information, connect to different cloud platforms through the interface adapter module, and obtain resource information of each cloud platform; S2. Collect the service characteristics of each cloud platform and organize and store the service characteristic information in a unified format to build a multi-cloud environment resource information database; S3. Conduct analysis and modeling of containerized applications, and conduct in-depth analysis of containerized applications that need to be deployed; S4. Build an application model based on the analysis results; S5. Formulate dynamic resource scheduling strategies; S6. Use the containerized operation and maintenance orchestration engine to execute the scheduling strategy.

[0005] Preferably, the information obtained in step S1 includes the number of virtual machines, memory size, storage capacity, and network bandwidth, and step S3 analyzes the dependencies between containers within the application, data flow, and resource demand characteristics.

[0006] Preferably, in the analysis and modeling steps of the containerized application, the application analysis tool adopts a combination of static analysis and dynamic monitoring. In the static analysis stage, the code structure, configuration files, etc. of the application are deeply scanned to preliminarily determine the dependencies between containers and the range of resource requirements. In the dynamic monitoring stage, when the application is running in a small-scale test, the resource usage data of the container, including CPU, memory, disk I / O, etc., is collected in real time, and the static analysis results are corrected and improved to make the constructed application model more accurate. At the same time, in order to cope with dynamic changes in the application operation process, the application model has the ability to adaptively adjust. When it is detected that the application business volume has a large fluctuation, resulting in changes in resource requirements, the model automatically adjusts the resource allocation ratio and topology structure between containers according to preset rules and real-time data to ensure that the application can obtain reasonable resource allocation in different business scenarios and maintain stable operation.

[0007] Preferably, the containerized operation and maintenance orchestration engine executes the scheduling strategy. The engine automatically creates container instances on the corresponding cloud platform according to the formulated resource allocation plan, configures network connections to ensure smooth communication between containers, and monitors the running status of containers in real time. Once abnormal resource usage or load changes are found, resources are immediately reallocated according to the dynamic adjustment strategy to ensure that containerized applications run efficiently and stably in a multi-cloud environment.

[0008] Preferably, the formulation of the dynamic resource scheduling strategy comprehensively considers the resource status in the multi-cloud environment resource information database, the service cost of each cloud platform and the resource demand in the application model, and uses an intelligent algorithm, which includes a hybrid algorithm combining a genetic algorithm and a particle swarm algorithm, to calculate the optimal resource allocation plan, determine which cloud platform and which resource each container should be deployed on, and how to dynamically adjust resource allocation according to real-time resource usage during application operation, and introduce a prediction mechanism in the resource scheduling process. Using machine learning algorithms, we analyze and train historical resource usage data, application business volume change data, and cloud platform resource supply data to establish resource demand prediction models and resource supply prediction models. Through the prediction models, we can predict in advance the resource demand change trend of applications in the future and the possible resource supply fluctuations of each cloud platform, providing a forward-looking reference for the formulation of dynamic resource scheduling strategies. For example, when it is predicted that the application business volume will increase significantly within a certain period of time, the system will reserve sufficient resources on the appropriate cloud platform in advance to avoid application performance degradation due to insufficient resources; when it is predicted that a cloud platform will soon face resource supply shortages, the resource allocation plan will be adjusted in advance, and some containers will be migrated to other cloud platforms with sufficient resources, so as to improve the system's resource utilization efficiency and ability to respond to emergencies, and ensure the long-term stable and efficient operation of containerized applications in a multi-cloud environment.

[0009] Preferably, the containerized operation and maintenance orchestration engine is used to execute the scheduling strategy, and a distributed deployment architecture is adopted. An engine node is deployed on each cloud platform, and information is exchanged between the nodes through a secure and reliable communication protocol. When a container instance needs to be created, the master node distributes the creation task to the node of the corresponding cloud platform according to the resource allocation plan, and each node processes in parallel, which greatly shortens the container creation time. At the same time, in terms of monitoring the running status of the container, each node collects the running data of the local container in real time, including CPU usage, memory usage, network traffic, etc., and uploads this data to the master node for summary analysis. The master node determines whether dynamic resource adjustment is needed based on the summary data. If necessary, it sends adjustment instructions to the relevant nodes to realize distributed and efficient operation and maintenance management of containerized applications, and improve the reliability and scalability of the system.

[0010] Compared with the existing technology, the present invention provides a containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment, which has the following beneficial effects: 1. The containerized operation and maintenance orchestration engine and resource dynamic scheduling method in the multi-cloud environment. After obtaining the resource demand prediction results and resource supply prediction results, this application outputs them as a basis for formulating resource dynamic scheduling strategies. In the process of formulating resource dynamic scheduling strategies, the resource status in the multi-cloud environment resource information database, the service costs of each cloud platform, and the resource requirements in the application model are comprehensively considered. With accurate resource demand and supply predictions, the scheduling strategy can plan resource allocation in advance. For example, when it is predicted that the application business volume will increase significantly within a certain period of time, the system can reserve sufficient resources on the appropriate cloud platform in advance to avoid application performance degradation due to insufficient resources; when it is predicted that a cloud platform will soon have a tight resource supply, the resource allocation plan is adjusted in advance, and some containers are migrated to other cloud platforms with sufficient resources. This dynamic resource scheduling based on the prediction model greatly improves the resource scheduling efficiency compared to traditional post-responsive scheduling, ensuring that containerized applications run efficiently in a multi-cloud environment.

[0011] 2. This containerized operation and maintenance orchestration engine and dynamic resource scheduling method for a multi-cloud environment. The prediction mechanism and fault tolerance mechanism work together throughout the entire process. On the one hand, by predicting resource supply fluctuations, potential resource supply risks can be discovered in advance to avoid system failures caused by resource supply issues. For example, if a cloud platform's resource supply is predicted to be tight, containers can be migrated in advance to prevent container failures caused by insufficient resources, thereby ensuring that the overall application business is not affected. On the other hand, when a failure occurs during the operation of a containerized application, a deep learning algorithm is used to quantify the dependency strength between containers, combined with the container dependency graph in the application model, and the containers with the greatest business impact are restored first. At the same time, instances of the failed container are recreated on other available resources, and a resource matching algorithm is used to select appropriate cloud platform resource deployment. Software-defined networking (SDN) technology is used to optimize the network topology to ensure smooth communication between containers. These technical measures, assisted by the prediction mechanism, enhance the system's fault tolerance and self-healing capabilities in the face of various failures, ensuring the continuous and stable operation of containerized applications in a multi-cloud environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a schematic flow chart of the steps of the dynamic scheduling method of the present invention; Figure 2 This is a schematic diagram of the architecture of the containerized operation and maintenance orchestration system in a multi-cloud environment of the present invention; Figure 3 This is a flowchart of dynamic resource scheduling in a multi-cloud environment of the present invention Figure 4 This is a schematic diagram of the containerized application analysis and modeling process of the present invention; Figure 5 This is a diagram of the distributed deployment architecture of the containerized operation and maintenance orchestration engine of the present invention; Figure 6 This is a flow chart of the fault tolerance mechanism in a multi-cloud environment of the present invention. DETAILED DESCRIPTION

[0013] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0014] See Figure 2 , containerized operation and maintenance orchestration engine and resource dynamic scheduling method in multi-cloud environment, first, collect and integrate multi-cloud environment information, and connect with different cloud platforms through interface adapter modules to obtain resource information of each cloud platform, including but not limited to the number of virtual machines, memory size, storage capacity, network bandwidth, etc., and collect service characteristics of each cloud platform, such as billing model, service availability level, etc., and organize and store this information in a unified format to build a multi-cloud environment resource information database; carry out analysis and modeling of containerized applications, conduct in-depth analysis of containerized applications that need to be deployed, analyze the dependencies, data flow and resource demand characteristics between containers within the application, and build an application model based on the analysis results. The model not only includes the topological structure between containers, but also accurately marks the resource requirements of each container at different operation stages, such as CPU usage peak, memory usage Upper limit, etc. Based on the above information, a dynamic resource scheduling strategy is formulated, taking into account the resource status in the multi-cloud environment resource information database, the service cost of each cloud platform and the resource demand in the application model, and using intelligent algorithms, such as a hybrid algorithm combining genetic algorithm and particle swarm algorithm, to calculate the optimal resource allocation plan, determine which cloud platform and resource each container should be deployed on, and how to dynamically adjust resource allocation according to real-time resource usage during application operation, and use the containerized operation and maintenance orchestration engine to execute the scheduling strategy. The engine automatically creates container instances on the corresponding cloud platform according to the formulated resource allocation plan, configures network connections, ensures smooth communication between containers, and monitors the container operation status in real time. Once abnormal resource usage or load changes are found, resources are immediately reallocated according to the dynamic adjustment strategy to ensure that containerized applications run efficiently and stably in a multi-cloud environment.

[0015] See Figure 4As an embodiment of the present application, in the analysis and modeling steps of containerized applications, the application analysis tool adopts a combination of static analysis and dynamic monitoring. In the static analysis stage, the dependencies and resource demand ranges between containers are preliminarily determined by deeply scanning the code structure and configuration files of the application; in the dynamic monitoring stage, when the application is running in a small-scale test, the resource usage data of the container, including CPU, memory, disk I / O, etc., is collected in real time, and the static analysis results are corrected and improved to make the constructed application model more accurate. At the same time, in order to cope with dynamic changes in the application operation process, the application model has the ability to adaptively adjust. When it is detected that the application business volume has a large fluctuation, resulting in changes in resource demand, the model automatically adjusts the resource allocation ratio and topology structure between containers according to preset rules and real-time data to ensure that the application can obtain reasonable resource allocation in different business scenarios and maintain stable operation; Containerized application analysis and modeling uses an innovative approach that combines static analysis with dynamic monitoring to conduct in-depth analysis of applications. It not only preliminarily determines container dependencies and resource requirements from static levels such as code structure and configuration files, but also dynamically monitors resource usage data during small-scale application runtime to revise and improve the results, ensuring that the constructed application model accurately reflects the actual situation. The application model has the ability to automatically adjust the resource allocation ratio and topology structure between containers according to dynamic changes in business volume, and can flexibly adapt to different business scenarios to ensure the stable operation of applications in various situations.

[0016] See Figure 5 As an embodiment of the present application, the containerized operation and maintenance orchestration engine adopts a distributed deployment architecture when executing the scheduling strategy, deploys an engine node on each cloud platform, and exchanges information between the nodes through a secure and reliable communication protocol. When it is necessary to create a container instance, the master node distributes the creation task to the nodes of the corresponding cloud platform according to the resource allocation plan, and each node processes in parallel, which greatly shortens the container creation time. At the same time, in terms of monitoring the running status of the container, each node collects the running data of the local container in real time, including CPU usage, memory usage, network traffic, etc., and uploads this data to the master node for summary analysis. The master node determines whether dynamic resource adjustment is needed based on the summary data. If necessary, it sends adjustment instructions to the relevant nodes to realize distributed and efficient operation and maintenance management of containerized applications and improve the reliability and scalability of the system; A distributed deployment architecture is adopted, and engine nodes are deployed on each cloud platform. Each node communicates through a secure protocol. When creating a container instance, the master node distributes tasks, and the nodes process in parallel, which greatly shortens the creation time and improves system efficiency. Each node collects local container operation data in real time and uploads it to the master node. The master node determines whether resources need to be adjusted based on the summarized data. If necessary, it sends instructions in a timely manner to achieve efficient operation and maintenance management of containerized applications.

[0017] See Figure 6 As an embodiment of the present application, a fault tolerance mechanism is introduced in the entire method process. During the multi-cloud environment information collection and integration process, if a cloud platform interface fails, the interface adapter module automatically switches to the backup interface or uses cached historical data for emergency processing, and promptly sends fault alarm information to the administrator. During the operation of the containerized application, when a container fails, the containerized operation and maintenance orchestration engine immediately starts the fault recovery process, determines the other affected containers based on the container dependencies in the application model, recreates the instance of the faulty container on other available resources, and adjusts the network connection between the related containers to ensure that the overall application business is not affected. At the same time, an in-depth analysis of the cause of the failure is conducted, and the failure log is recorded to provide a basis for subsequent system optimization, improve the system's fault tolerance and self-repair capabilities in the face of various failures, and ensure the continuous and stable operation of containerized applications in a multi-cloud environment; in the process of dynamic resource scheduling, a prediction mechanism is introduced. Using machine learning algorithms, historical resource usage data, application business volume change data, and cloud platform resource supply data are analyzed and trained to establish resource demand prediction models and resource supply prediction models. Using predictive models, we can predict in advance the changing trends in application resource demand over time and the potential fluctuations in resource supply across various cloud platforms, providing a forward-looking reference for developing dynamic resource scheduling strategies. For example, when a significant increase in application traffic is predicted within a certain timeframe, the system reserves sufficient resources on appropriate cloud platforms to prevent application performance degradation due to insufficient resources. When resource constraints are predicted for a particular cloud platform, resource allocation can be adjusted in advance, migrating some containers to other cloud platforms with sufficient resources. This improves the system's resource utilization efficiency and ability to respond to emergencies, ensuring the long-term stable and efficient operation of containerized applications in a multi-cloud environment. Fault tolerance mechanisms are implemented for both multi-cloud environment information collection and containerized application operation. In the event of an interface failure, the system automatically switches to a backup interface or uses cached data as an emergency response. In the event of a container failure, the system identifies the affected container based on dependencies and recreates the instance on other resources, adjusting network connections to minimize the impact on business operations. Furthermore, the system conducts in-depth analysis of the failure cause and records detailed fault logs, providing a basis for subsequent system optimization and continuously improving the system's fault tolerance and self-healing capabilities.

[0018] See Figure 3As an embodiment of the present application, a prediction mechanism is introduced in the process of dynamic resource scheduling, and a machine learning algorithm is used to analyze and train historical resource usage data, application business volume change data, and cloud platform resource supply data to establish a resource demand prediction model and a resource supply prediction model. Through the prediction model, the resource demand change trend of the application in the future and the possible resource supply fluctuations of each cloud platform are predicted in advance, providing a forward-looking reference for the formulation of dynamic resource scheduling strategies. For example, when it is predicted that the application business volume will increase significantly in a certain period of time, the system reserves sufficient resources on the appropriate cloud platform in advance to avoid application performance degradation due to insufficient resources; when it is predicted that a cloud platform will soon face resource supply shortages, the resource allocation plan is adjusted in advance, and some containers are migrated to other cloud platforms with sufficient resources, so as to improve the system's resource utilization efficiency and ability to respond to emergencies, and ensure the long-term stable and efficient operation of containerized applications in a multi-cloud environment; Prediction mechanism, data-driven prediction, uses machine learning algorithms to analyze and train historical resource usage, application business volume changes and cloud platform resource supply data, builds a resource demand and supply prediction model, and uses the prediction model to predict resource demand change trends and supply fluctuations in advance, providing forward-looking reference for the formulation of dynamic resource scheduling strategies, helping the system to respond to resource changes in advance, improve resource utilization efficiency and the ability to respond to emergencies.

[0019] The specific steps of the resource dynamic scheduling prediction mechanism algorithm process are as follows: Input: historical resource usage data, application business volume change data, and cloud platform resource supply data.

[0020] Output: resource demand forecast results, resource supply forecast results.

[0021] Step 1: Initialize the resource demand prediction model ← to an empty model frame, and initialize the resource supply prediction model ← to an empty model frame.

[0022] Step 2: Extract features from historical resource usage data, such as extracting features such as peak and average resource usage in different time periods; organize application business volume change data, such as sorting out business volume growth and decline trends by time series; analyze cloud platform resource supply data to extract features such as the fluctuation cycle of resource supply and the upper limit of supply capacity.

[0023] Step three: Divide the processed historical resource usage data and application business volume change data into training sets and test sets according to a certain ratio, and process the cloud platform resource supply data in the same way.

[0024] Step 4: Select appropriate machine learning algorithms, such as recurrent neural networks (RNNs) and their variants, long short-term memory networks (LSTMs), for resource demand forecasting model training, as they are suitable for processing time series data; select decision tree ensemble algorithms, such as random forests, for resource supply forecasting model training, as they can handle complex data relationships.

[0025] Step 5: Use the training set data to train the resource demand prediction model, and continuously adjust the model parameters, such as the number of hidden layer neurons in LSTM and the learning rate, so that the model's fit to the changing trend of resource demand continues to improve.

[0026] Step 6. At the same time, use the training set data to train the resource supply prediction model, and adjust the parameters such as the number of trees and maximum depth in the random forest model so that the model can accurately predict the fluctuation of cloud platform resource supply.

[0027] Step 7: Use the test set data to verify the trained resource demand forecasting model and calculate the prediction error, such as the root mean square error (RMSE), to evaluate the accuracy of the model.

[0028] In step eight, similarly, the trained resource supply forecasting model is verified using the test set data, the forecast error is calculated, and the accuracy of the model is evaluated.

[0029] Step nine: If the errors of the resource demand forecast model and the resource supply forecast model are both within an acceptable range, the model is saved.

[0030] Step 10: Collect current application business volume data and cloud platform resource supply data in real time.

[0031] Step 11: Input the real-time collected data into the saved resource demand forecasting model to obtain the resource demand forecast results for application in the future period.

[0032] Step 12: Input the real-time collected data into the saved resource supply prediction model to obtain the resource supply prediction results of each cloud platform in the future.

[0033] Step 13: Output the resource demand forecast results and resource supply forecast results to provide a basis for the formulation of dynamic resource scheduling strategies.

[0034] Working principle: This application builds a real-time monitoring and intelligent diagnosis system, using artificial intelligence-based anomaly detection algorithms to monitor the operating status of each cloud platform interface in real time, and analyze multi-dimensional indicators such as interface response time change trends and data traffic fluctuations. Once a cloud platform interface is detected to have signs of failure, such as the interface response time exceeding the preset threshold three times in a row, the system immediately initiates the fault diagnosis process. During the diagnosis process, the intelligent algorithm comprehensively analyzes the historical operating data of the interface, the current network environment, and the real-time status of the cloud platform to locate the cause of the failure and determine whether it is network congestion, a cloud platform failure itself, or an interface program vulnerability. When an interface failure is identified, the interface adapter module responds. If a backup interface exists, a switching strategy selects a backup interface with similar performance and lower current load to the failed interface, ensuring continuous information collection. Furthermore, data caching technology utilizes cached historical data for emergency processing during the backup interface switchover, ensuring real-time resource information. During this process, the system sends a fault alert to the administrator, along with the fault diagnosis results and the emergency measures implemented, to facilitate subsequent decision-making.

[0035] During the operation of containerized applications, intelligent sensors are deployed within each container to collect real-time data on key operational indicators such as CPU, memory, disk I / O, and network communication. Edge computing technology is then used to perform preliminary local analysis to determine if there are any anomalies. If a container failure is detected, the containerized operations and maintenance orchestration engine immediately initiates the recovery process.

[0036] Based on the container dependency graph pre-built in the application model, the system identifies other containers affected by the failed container. Using deep learning algorithms, it quantifies the inter-container dependency strength and prioritizes restoring containers with the greatest business impact. When recreating a failed container instance on other available resources, the system uses a resource matching algorithm to select the appropriate cloud platform resource deployment based on the failed container's historical resource usage data and the current real-time resource load of each cloud platform, ensuring the efficient operation of the newly created container instance. Simultaneously, it automatically adjusts the network connections between related containers, leveraging software-defined networking (SDN) technology to optimize the network topology, ensuring smooth and efficient communication between containers and minimizing the impact of the failure on the overall application business.

[0037] After troubleshooting is complete, big data analytics are used to integrate and analyze the faulty container's operational data, system environment data, and relevant event data before and after the failure to construct a fault cause analysis model. This model identifies the direct cause of the failure and analyzes potential indirect causes, providing a basis for subsequent system optimization. Fault logs are also recorded to create a fault archive, facilitating subsequent query and experience summary, improving the system's fault tolerance and self-healing capabilities in the face of various failures.

[0038] In resource demand forecasting scenarios, changes in resource usage and application traffic typically exhibit time-varying characteristics, representing typical time series data. RNNs and their variant, LSTMs, offer unique advantages in processing time series data. Traditional feedforward neural networks are unable to handle long-term dependencies within a sequence. However, RNNs, by introducing recurrent connections, enable information to flow across time, thereby modeling sequence data. LSTMs further address the vanishing and exploding gradient issues inherent in RNNs when learning long-term dependencies. Using a gating mechanism, LSTMs selectively memorize and forget information, effectively capturing trends in resource demand over long time spans. Compared to other similar time series forecasting algorithms, such as the autoregressive integrated moving average (ARIMA) model, RNNs and LSTMs do not require strict assumptions about data stationarity and can handle more complex and dynamic resource demand data.

[0039] When it comes to resource supply forecasting, cloud platform resource provisioning is influenced by a variety of complex factors, including hardware maintenance schedules, fluctuations in user demand, and cloud service providers' resource allocation strategies, resulting in complex data relationships. Random forest, a decision tree ensemble algorithm, is capable of handling high-dimensional, nonlinear data relationships. It improves prediction accuracy by constructing multiple decision trees and combining their predictions. Compared to a single decision tree, random forest effectively avoids overfitting and is more robust to noise and outliers in the data. Compared to other algorithms that handle complex data relationships, such as support vector machines (SVMs), random forests offer greater computational efficiency when processing large datasets. They eliminate the need for complex kernel function selection and parameter tuning, making them more suitable for the complex nature and large-scale data volumes of cloud platform resource provisioning data.

[0040] This solution uses an innovative approach combining static analysis with dynamic monitoring to deeply analyze applications. This not only preliminarily determines container dependencies and resource requirements from static perspectives like code structure and configuration files, but also dynamically monitors resource usage data during small-scale application runtimes to refine and improve the results, ensuring that the constructed application model accurately reflects actual conditions. The application model automatically adjusts the resource allocation ratio and topology between containers based on dynamic changes in business volume, flexibly adapting to diverse business scenarios and ensuring stable application operation under various circumstances. The entire solution utilizes a comprehensive fault tolerance mechanism with corresponding response measures for everything from interface failures to container failures, ensuring the system maintains a certain level of operational capability in the face of various faults and guaranteeing the continuous and stable operation of containerized applications. This is a crucial safeguard for system stability. The fault tolerance mechanism should be continuously verified and refined through system testing and fault simulation drills to ensure its effectiveness.

[0041] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A containerized operation and maintenance orchestration engine and dynamic resource scheduling method in a multi-cloud environment, characterized by: The following steps are involved: S1. Collect and integrate multi-cloud environment information, connect to different cloud platforms through the interface adapter module, and obtain resource information of each cloud platform; S2. Collect the service characteristics of each cloud platform and organize and store the service characteristic information in a unified format to build a multi-cloud environment resource information database; S3. Conduct analysis and modeling of containerized applications, and conduct in-depth analysis of containerized applications that need to be deployed; S4. Build an application model based on the analysis results; S5. Formulate dynamic resource scheduling strategies; S6. Use the containerized operation and maintenance orchestration engine to execute the scheduling strategy.

2. The containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment according to claim 1 is characterized by: The information obtained in step S1 includes the number of virtual machines, memory size, storage capacity, and network bandwidth. In step S3, the dependencies between containers within the application, data flow, and resource demand characteristics are analyzed.

3. The containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment according to claim 1 is characterized by: In the analysis and modeling steps of the containerized application, the application analysis tool adopts a method combining static analysis and dynamic monitoring.

4. The containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment according to claim 1, characterized in that: The containerized operation and maintenance orchestration engine adopts a distributed deployment architecture to execute scheduling strategies, deploying an engine node on each cloud platform, and information exchange between nodes is carried out through a secure and reliable communication protocol.

5. The containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment according to claim 1, characterized in that: The formulation of the dynamic resource scheduling strategy comprehensively considers the resource status in the multi-cloud environment resource information database, the service cost of each cloud platform and the resource demand in the application model, and uses an intelligent algorithm, which includes a hybrid algorithm combining a genetic algorithm and a particle swarm algorithm, to calculate the optimal resource allocation plan, determine which cloud platform and which resource each container should be deployed on, and how to dynamically adjust resource allocation according to real-time resource usage during application operation.

6. The containerized operation and maintenance orchestration engine and resource dynamic scheduling method in a multi-cloud environment according to claim 4, characterized in that: The containerized operation and maintenance orchestration engine is used to execute the scheduling strategy. The engine automatically creates container instances on the corresponding cloud platform according to the formulated resource allocation plan, configures network connections to ensure smooth communication between containers, and monitors the running status of containers in real time. Once abnormal resource usage or load changes are found, resources are immediately reallocated according to the dynamic adjustment strategy to ensure that containerized applications can run efficiently and stably in a multi-cloud environment.

Images