Automatic resource planning method, system and equipment for hybrid cloud and medium

Abstract

The invention discloses an automatic resource planning method, system and device for hybrid clouds and a medium, and the method specifically comprises the steps: determining a business change trend through employing a time sequence analysis algorithm, and obtaining a demand prediction model; obtaining dynamic resource demand parameters from the demand prediction model, judging a resource allocation proportion of the private cloud and the public cloud, and determining an initial configuration scheme of a cross-cloud resource pool; based on the initial configuration scheme, a reinforcement learning algorithm is adopted to optimize resource scheduling dynamics, and a real-time scheduling plan is obtained; specific parameters of resource allocation are extracted from the real-time scheduling plan, whether the requirements for safety and flexibility are met or not is judged, and a final scheduling instruction is determined; and based on the final scheduling instruction, generating a configuration script and an automatic deployment file for the cross-cloud environment by adopting a preset template engine. According to the method, the hybrid cloud resource delivery efficiency and the resource utilization rate are improved, and deepening and breakthrough of the hybrid cloud technology in enterprise application are promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of computer cloud platforms, and in particular to a hybrid cloud automated resource planning method, system, device and medium. Background Art

[0002] Cloud computing, a core pillar of modern enterprise information technology infrastructure, is driving digital transformation and significantly enhancing business agility with its unique appeal. Among the various cloud computing models, hybrid cloud is increasingly favored by enterprises due to its unique integration capabilities. It cleverly blends the security and controllability of private cloud with the flexibility and scalability of public cloud, providing strong support for enterprises to meet changing and complex business needs.

[0003] However, although hybrid cloud has many advantages in theory, in actual application, existing management solutions face many challenges, which prevents its potential from being fully unleashed.

[0004] Traditional hybrid cloud management solutions often rely on manual configuration and decentralized management strategies. This approach may be adequate when the business environment is relatively stable, but it becomes inadequate when business needs rapidly change. Resource scheduling is severely lacking in dynamism and automation, and manual operations are time-consuming, labor-intensive, and error-prone, resulting in long configuration cycles and high error rates. This not only severely impacts resource delivery efficiency but also makes it difficult for enterprises to quickly respond to market changes and seize business opportunities.

[0005] More specifically, traditional methods have significant shortcomings in predicting business load. Due to the difficulty in accurately predicting changing business load trends, resources cannot be quickly aligned with business needs. When business demand surges, resources may be insufficient, impacting service availability; whereas, when demand decreases, resources may become excessive, leading to waste or performance bottlenecks. This resource mismatch not only increases an enterprise's operating costs but also reduces resource utilization efficiency.

[0006] Furthermore, resource scheduling between private and public clouds lacks dynamism and flexibility. Due to the lack of a unified scheduling mechanism, resources cannot be efficiently allocated and utilized based on actual demand. This limits the elastic scalability of hybrid clouds and prevents them from fully leveraging cloud resources to meet peak business demands. During peak business periods, enterprises may face the risk of resource shortages; during low business periods, they may face the waste of idle resources.

[0007] Furthermore, the configuration and deployment process for hybrid cloud environments is cumbersome and error-prone. Due to the lack of unified automated configuration tools, enterprises must perform a significant amount of manual configuration and deployment work. This not only prolongs deployment cycles but also makes it difficult to ensure configuration consistency and correctness. Configuration errors can trigger a host of issues, increasing operational complexity and risk.

[0008] In summary, traditional hybrid cloud management solutions face significant technical bottlenecks in resource forecasting, resource scheduling, and configuration and deployment. These bottlenecks collectively lead to inefficient hybrid cloud resource delivery and limited optimization capabilities. This not only severely impacts the widespread adoption of hybrid cloud but also hinders enterprises' business agility and market competitiveness. Summary of the Invention

[0009] The purpose of the present invention is to provide a hybrid cloud automated resource planning method, system, equipment and medium. Through a systematic automated delivery solution, it bridges the shortcomings of existing methods, improves the efficiency and resource utilization of hybrid cloud resource delivery, promotes the deepening and breakthrough of hybrid cloud technology in enterprise applications, and solves at least one of the above-mentioned existing technical problems.

[0010] In a first aspect, the present invention provides a method for automated resource planning for a hybrid cloud, the method specifically comprising:

[0011] By collecting business load data and historical operation records in real time, using time series analysis algorithms to determine business change trends, we can obtain a demand forecasting model.

[0012] Obtain dynamic resource demand parameters from the demand forecast model, determine the resource allocation ratio between private cloud and public cloud, and determine the initial configuration plan for cross-cloud resource pools;

[0013] Based on the initial configuration plan, a reinforcement learning algorithm is used to optimize resource scheduling dynamics and obtain a real-time scheduling plan;

[0014] Extract specific resource allocation parameters from the real-time scheduling plan, determine whether security and flexibility requirements are met, and determine the final scheduling instructions;

[0015] Based on the final scheduling instructions, a preset template engine is used to generate configuration scripts and automated deployment files for cross-cloud environments.

[0016] In a second aspect, the present invention provides an automated resource planning system for a hybrid cloud, the system specifically comprising:

[0017] The first planning module is used to collect real-time business load data and historical operation records, use time series analysis algorithms to determine business change trends, and obtain a demand forecast model;

[0018] The second planning module is used to obtain dynamic resource demand parameters from the demand forecast model, determine the resource allocation ratio between private cloud and public cloud, and determine the initial configuration plan for cross-cloud resource pools;

[0019] The third planning module is used to optimize resource scheduling dynamics based on the initial configuration plan using reinforcement learning algorithms to obtain a real-time scheduling plan;

[0020] The fourth planning module is used to extract specific parameters of resource allocation from the real-time scheduling plan, determine whether the security and flexibility requirements are met, and determine the final scheduling instructions;

[0021] The fifth planning module is used to generate configuration scripts and automated deployment files for cross-cloud environments based on the final scheduling instructions using a preset template engine.

[0022] In a third aspect, the present invention provides a computer device comprising: a memory and a processor and a computer program stored in the memory, wherein when the computer program is executed on the processor, the method for automated resource planning of a hybrid cloud as described in any one of the above methods is implemented.

[0023] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for automated resource planning of a hybrid cloud as described in any one of the above methods is implemented.

[0024] Compared with the prior art, the present invention has at least one of the following technical effects:

[0025] 1. This invention bridges the gaps of existing methods through a systematic automated delivery solution, improves the efficiency of hybrid cloud resource delivery and resource utilization, and promotes the deepening and breakthrough of hybrid cloud technology in enterprise applications.

[0026] 2. The present invention improves the automation and accuracy of hybrid cloud resource scheduling by predicting business needs in real time and dynamically adjusting cloud resource allocation.

[0027] 3. By combining time series analysis and deep learning, the present invention improves the accuracy of business load forecasting and provides a reliable data basis for resource planning.

[0028] 4. By optimizing the objective function and heuristic rules, the present invention determines a reasonable cross-cloud resource allocation ratio, reduces costs and risks, and improves resource utilization efficiency.

[0029] 5. The present invention introduces carbon cost prediction and double-Q reinforcement learning to achieve environmental friendliness of resource scheduling and maximize long-term benefits.

[0030] 6. The present invention constructs a detailed carbon cost prediction function, which provides a quantitative basis for carbon emissions for real-time scheduling and promotes the development of green cloud computing.

[0031] 7. The present invention ensures the reliability and adaptability of scheduling instructions by comprehensively evaluating the security and flexibility of resource allocation parameters.

[0032] 8. The present invention uses a template engine and cross-cloud syntax conversion rules to automatically generate configuration scripts and deployment files, simplifying the configuration and deployment process across cloud environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 This is a flow chart of a hybrid cloud automated resource planning method provided by one embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of the structure of an automated resource planning system for a hybrid cloud provided by one embodiment of the present invention;

[0036] Figure 3 It is a structural diagram of a computer device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0037] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0038] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0039] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0040] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0041] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0042] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0043] In the embodiments of the present application, the execution subject of the process includes a terminal device, which includes but is not limited to: a server, a computer, a smart phone, a tablet computer, and other devices capable of executing the method disclosed in the present application. Figure 1 A flow chart of the method for automated resource planning for a hybrid cloud disclosed in the first embodiment of the present invention is shown, and is described in detail as follows:

[0044] S101, by collecting business load data and historical operation records in real time, using time series analysis algorithm to determine business change trends and obtain a demand forecast model.

[0045] In this embodiment, business load data is acquired from the business system via a sensor interface and stored in a database using real-time acquisition technology to obtain a load data set. The load data set and historical operation records are extracted from the database and merged into complete time series data to obtain a sequence data set. For the sequence data set, a time series analysis algorithm is used to decompose periodic fluctuations to obtain preliminary trend characteristics. The preliminary trend characteristics are fitted using a regression analysis method to determine the business change trend and obtain a trend function. If the slope of the trend function is greater than a preset threshold, the trend function is adjusted using an exponential smoothing algorithm to obtain an optimized trend function. The load value at a future time point is calculated based on the optimized trend function to generate a demand forecast result and obtain a forecast data set. By comparing the forecast data set with the historical operation records, the forecast deviation is determined and the model parameters are updated to obtain the final forecast model.

[0046] In this embodiment, based on accurate business load forecasting, enterprises can plan and adjust cloud resources in advance to ensure that sufficient computing, storage, and network resources are available during business peak periods.

[0047] S102, obtaining dynamic resource demand parameters from the demand forecasting model, determining the resource allocation ratio of the private cloud and the public cloud, and determining the initial configuration plan of the cross-cloud resource pool.

[0048] In this embodiment, business change characteristics are extracted through trend analysis, and a time series analysis algorithm is used to obtain the changing trend of dynamic resource demand. Based on the changing trend, demand parameters are obtained from the prediction model, and a support vector machine algorithm is used to determine the peak and valley values of resource demand. If the demand parameter exceeds the preset threshold, the allocation ratio of private cloud resources to public cloud resources is adjusted to a dynamic ratio to obtain a cross-cloud resource allocation plan. Through the cross-cloud resource allocation plan, the initial configuration plan is determined, and the load balancing tool is used to obtain the optimized layout of resource pool management. Resource pool management data is extracted from the optimized layout, the real-time load distribution of the private cloud and the public cloud is determined, and the resource scheduling priority is determined. Based on the resource scheduling priority, the dynamic adjustment parameters of the cross-cloud resource pool are obtained to determine whether to trigger the expansion or reduction operation of the resource pool. By dynamically adjusting the parameters, the initial configuration plan is updated to obtain the final allocation result of the cross-cloud resource pool.

[0049] In this embodiment, the demand forecasting model accurately predicts future business trends and resource requirements, providing a scientific basis for resource allocation between private and public clouds. This enables dynamic resource adjustment, flexible resource allocation based on real-time demand, and improved resource utilization efficiency. By comprehensively considering factors such as the cost structure and performance characteristics of private and public clouds, a reasonable resource allocation strategy can be formulated to reduce enterprise operating costs.

[0050] S103, based on the initial configuration plan, a reinforcement learning algorithm is used to optimize the resource scheduling dynamics to obtain a real-time scheduling plan.

[0051] In this embodiment, the current system resource usage (such as CPU utilization, memory usage, network bandwidth usage, etc.) and business load (such as user request volume and transaction success rate) are used as state inputs. A series of possible scheduling actions are defined, such as starting or stopping virtual machines on different cloud platforms, adjusting virtual machine configurations (such as the number of CPU cores and memory size), and migrating virtual machines. A reward function is designed to reflect the scheduling objectives, such as minimizing task response time, maximizing resource utilization, and reducing energy consumption. For example, the shorter the task response time, the higher the resource utilization, and the lower the energy consumption, the greater the reward value. A reinforcement learning algorithm suitable for resource scheduling problems is selected, such as the Deep Q-Network (DQN) or a policy gradient algorithm (such as REINFORCE or PPO), and the algorithm parameters, such as the neural network structure, learning rate, and discount factor, are initialized. The reinforcement learning model is trained using historical data and a simulation environment. By simulating different resource demand scenarios, the agent interacts with the environment and adjusts its policy based on the reward function feedback. During training, the agent continuously tries different action combinations, observes their impact on the system state and reward, and gradually learns the optimal scheduling policy. When new business load data arrives, the current system state is input into the trained reinforcement learning model. Based on the learned strategy, the model quickly outputs the optimal scheduling action for the current state, i.e., the real-time scheduling plan.

[0052] In this embodiment, the reinforcement learning algorithm dynamically adjusts resource allocation strategies based on real-time business load and system status, rapidly responding to changes in resource demand. Through continuous learning and optimization, the intelligent agent gradually learns the optimal scheduling strategy for different business scenarios. By optimizing resource scheduling, task response time is reduced, user experience is enhanced, resource utilization is improved, and resource waste and operating costs are reduced. The reinforcement learning algorithm automates resource scheduling, reduces manual intervention, and improves work efficiency.

[0053] S104, extracting specific parameters of resource allocation from the real-time scheduling plan, judging whether the security and flexibility requirements are met, and determining the final scheduling instructions.

[0054] In this embodiment, specific resource allocation parameters are extracted from the real-time scheduling plan generated by the reinforcement learning algorithm. These parameters include: the number of virtual machines to be started or stopped on the private cloud and public cloud, respectively; virtual machine configuration information, such as the number of CPU cores, memory size, and storage capacity; network bandwidth allocation, including upload and download bandwidth; and other parameters related to resource allocation, such as load balancing strategies and data migration plans.

[0055] Security requirements are determined by the following conditions: (1) Resource isolation and access control: Ensure that resources between different users or applications are isolated to prevent unauthorized access to resources. Implement strict access control policies, such as role-based access control (RBAC), to ensure that only authorized users can access specific resources. (2) Fault recovery and redundancy design: Check whether the scheduling plan includes fault recovery mechanisms, such as virtual machine backup, data redundant storage, etc. Ensure that when key components or resources fail, the system can quickly switch to backup resources to ensure service continuity. (3) Resource utilization monitoring: Monitor system resource utilization in real time to ensure that resource allocation does not cause system overload or resource idleness. Set resource utilization thresholds, and trigger an early warning mechanism when resource utilization approaches or exceeds the threshold.

[0056] Flexibility requirements are determined by the following criteria: (1) Elastic scalability: Evaluate whether the scheduling plan supports rapid adjustment of resource allocation based on real-time changes in business load. Ensure that the system can automatically expand or reduce resources to meet business needs of different time periods and scales. (2) Resource migration and reallocation: Check whether the scheduling plan includes resource migration strategies to quickly migrate virtual machines or data when needed. Ensure that the system can flexibly reallocate resources to respond to emergencies or optimize resource utilization efficiency. (3) Cross-cloud collaboration and integration: Verify whether the scheduling plan supports seamless collaboration and integration between private and public clouds. Ensure that cross-cloud resource pools can be uniformly managed, scheduled, and optimized to improve overall resource utilization and service quality.

[0057] If the resource allocation parameters meet the security and flexibility requirements, the final scheduling instructions are generated and sent to the corresponding resource management components for execution. If they do not meet the requirements, adjustments or optimizations are made according to the specific situation until they are met.

[0058] In this embodiment, security measures such as resource isolation, access control, fault recovery, and redundancy are implemented to ensure system stability and security. Unauthorized access, resource abuse, and malicious attacks are prevented, protecting user data and enterprise assets. Flexible resource management strategies such as elastic scaling, resource migration, and reallocation are supported, enabling the system to quickly respond to changes in business load. This improves resource utilization efficiency and service quality, while reducing operating costs.

[0059] S105 , based on the final scheduling instruction, a preset template engine is used to generate configuration scripts and automated deployment files for cross-cloud environments.

[0060] In this embodiment, key information is extracted from the final scheduling instructions output by the reinforcement learning algorithm, such as virtual machine specifications (CPU, memory, storage), network configuration, security group rules, API credentials of the cloud service provider, etc. This information will be used to fill in the variables in the template engine to generate configuration scripts and automated deployment files that meet the requirements of cross-cloud environments. For example, the Jinja2 template engine is chosen because it is based on Python, easy to integrate into the existing automation tool chain, and supports rich template syntax and logical control. The Jinja2 template engine allows the structure of configuration scripts and automated deployment files to be separated from dynamic content, improving the reusability and maintainability of the code.

[0061] Design and create Jinja2 template files for generating configuration scripts and automated deployment files. Define variables, conditionals, loop structures, and other structures in the template files to dynamically generate content based on the final scheduling instructions. For example, you can create a template file to generate a configuration script for an AWS EC2 instance, including variables such as the instance type, image ID, and security group ID.

[0062] Using the rendering capabilities of the Jinja2 template engine, key information from the final scheduling instructions is populated into template files, generating specific configuration scripts and automated deployment files. The rendering process can be automated by writing Python scripts that call Jinja2's API. The generated configuration scripts and automated deployment files contain detailed cross-cloud configuration information, such as network connection settings between private and public clouds, load balancer configuration, and database connection strings.

[0063] Use automation tools (such as Ansible and Terraform) to read the generated configuration scripts and automated deployment files to execute cross-cloud resource deployment and configuration. The automation tools will follow the instructions in the files to create virtual machines, configure networks, install software, and start services on private and public clouds. During deployment, the automation tools will monitor progress and provide error messages and retry mechanisms if any errors occur.

[0064] In this embodiment, by using a template engine to generate configuration scripts and automated deployment files, it is possible to quickly respond to business changes and achieve rapid deployment and configuration of resources. Automated deployment tools can reduce manual intervention, lower deployment error rates, and improve overall deployment efficiency. The template engine separates the structure of configuration scripts and automated deployment files from dynamic content, making the code more modular and easy to maintain. When the configuration or deployment process needs to be modified, only the template file or scheduling instructions need to be updated without modifying a large amount of script code. The template engine supports complex logical control and conditional judgment, and can generate different configuration scripts and automated deployment files based on different scheduling instructions. This flexibility enables companies to easily expand their hybrid cloud environments and support more types of applications and resources. By performing resource deployment and configuration across cloud environments through automated deployment tools, it can be ensured that each deployment follows a unified process and standard. Automation tools can also integrate security checks and compliance verification functions to ensure that the deployment process complies with the company's security policies and regulatory requirements.

[0065] In some embodiments, in step S101, the process of collecting business load data and historical operation records in real time and using a time series analysis algorithm to determine business change trends and obtain a demand forecasting model specifically includes:

[0066] Distributed probes deployed on private cloud nodes and public cloud nodes periodically collect business load data and time series data of the same dimension to form a time series training dataset. The time series training dataset includes CPU utilization sequence, memory usage sequence, network throughput sequence, and disk IOPS sequence.

[0067] Performing differential processing on the time series training data set, modeling the differentially processed time series training data set through ARIMA, and performing residual compensation through LSTM network to form a time series prediction model;

[0068] A correlation formula is set for converting the predicted value output by the time series prediction model into the number of virtual machine instances and the cross-cloud resource configuration vector, and the correlation formula is combined with the time series prediction model to form a demand prediction model.

[0069] In this embodiment, business load data is periodically collected from private cloud nodes and public cloud nodes through distributed probes to generate a time series training data set containing a CPU utilization sequence, a memory usage sequence, a network throughput sequence, and a disk IOPS sequence. Differential processing is applied to the time series training data set to obtain differential time series data. The ARIMA model is used to model the differential time series data to obtain preliminary prediction results. The residual of the ARIMA model is compensated by the LSTM network to obtain an optimized prediction model. The change trend of each sequence is extracted from the optimized prediction model to determine the resource usage status. If the resource usage status exceeds the preset threshold, the resource allocation of the private cloud node and the public cloud node is adjusted through the load balancing tool to obtain the adjusted resource configuration. According to the adjusted resource configuration, the collection parameters of the distributed probe are updated to obtain a new time series training data set.

[0070] The predicted value output by the time series prediction model is obtained. The number of virtual machine instances is calculated using a preset correlation formula to obtain a preliminary configuration result. Based on the preliminary configuration result, a linear regression algorithm is used to adjust the ratio of the number of instances to cloud resources to determine the optimized number of instances. The cross-cloud resource configuration vector is calculated based on the optimized number of instances to obtain a resource allocation plan. For the resource allocation plan, if the value of the configuration vector exceeds the preset threshold, the cloud resources are reallocated using the support vector machine algorithm to obtain balanced configuration data. The balanced configuration data is used, combined with the historical trend of the predicted value, to determine the need for dynamic adjustment of resource configuration. Based on the dynamic adjustment needs, the predicted value is updated using the time series prediction model to obtain a new resource configuration vector. The new resource configuration vector is used to determine the load balancing status of cloud resources and obtain the final resource scheduling plan.

[0071] Specifically, the association formula satisfies

[0072]

[0073] Among them, VM num Indicates the number of virtual machine instances. represents the predicted CPU utilization at time t, represents the predicted memory usage at time t, represents the predicted network throughput at time t, R threshold Indicates the resource carrying threshold of a single virtual machine. Indicates rounding up, Q t represents the cross-cloud resource configuration vector, Indicates the number of virtual machine instances in the private cloud. Indicates the number of virtual machine instances in the public cloud. Represents a multidimensional prediction matrix, S cost represents the cost constraint matrix, S securityrepresents the security policy matrix, and f(·) represents the multi-objective optimization decision function.

[0074] For example, in an enterprise data center, probes collect data every five minutes, generating a time series training dataset consisting of CPU utilization, memory usage, network throughput, and disk IOPS. Assume that the CPU utilization series might be a percentage of data for each minute over a continuous 24-hour period, such as 30%, 35%, and 40%, reflecting the temporal trend of system load. This collection method comprehensively captures business operational status and provides a reliable foundation for subsequent forecasting. In one possible implementation, differencing is performed on the time series training dataset to eliminate data non-stationarity and make it more suitable for modeling. For example, suppose the memory usage series is 500MB, 550MB, and 580MB. After differencing, it becomes 50MB and 30MB, highlighting the trend rather than the absolute value. ARIMA modeling is then used to capture the linear patterns in the data. For example, for a network throughput series, ARIMA can predict the throughput at the next moment based on the traffic data of the past hour (e.g., 100Mbps, 120Mbps, and 130Mbps). This method is particularly effective when processing time series data with high stationarity and can logically extract key trends. It should be noted that ARIMA modeling may not be able to fully fit the nonlinear residuals, so the LSTM network is introduced to compensate for the residuals. Specifically, assume that ARIMA predicts that the CPU utilization is 45%, while the actual value is 48% and the residual is 3%. LSTM further corrects the predicted value to 47.5% by learning historical residual patterns (such as the residual sequence of 2%, 3%, and 1% in the previous few minutes), thereby improving accuracy. This combination makes full use of the linear modeling capabilities of ARIMA and the nonlinear learning capabilities of LSTM, significantly improving prediction accuracy. Preferably, the output of the time series prediction model needs to be converted into the number of virtual machine instances and the cross-cloud resource configuration vector. For example, the prediction results show that the CPU utilization will reach 80% and the memory usage will be 2GB in the next hour. The associated formula can be used to calculate that 2 new virtual machines need to be added, each with 1 CPU core and 1GB of memory. In one embodiment, the formula can be set as If the utilization rate is 80%, then 2 units are needed. This conversion logic is clear and can directly guide resource allocation. In one possible implementation, the demand forecasting model integrates time series forecasting and correlation formulas. For example, during a promotion on an e-commerce platform, the probe collects a surge in the disk IOPS sequence (such as 1000, 1500, and 2000). The model predicts that the IOPS will reach 2500 in the next hour. Combined with the formula The conclusion is that three virtual machines are required. This approach not only predicts load but also dynamically adjusts resources to ensure service stability. As you can imagine, this model delivers technical benefits including improved resource utilization and cost optimization. For example, statically allocating five virtual machines without prediction would waste the resources of two. However, through dynamic adjustment using the model, only three can meet demand, saving approximately 40% of cloud resource costs. Furthermore, improved prediction accuracy reduces the risk of business interruption due to insufficient resources, logically forming a complete closed loop from data collection to resource allocation. For example, if a private cloud node experiences a sudden increase in load, the model can predict and preemptively migrate some tasks to the public cloud, smoothing cross-cloud resource scheduling. This multi-faceted approach ensures the rigor of the solution while demonstrating its practicality through specific scenarios, highlighting the value of technology in business continuity.

[0075] In some embodiments, in step S102 above, obtaining dynamic resource demand parameters from the demand forecasting model, determining the resource allocation ratio between the private cloud and the public cloud, and determining the initial configuration plan for the cross-cloud resource pool specifically include:

[0076] Extracting peak resource demand and baseline resource demand from the forecast value output by the demand forecast model;

[0077] Minimizing cost, latency, and risk is used as the optimization goal. The resource capacity constraints of the public and private clouds, the security isolation constraints between the public and private clouds, and the elastic buffer constraints between the public and private clouds are used as constraints to form the first optimization objective function.

[0078] Based on the peak resource demand, the baseline resource demand and the first optimization objective function, a mixed integer programming algorithm is used to determine the virtual machine allocation ratio between the public cloud and the private cloud, and bandwidth and storage allocation are optimized through heuristic rules to form an initial configuration plan for the cross-cloud resource pool.

[0079] In this embodiment, the demand forecasting model outputs predicted values, extracts peak resources and baseline resources, and uses statistical analysis to determine the resource demand range. Based on the peak resources and baseline resources, combined with resource capacity constraints, a mixed integer programming algorithm is applied to obtain the virtual machine allocation ratio between the public cloud and the private cloud. Security isolation constraints are obtained for the virtual machine allocation ratio. If the isolation requirements are met through logical judgment, the allocation ratio is adjusted to determine a preliminary cross-cloud allocation plan. Bandwidth requirements are extracted from the preliminary cross-cloud allocation plan, and heuristic rules are used to optimize bandwidth allocation to obtain bandwidth optimization results. Based on the bandwidth optimization results and baseline resources, combined with elastic buffer constraints, storage allocation is adjusted using heuristic rules to determine the storage allocation plan. The optimization objectives are integrated by using the virtual machine allocation ratio, bandwidth optimization results, and storage allocation plan to generate an initial configuration plan for the cross-cloud resource pool. Resource usage data is extracted from the initial configuration plan, and a linear regression algorithm is used to analyze the stability of the configuration plan to determine the final configuration plan.

[0080] For example, suppose an enterprise's private cloud node is predicted to have a CPU utilization of 90% and a memory requirement of 3GB during peak business hours, while the baseline requirement during off-peak hours is 30% CPU utilization and 1GB of memory. This extraction method, by analyzing historical time series data, clearly defines upper and lower limits for resource usage, laying the foundation for subsequent optimization. In one possible implementation, minimizing cost, latency, and risk is the optimization objective, reflecting a multi-dimensional balance approach. Specifically, cost refers to the cost of cloud resource rental, latency focuses on task response speed, and risk involves service interruptions caused by insufficient resources. Among the constraints, the resource capacity constraints for public and private clouds can be understood as follows: the private cloud has a fixed capacity of 10 CPU cores and 20GB of memory, while the public cloud can be elastically scaled to 50 cores and 100GB of memory. The security isolation constraint requires that sensitive data be processed only in the private cloud, while the elastic buffer constraint ensures that the public cloud reserves 20% of spare capacity. For example, an enterprise can achieve both isolation and elasticity by prioritizing the deployment of core databases in the private cloud and running front-end services in the public cloud. It should be noted that the construction of the first optimization objective function quantifies the above objectives and constraints into a computable model. Preferably, the cost can be defined as 0.5 yuan per core per hour, and the latency as a risk factor of 0.1 per millisecond, forming a function through weighted combination. For example, a solution with a cost of 5 yuan, a latency of 10 milliseconds, and a risk probability of 0.2 aims to minimize the overall score. This approach is logically clear and facilitates subsequent algorithmic solution. Using a mixed integer programming algorithm to determine the virtual machine allocation ratio based on peak resource demand, baseline resource demand, and the optimization objective function is a rigorous approach. In one embodiment, if peak demand requires 15 CPU cores, the baseline requires 5, and the private cloud has a capacity of 10 cores, the algorithm can determine a ratio of 10 to the private cloud and 5 to the public cloud. Next, bandwidth and storage allocation are optimized using heuristic rules. For example, bandwidth is prioritized for low-latency front-end tasks, while storage is preferred to the private cloud to meet security requirements. It is understood that the heuristic rule can be set to "bandwidth allocation ratio is inversely proportional to latency requirement." For example, if the front-end task requires a 1 millisecond response time, 2 Gbps of bandwidth is allocated. This method flexibly adjusts resource distribution to ensure that the initial configuration plan is efficient. Specifically, the initial configuration plan for cross-cloud resource pools needs to take into account the actual scenario. For example, during business peak periods, the memory demand is predicted to be 3GB, the private cloud provides 2GB, and the public cloud supplements 1GB. At the same time, the bandwidth is configured to 1Gbps to support data transmission. This configuration not only meets peak demand, but also avoids idle resources through baseline optimization. In one possible implementation, if the baseline demand only requires 1GB of memory, the solution can dynamically reduce the public cloud allocation, reflecting the advantage of elasticity. This multi-faceted support solution not only ensures resource adequacy, but also improves overall efficiency through fine allocation.

[0081] In some embodiments, in step S103, optimizing resource scheduling dynamics based on the initial configuration scheme using a reinforcement learning algorithm to obtain a real-time scheduling plan specifically includes:

[0082] Setting a carbon cost prediction function, wherein the carbon cost prediction function is used to predict the carbon cost generated by the private cloud and the public cloud under different resource allocation conditions;

[0083] Resource utilization, queue status, and carbon cost are set as the state space, elastic scaling of private or public clouds, cross-cloud traffic migration, and priority task quotas are set as the action space, resource efficiency is set as the immediate reward, and carbon efficiency is set as the long-term reward;

[0084] Based on the initial configuration scheme, state space, action space, immediate reward and long-term reward, the double-Q reinforcement learning algorithm is used to dynamically optimize the resource scheduling between private cloud and public cloud to obtain a real-time scheduling plan.

[0085] In this embodiment, resource utilization and queue status data are obtained, and the initial value of the state space is determined by collecting runtime information of the private cloud and the public cloud. Resource utilization is extracted from the initial value of the state space, and the immediate reward value is obtained by calculating the current resource efficiency. Based on the queue status and task quota, a priority sorting algorithm is used to determine the task allocation plan, and the adjusted task quota data is output. Based on the adjusted task quota data and cross-cloud migration properties, a linear regression algorithm is used to predict the impact of cross-cloud traffic migration on carbon efficiency, and a migration optimization plan is obtained. The migration optimization plan and resource efficiency data are obtained, the action combination of private cloud scaling and public cloud scaling is determined, and scaling adjustment instructions are output. The action space update value is extracted from the scaling adjustment instruction, and the carbon cost change trend is calculated by the prediction function to obtain the long-term carbon efficiency optimization result. According to the carbon cost change trend and the state space update value, the action space parameters are adjusted using a reinforcement learning algorithm, and the final resource allocation plan is output. Specifically, the Q-value table is iteratively updated through the dual-Q learning algorithm, the resource scheduling strategy is dynamically optimized, and short-term and long-term benefits are balanced. The current cloud environment status, including the computing, storage, and network resource usage of private and public clouds, is obtained. The Q-value table is queried based on the current status, and the optimal action is selected for resource scheduling to achieve dynamic load balancing between private and public clouds. The changes in system performance indicators after resource scheduling are monitored, the reward value is calculated and the Q-value table is updated. The scheduling strategy is continuously optimized for sudden business needs or abnormal situations, triggering the emergency scheduling mechanism to quickly adjust resource allocation to meet business needs.

[0086] For example, a carbon cost prediction function can consider factors such as energy mix and equipment efficiency to predict the carbon emissions of different resource allocation scenarios. In one possible implementation, a private cloud uses 80% clean energy and emits 0.2 kg of CO2 per kilowatt-hour, while a public cloud has a poorer energy mix, emitting 0.5 kg per kilowatt-hour. This prediction allows enterprises to balance performance with environmental requirements. It's important to note that the design of the state space directly impacts decision accuracy. Specifically, resource utilization represents computing load, queue status reflects the level of task backlog, and carbon cost quantifies environmental impact. For example, at a certain moment, the private cloud's CPU utilization is 70%, memory usage is 2 GB, the task queue length is 5, and carbon emissions are expected to be 10 kg. These metrics together constitute the current state vector. In the action space, elastic scaling determines resource scalability, cross-cloud migration influences load distribution, and priority quotas regulate task order. In one possible implementation, a set of actions could include expanding the private cloud's CPU capacity by two cores, migrating data analysis tasks to the public cloud, and increasing the quota for critical services to 50%. This multi-dimensional control provides a rich set of options for resource scheduling. Preferably, resource efficiency is assigned an immediate reward, while carbon efficiency is assigned a long-term reward, balancing short-term performance with sustainable development. For example, an action that increases task processing speed by 20% earns a positive immediate reward; at the same time, it is expected to reduce annual carbon emissions by 5%, earning a long-term bonus. This design guides the system to prioritize environmental impact while improving performance. Understandably, the Dual-Q reinforcement learning algorithm gradually optimizes its decision-making strategy through continuous trial and evaluation. In practical applications, the algorithm might first attempt to migrate compute-intensive tasks to a private cloud, observing improved resource utilization and reduced carbon emissions, and then reinforce this strategy. Through continuous learning and adjustment, the system ultimately develops a dynamic scheduling solution that balances efficiency and environmental protection.

[0087] Furthermore, the setting of the carbon cost prediction function specifically includes:

[0088] Based on the real-time server power consumption of the private cloud and the regional carbon intensity and virtual machine instance power consumption of various energy structures in the power grid and the public cloud, a carbon cost prediction function is constructed. The carbon cost prediction function satisfies

[0089] C total =α·C private +(1-α)·C public

[0090]

[0091] Among them, C total represents the carbon cost of the carbon cost prediction function, C private represents the carbon cost of private cloud, C publicrepresents the carbon cost of the public cloud, α represents the dynamic weight of the private cloud load ratio, t represents the total time step, ΔT represents the integration time window, The formula for real-time power consumption of private cloud servers is: express The proportion of the i-th type of energy in the power grid at the moment, F i represents the carbon emission factor of energy source i, The formula for the power consumption of a public cloud virtual machine instance is: express The real-time carbon intensity of the region where the public cloud is located at the moment, δ j represents the running time of the jth virtual machine instance in the time window ΔT, and J represents the total number of virtual machines in the public cloud cluster;

[0092]

[0093] in, represents the basic power consumption of the kth server in idle state, represents the peak power consumption of the kth server at full load, express The CPU utilization of the kth server at the moment, K represents the total number of servers in the private cloud cluster, η j represents the energy efficiency correction coefficient of the j-th virtual machine instance, express The CPU usage of the jth virtual machine instance at time, Indicates the maximum CPU quota available for any virtual machine instance. Indicates the thermal design power consumption of the physical CPU corresponding to any virtual machine instance;

[0094] By performing trend term modeling, period term extraction and residual term learning on the carbon cost output by the carbon cost prediction function, the intensity trend of the carbon cost in a future preset time period is obtained.

[0095] In this embodiment, the dynamic weighting of the private cloud load contribution allows the relative importance of private and public cloud carbon costs to be adjusted based on actual conditions. This flexibility enables the prediction function to adapt to different usage scenarios and strategies, such as optimizing cloud service configuration under different time periods or load conditions.

[0096] The total time step is the total time of the forecast, and the integration time window is the time interval when the carbon cost is calculated. These two parameters define the time range and accuracy of the forecast, allowing users to adjust the granularity and scope of the forecast as needed.

[0097] The proportion of type i energy in the power grid and the carbon emission factor of energy i describe the proportion of each type of energy in the power grid and their carbon emission characteristics.

[0098] The real-time carbon intensity of the region where the public cloud is located reflects the carbon emission intensity of electricity production in the region where the public cloud is located.

[0099] By modeling the carbon cost output from the carbon cost prediction function through trend terms, extracting periodic terms, and learning residual terms, this technical solution can predict carbon cost intensity trends over a preset time period. This predictive capability provides decision support for businesses and organizations in developing long-term emission reduction strategies and responding to short-term fluctuations.

[0100] Detailed calculation of public cloud virtual machine instance power consumption, including energy efficiency correction factors, CPU utilization, and physical CPU thermal design power, enables the prediction function to accurately assess the impact of public cloud usage on carbon costs. This helps optimize the use of public cloud resources and improve energy efficiency.

[0101] By comprehensively considering the real-time power consumption of private cloud servers, the share of various energy sources in the power grid and their carbon emission factors, as well as the power consumption and regional carbon intensity of public cloud virtual machine instances, this carbon cost prediction function accurately calculates the total carbon cost within a given time window. This precise prediction provides businesses and organizations with a quantitative metric to assess the environmental impact of their cloud service usage. By incorporating the share of various energy sources in the power grid and their carbon emission factors, the prediction function reflects the impact of the energy mix on carbon costs. This helps businesses and organizations choose low-carbon energy sources, thereby reducing their carbon footprint.

[0102] In some embodiments, in step S104, extracting specific parameters of resource allocation from the real-time scheduling plan, determining whether security and flexibility requirements are met, and determining the final scheduling instructions specifically include:

[0103] Extracting resource allocation parameters from the real-time scheduling plan, the resource allocation parameters including a resource type code, a cloud platform identifier, a geographic region, and a resource parameter vector, the resource parameter vector including the number of virtual machine instances, the number of CPU cores, memory capacity, network bandwidth, and storage capacity;

[0104] Determine the security assessment of the resource allocation parameters by comprehensively analyzing the data sensitivity, encryption brightness, compliance, number of role access control levels, and audit log retention days of the resource allocation parameters;

[0105] Determining the flexibility evaluation degree of the resource allocation parameter by comprehensively analyzing the resource expansion and contraction delay, the ratio of multiple cloud providers available, and the proportion of idle resource pool capacity of the resource allocation parameter;

[0106] The security assessment degree and the flexibility assessment degree are compared with respective preset thresholds to determine whether the resource allocation parameters meet the security and flexibility requirements. If so, the resource allocation parameters are determined as the final scheduling instructions.

[0107] In this embodiment, an initial dataset for resource allocation is obtained by extracting parameters such as resource type, cloud platform, geographic region, and number of virtual machines from the real-time scheduling plan. A random forest algorithm is used to analyze the characteristics of the number of virtual machines, CPU cores, memory capacity, network bandwidth, and storage capacity in the initial dataset to obtain the distribution characteristics of the resource parameter vector. If the correlation between the number of virtual machines and the number of CPU cores in the distribution characteristics exceeds a preset threshold, the allocation ratio of cloud platform and geographic region is adjusted to determine the optimized resource parameter vector.

[0108] Based on the optimized resource parameter vector, the match between data sensitivity and encryption strength is analyzed to obtain a preliminary security configuration assessment. By comparing this preliminary assessment result with compliance and access control requirements, any security configuration deviations are determined and the adjusted security configuration parameters are obtained. The final security assessment is determined based on the adjusted security configuration parameters and the number of days audit logs are retained.

[0109] By collecting real-time data on resource scaling, a baseline value for latency analysis is determined. Based on the latency analysis results, availability data for multiple cloud providers is obtained to determine the distribution of available ratios. Based on the distribution of available ratios, a preset threshold is used to determine the allocation priority of idle resources. Based on the allocation priority of idle resources, the dynamic changes in pool capacity percentage are analyzed to obtain preliminary results of the resource pool analysis. If the results of the resource pool analysis exceed the preset range, an optimization solution for the comprehensive analysis is determined by adjusting the call ratio of multiple cloud providers. Based on the optimization solution from the comprehensive analysis, a random forest algorithm is used to determine the final value of the flexibility evaluation degree. Using the final value of the flexibility evaluation degree, updated data for parameter determination is obtained to determine an adjustment plan for the resource allocation parameters.

[0110] For example, when extracting resource allocation parameters from a real-time scheduling plan, they can be viewed as a multidimensional information set. For example, the resource type code might be "VM-001" to represent a virtual machine type, the cloud platform identifier "PC-01" to refer to a private cloud, and the geographic region set to "East Asia - Tokyo." The resource parameter vector specifically includes 5 virtual machine instances, 8 CPU cores, 32GB of memory, 500Mbps of network bandwidth, and 1TB of storage. These parameters together outline the basic outline of resource allocation, laying the foundation for subsequent analysis. In one possible implementation, the security assessment of resource allocation parameters can be based on data sensitivity. For example, if a task involves user privacy data and is rated as high in sensitivity, encryption strength of AES-256 is required. Regarding compliance, if regional regulations require that data not be exported, a private cloud in the Tokyo region naturally meets this requirement. The role-based access control hierarchy is set to three, restricting access to specific administrators only, and audit log retention is set to 180 days to meet audit requirements. A comprehensive evaluation of these factors might yield a security score of 85, exceeding the preset threshold of 75 and indicating that security meets the standard. It should be noted that the determination of flexibility focuses on the dynamic nature of resource scheduling. Specifically, a resource scaling delay of 2 seconds indicates a responsive system. A 60% availability of multi-cloud providers indicates that scheduling can flexibly switch to other cloud platforms, such as public clouds. A 30% idle resource pool capacity indicates sufficient reserved resources to support sudden demand. After comprehensive analysis, the flexibility score might reach 80, exceeding the threshold of 70, demonstrating strong adaptability. In one possible implementation, the comparison between security and flexibility scores can be further refined. For example, in a scenario where the security score is 90 but the flexibility score is only 65, failing to meet the standard, parameters can be adjusted, such as reducing the number of virtual machine instances to 3 to reduce load and increasing the idle resource percentage to 40%, thereby raising the flexibility score to 72, meeting the requirement. This adjustment demonstrates the practicality of parameter optimization and helps the system maintain flexible scheduling capabilities while maintaining high security. Preferably, the determination of the final scheduling instruction can also consider the balance between parameters. For example, increasing the number of CPU cores to 10 may improve performance, but if the energy supply in the geographical area is tight, it will lead to increased costs. On the contrary, keeping the number of instances at 5 and optimizing the network bandwidth to 600Mbps can both meet the task requirements and avoid resource waste. This trade-off ensures the efficiency of the instructions while saving operating costs for the enterprise. In one possible implementation, assuming that an enterprise needs to process real-time data analysis tasks, the resource allocation parameters are initially determined to be 4 virtual machine instances, 6 CPU cores, and 16GB of memory. If the security assessment is 88 and the flexibility assessment is 78, both exceeding the threshold, they are directly solidified as scheduling instructions. This method not only ensures the security of data processing, but also can quickly respond to traffic fluctuations, reflecting the practical value of the scheduling solution.It's understandable that the above analysis process, through multi-faceted scrutiny, forms a rigorous logic for parameter determination. For example, security assessments are layered, examining everything from sensitivity to compliance, while flexibility assessments comprehensively consider everything from latency to resource pools. These two factors mutually support each other, ultimately ensuring that instructions are both secure and flexible. This approach effectively improves resource scheduling reliability and provides enterprises with stable and sustainable cloud resource management support.

[0111] In some embodiments, in step S105 above, generating a configuration script and automated deployment file for a cross-cloud environment using a preset template engine based on the final scheduling instruction specifically includes:

[0112] Associate and map resource allocation parameters with configuration parameters and deployment parameters to form a template engine;

[0113] Parsing the template engine using context-free grammar to construct a template parsing tree, and inputting the parameters of the final scheduling instruction into the template parsing tree to obtain an initial configuration deployment template;

[0114] A cross-cloud syntax conversion rule library is set up, the initial configuration deployment template is converted according to the cross-cloud syntax conversion rule library, and semantic verification and permission verification are performed to generate a configuration script and automated deployment file for a cross-cloud environment.

[0115] In this embodiment, a template engine is generated by establishing an association mapping between resource allocation parameters and configuration parameters; the template engine is parsed using a context-free grammar to construct a parsing tree and obtain tree structure data. Parameter input is obtained through a scheduling instruction, and the input processing result is extracted from the tree structure data. If the input processing result meets the preset conditions, the initial configuration content is determined using a logical construction method. Configuration generation data is generated based on the initial configuration content, and the configuration generation data is adjusted using a tree structure. For the adjusted configuration generation data, a deployment template framework is obtained through deployment template mapping. Logical construction features are extracted from the deployment template framework to determine whether the features comply with context-free rules. If the features comply with context-free rules, the deployment template framework is updated through parameter input to obtain the final template data.

[0116] By analyzing cross-cloud environment requirements, we obtain syntax rules and conversion rules and build a rule base. Based on the rule base, we convert the initial configuration deployment template to obtain the conversion result. We perform semantic verification on the conversion result to determine content consistency and generate a template that passes the verification. We perform permission verification on the template that passes the verification to determine access control compliance and obtain the permission verification result. Based on the permission verification result, we generate a configuration script and use a script generation tool to process it into a cross-cloud compatible format. We generate deployment files from the configuration script to obtain the complete set of files required for automated deployment. We load the deployment file in the cross-cloud environment, execute automated deployment, and complete the environment adaptation.

[0117] For example, the number of virtual machine instances can be mapped to "instance_count," the number of CPU cores to "cpu_cores," and the memory capacity to "memory_size." This mapping creates a template engine that can flexibly adapt to the parameter requirements of different cloud platforms. In one possible implementation, when parsing the template engine using a context-free grammar, production rules such as S→Resource|Config|Deploy, Resource→TypeCount, and Config→ParamValue can be defined. The template parse tree constructed using these rules effectively expresses the hierarchical structure of resource configurations. It should be noted that during the generation of the initial configuration deployment template, scheduling instruction parameters such as "VM-002,3,8 cores,16GB" can be input into the parse tree. After syntactic analysis and semantic processing, structured template content is generated. This approach improves the accuracy and efficiency of template generation. Specifically, the cross-cloud syntax conversion rule library can include proprietary syntax comparison tables for different cloud platforms. For example, Alibaba Cloud's instance type "ecs.g6.large" might correspond to AWS's "t3.medium." Through this rule conversion, the initial template can adapt to multiple cloud environments, enhancing deployment compatibility. In one embodiment, semantic verification can check the logical relationship between parameters, such as ensuring that the number of CPU cores matches the memory capacity. Permission verification can check whether the operator has the necessary permissions for cross-cloud deployment, such as checking the validity of the API key. These verification steps significantly improve the reliability of the configuration script. Preferably, the generated configuration script can be in YAML format to facilitate human-computer interaction. The automated deployment file can use the template language of tools such as Terraform, support declarative infrastructure as code, and simplify the resource management process across cloud environments. It can be understood that this method uses technologies such as template engines, syntax parsing, and rule conversion to achieve efficient conversion from abstract resource parameters to specific deployment instructions, providing enterprises with a flexible and standardized cross-cloud resource scheduling solution.

[0118] Reference Figure 2 An embodiment of the present invention provides a hybrid cloud automated resource planning system 2, wherein the system 2 specifically includes:

[0119] The first planning module 201 is used to collect business load data and historical operation records in real time, use time series analysis algorithms to determine business change trends, and obtain a demand forecast model;

[0120] The second planning module 202 is used to obtain dynamic resource demand parameters from the demand forecast model, determine the resource allocation ratio between the private cloud and the public cloud, and determine the initial configuration plan of the cross-cloud resource pool;

[0121] The third planning module 203 is used to optimize resource scheduling dynamics based on the initial configuration plan using a reinforcement learning algorithm to obtain a real-time scheduling plan;

[0122] The fourth planning module 204 is used to extract specific parameters of resource allocation from the real-time scheduling plan, determine whether the security and flexibility requirements are met, and determine the final scheduling instructions;

[0123] The fifth planning module 205 is used to generate configuration scripts and automated deployment files for cross-cloud environments using a preset template engine based on the final scheduling instruction.

[0124] It is understandable that if Figure 1 The contents of the embodiment of the hybrid cloud automated resource planning method shown in FIG. 1 are applicable to the embodiment of the hybrid cloud automated resource planning system. The functions specifically implemented by the embodiment of the hybrid cloud automated resource planning system are similar to those in FIG. Figure 1 The embodiment of the hybrid cloud automated resource planning method shown is the same as that shown in FIG. Figure 1 The beneficial effects achieved by the embodiment of the hybrid cloud automated resource planning method shown are also the same.

[0125] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0126] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0127] Reference Figure 3An embodiment of the present invention further provides a computer device 3, comprising: a memory 302, a processor 301, and a computer program 303 stored on the memory 302. When the computer program 303 is executed on the processor 301, the automatic resource planning method for the hybrid cloud as described in any one of the above methods is implemented.

[0128] The computer device 3 may be a desktop computer, a notebook computer, a PDA, a cloud server or other computing devices. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that Figure 3 This is merely an example of the computer device 3 and does not constitute a limitation on the computer device 3 . The computer device 3 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 3 may also include input and output devices, network access devices, etc.

[0129] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0130] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device 3. Furthermore, the memory 302 may include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 302 may also be used to temporarily store data that has been output or is about to be output.

[0131] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for automated resource planning of a hybrid cloud as described in any one of the above methods is implemented.

[0132] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process of the above-mentioned method embodiment by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can at least include: any entity or device capable of carrying computer program code to the camera / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. For example, a USB flash drive, mobile hard drive, magnetic disk, or optical disk. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0133] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0134] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0135] In the embodiments disclosed in this application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0136] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

Claims

1. A hybrid cloud automated resource planning method, characterized in that: The method specifically includes: By collecting business load data and historical operation records in real time, using time series analysis algorithms to determine business change trends, we can obtain a demand forecasting model. Obtain dynamic resource demand parameters from the demand forecast model, determine the resource allocation ratio between private cloud and public cloud, and determine the initial configuration plan for cross-cloud resource pools; Based on the initial configuration plan, a reinforcement learning algorithm is used to optimize resource scheduling dynamics and obtain a real-time scheduling plan; Extract specific resource allocation parameters from the real-time scheduling plan, determine whether security and flexibility requirements are met, and determine the final scheduling instructions; Based on the final scheduling instructions, a preset template engine is used to generate configuration scripts and automated deployment files for cross-cloud environments.

2. The method according to claim 1, characterized in that The method involves collecting business load data and historical operation records in real time, using a time series analysis algorithm to determine business change trends, and obtaining a demand forecasting model, specifically including: Distributed probes deployed on private cloud nodes and public cloud nodes periodically collect business load data and time series data of the same dimension to form a time series training dataset. The time series training dataset includes CPU utilization sequence, memory usage sequence, network throughput sequence, and disk IOPS sequence. Performing differential processing on the time series training data set, modeling the differentially processed time series training data set through ARIMA, and performing residual compensation through LSTM network to form a time series prediction model; A correlation formula is set for converting the predicted value output by the time series prediction model into the number of virtual machine instances and the cross-cloud resource configuration vector, and the correlation formula is combined with the time series prediction model to form a demand prediction model.

3. The method according to claim 1, characterized in that The process of obtaining dynamic resource demand parameters from the demand forecasting model, determining the resource allocation ratio between the private cloud and the public cloud, and determining the initial configuration plan for the cross-cloud resource pool specifically includes: Extracting peak resource demand and baseline resource demand from the forecast value output by the demand forecast model; Minimizing cost, latency, and risk is used as the optimization goal. The resource capacity constraints of the public and private clouds, the security isolation constraints between the public and private clouds, and the elastic buffer constraints between the public and private clouds are used as constraints to form the first optimization objective function. Based on the peak resource demand, the baseline resource demand and the first optimization objective function, a mixed integer programming algorithm is used to determine the virtual machine allocation ratio between the public cloud and the private cloud, and bandwidth and storage allocation are optimized through heuristic rules to form an initial configuration plan for the cross-cloud resource pool.

4. The method according to claim 1, wherein Based on the initial configuration scheme, the reinforcement learning algorithm is used to optimize the resource scheduling dynamics and obtain a real-time scheduling plan, specifically including: Setting a carbon cost prediction function, wherein the carbon cost prediction function is used to predict the carbon cost generated by the private cloud and the public cloud under different resource allocation conditions; Resource utilization, queue status, and carbon cost are set as the state space, elastic scaling of private or public clouds, cross-cloud traffic migration, and priority task quotas are set as the action space, resource efficiency is set as the immediate reward, and carbon efficiency is set as the long-term reward; Based on the initial configuration scheme, state space, action space, immediate reward and long-term reward, the double-Q reinforcement learning algorithm is used to dynamically optimize the resource scheduling between private cloud and public cloud to obtain a real-time scheduling plan.

5. The method according to claim 4, characterized in that The setting of the carbon cost prediction function specifically includes: Based on the real-time server power consumption of the private cloud and the regional carbon intensity and virtual machine instance power consumption of various energy structures in the power grid and the public cloud, a carbon cost prediction function is constructed. The carbon cost prediction function satisfies C total =α·C private +(1-α)·C public Among them, C total represents the carbon cost of the carbon cost prediction function, C private represents the carbon cost of private cloud, C public represents the carbon cost of the public cloud, α represents the dynamic weight of the private cloud load ratio, t represents the total time step, ΔT represents the integration time window, The formula for real-time power consumption of private cloud servers is: express The proportion of the i-th type of energy in the power grid at the moment, F i represents the carbon emission factor of energy source i, The formula for the power consumption of a public cloud virtual machine instance is: express The real-time carbon intensity of the region where the public cloud is located at the moment, δ j represents the running time of the jth virtual machine instance in the time window ΔT, and J represents the total number of virtual machines in the public cloud cluster; in, represents the basic power consumption of the kth server in idle state, represents the peak power consumption of the kth server at full load, represents the CPU utilization of the kth server at time τ, K represents the total number of servers in the private cloud cluster, and η j represents the energy efficiency correction coefficient of the j-th virtual machine instance, express The CPU usage of the jth virtual machine instance at time, Indicates the maximum CPU quota available for any virtual machine instance. Indicates the thermal design power consumption of the physical CPU corresponding to any virtual machine instance; By performing trend term modeling, period term extraction and residual term learning on the carbon cost output by the carbon cost prediction function, the intensity trend of the carbon cost in a future preset time period is obtained.

6. The method according to claim 1, characterized in that The process of extracting specific resource allocation parameters from the real-time scheduling plan, determining whether security and flexibility requirements are met, and determining the final scheduling instructions specifically includes: Extracting resource allocation parameters from the real-time scheduling plan, the resource allocation parameters including a resource type code, a cloud platform identifier, a geographic region, and a resource parameter vector, the resource parameter vector including the number of virtual machine instances, the number of CPU cores, memory capacity, network bandwidth, and storage capacity; Determine the security assessment of the resource allocation parameters by comprehensively analyzing the data sensitivity, encryption brightness, compliance, number of role access control levels, and audit log retention days of the resource allocation parameters; Determining the flexibility evaluation degree of the resource allocation parameter by comprehensively analyzing the resource expansion and contraction delay, the ratio of multiple cloud providers available, and the proportion of idle resource pool capacity of the resource allocation parameter; The security assessment degree and the flexibility assessment degree are compared with respective preset thresholds to determine whether the resource allocation parameters meet the security and flexibility requirements. If so, the resource allocation parameters are determined as the final scheduling instructions.

7. The method according to claim 6, characterized in that Based on the final scheduling instructions, a preset template engine is used to generate configuration scripts and automated deployment files for cross-cloud environments, specifically including: Associate and map resource allocation parameters with configuration parameters and deployment parameters to form a template engine; Parsing the template engine using context-free grammar to construct a template parsing tree, and inputting the parameters of the final scheduling instruction into the template parsing tree to obtain an initial configuration deployment template; A cross-cloud syntax conversion rule library is set up, the initial configuration deployment template is converted according to the cross-cloud syntax conversion rule library, and semantic verification and permission verification are performed to generate a configuration script and automated deployment file for a cross-cloud environment.

8. An automated resource planning system for hybrid cloud, characterized in that: The system specifically includes: The first planning module is used to collect real-time business load data and historical operation records, use time series analysis algorithms to determine business change trends, and obtain a demand forecast model; The second planning module is used to obtain dynamic resource demand parameters from the demand forecast model, determine the resource allocation ratio between private cloud and public cloud, and determine the initial configuration plan for cross-cloud resource pools; The third planning module is used to optimize resource scheduling dynamics based on the initial configuration plan using reinforcement learning algorithms to obtain a real-time scheduling plan; The fourth planning module is used to extract specific parameters of resource allocation from the real-time scheduling plan, determine whether the security and flexibility requirements are met, and determine the final scheduling instructions; The fifth planning module is used to generate configuration scripts and automated deployment files for cross-cloud environments based on the final scheduling instructions using a preset template engine.

9. A computer device, characterized in that: include: A memory, a processor, and a computer program stored in the memory, which, when executed on the processor, implements the method for automatic resource planning of a hybrid cloud according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the method for automatic resource planning of a hybrid cloud according to any one of claims 1 to 7 is implemented.

Images