Enterprise-level cloud computing resource dynamic allocation and management system and implementation method thereof

Abstract

The invention discloses an enterprise-level cloud computing resource dynamic allocation and management system and an implementation method thereof, and relates to the technical field of computing resource management. According to the invention, cloud resource use data is collected and processed in real time through the resource state sensing and preprocessing module, and a load change rule is accurately extracted in combination with the load analysis and period detection module, so that intelligence and adaptive adjustment of resource allocation are realized, and the resource utilization efficiency and the service response speed are remarkably improved; meanwhile, by recording an operation log, generating a trust certificate and dynamically adjusting the access authority, the credibility and transparency of the resource allocation process are enhanced by using a block chain technology and a zero-knowledge proof protocol, and data tampering and trust crisis are effectively prevented.

Description

Technical Field

[0001] This invention relates to the field of computing resource management technology, specifically to an enterprise-level cloud computing resource dynamic allocation and management system and its implementation method. Background Technology

[0002] As a core pillar of the information technology field, cloud computing resource management plays a crucial role in the efficient operation and digital transformation of enterprises. It not only concerns the optimization of resource utilization but also directly affects business response speed and system security, making it an important manifestation of modern enterprise competitiveness. However, many current resource management solutions often reveal their inability to adapt to dynamic environments when faced with complex and ever-changing enterprise needs, especially in terms of the accuracy of resource allocation and the reliability of data recording, where there is significant room for improvement. These solutions mostly rely on traditional centralized management models, lacking the ability to perceive the real-time status of resource usage and struggling to ensure the credibility and transparency of data in multi-party collaboration.

[0003] Against this backdrop, the field faces significant technical challenges. The most pressing is how to achieve intelligent and adaptive resource allocation in the face of rapidly changing business needs. Due to a lack of real-time monitoring of resource status, systems often fail to respond promptly to sudden load fluctuations, leading to resource waste or service delays. More critically, this lack of real-time awareness further exacerbates the credibility issues in the resource allocation process, as the absence of a reliable mechanism to record and verify every step of resource usage easily leads to data tampering or trust crises. Summary of the Invention

[0004] The purpose of this invention is to provide an enterprise-level cloud computing resource dynamic allocation and management system and its implementation method, which solves the shortcomings of traditional cloud computing resource management in terms of dynamic adaptability, reliability and continuous improvement, and significantly improves resource utilization efficiency, system security and enterprise competitiveness.

[0005] The objective of this invention can be achieved through the following technical solutions: This application provides an enterprise-level cloud computing resource dynamic allocation and management system, including: The resource status perception and preprocessing module collects resource usage data and load status information of each node in the cloud infrastructure in real time through a distributed sensor network. It then classifies and stores the collected data, performs preliminary feature extraction, and performs noise reduction and standardization processing. The load analysis and periodic detection module performs in-depth analysis on the preprocessed resource status data, extracts the trends and periodic patterns of load changes, and constructs a business load feature map. The resource demand forecasting and assessment module uses a predictive model to estimate resource demand over a future period based on historical resource monitoring data and business load characteristic maps, and to identify potential bottleneck areas in resource allocation. The resource allocation optimization and execution module predicts potential bottleneck areas, optimizes resource allocation schemes through genetic algorithms, and updates resource allocation parameters of each node in the cloud in real time based on the optimization scheme. The security and trust management module performs security monitoring and trust verification throughout the entire resource allocation process, and then records operation logs, generates trust credentials, and dynamically adjusts access permissions. The optimization and continuous monitoring module continuously monitors changes in the resource status dataset, iterates through the data collection and processing flow according to the requirements of real-time status awareness, obtains the latest dynamic information on resource management, and then optimizes the system based on the feedback results.

[0006] Furthermore, the resource status perception and preprocessing module includes a data acquisition and classification unit, a feature extraction unit, and a data preprocessing unit. The data acquisition and classification unit collects resource usage data and load status information of each node in the cloud infrastructure in real time by deploying a distributed sensor network. Then, according to the preset sensor network protocol, it marks the node number of the data source and matches the preset business type label according to the node number, and stores the resource data into the time series database according to the label classification. The feature extraction unit uses a sliding window algorithm to calculate the mean and variance of each type of business data to obtain resource usage fluctuation characteristics. By merging high-frequency sampled data with regular data, a resource status feature matrix with timestamps is generated. The data preprocessing unit uses wavelet transform to decompose the resource status dataset into multiple scales, separating high-frequency noise components and low-frequency signal components. It removes abnormal fluctuations caused by sudden traffic through a threshold function, and then constructs the state equation of a Kalman filter to perform recursive filtering on the resource utilization rate sequence to obtain a smoothed load status sequence.

[0007] Furthermore, the load analysis and periodicity detection module includes a frequency domain feature extraction unit and a service load feature map construction unit. The frequency domain feature extraction unit uses a sliding window technique to segment and sample the smooth load sequence. The window length is set to a fixed time interval. Fourier transform is used to calculate the frequency domain distribution characteristics of the load data within each window, and the amplitude and phase information of the main frequency components are extracted. The load data of the corresponding time period is input into the LSTM network for training to learn the temporal dependency of resource consumption and identify the periodic patterns and trends of load changes. The business load feature map construction unit acquires the topological structure information of the microservice call chain, establishes the mapping relationship between service nodes and resource consumption data, constructs an associated index table to record the correspondence between call paths and load peaks, calculates the resource impact factor of each microservice based on the node weights and edge connection strengths in the feature map, generates a multi-dimensional resource state vector by combining time-series features, constructs a vector space model to represent the overall system load distribution, calculates the similarity of resource state vectors at different time points through a distance metric algorithm in the vector space, identifies key time nodes of load changes, and determines the triggering conditions and optimization strategies for resource scheduling.

[0008] Furthermore, the resource demand forecasting and assessment module includes a multi-dimensional time series construction unit and a resource demand forecasting unit. The multidimensional time series construction unit acquires key indicators from historical resource monitoring datasets, constructs a multidimensional time series data matrix according to timestamp order, and performs normalization processing on the data to obtain a standardized resource status sequence. The resource demand forecasting unit specifically includes: A Prophet prediction model is constructed based on a standardized resource state sequence. The trend prediction values ​​and confidence intervals of each resource indicator within the future time window are obtained through model training. The resource propagation weights between nodes are learned through a graph neural network to obtain the resource dependency strength matrix between services. The prediction results of the Prophet model are randomly perturbed by the Monte Carlo sampling method to generate multiple resource demand scenarios. The probability density function of resource consumption in each scenario is calculated to form a probability distribution model of resource demand. Based on the abnormal probability values ​​in the probability distribution model, abrupt nodes in the resource demand sequence are detected, and the abrupt time and the corresponding resource type are marked. By combining Prophet predictions, graph neural network dependency weights, and mutation node information, a multidimensional resource demand prediction matrix is ​​constructed to identify time periods and service nodes with concentrated resource demand, and a bottleneck risk assessment report is generated.

[0009] Furthermore, the resource allocation optimization and execution module includes a genetic algorithm optimization unit and a resource allocation update unit. The genetic algorithm optimization unit determines potential bottleneck regions based on the bottleneck risk assessment report, constructs a resource allocation scheme using chromosome encoding, and determines the multi-objective fitness function value through a genetic algorithm to obtain an optimized resource allocation scheme. The selection operation is performed based on the multi-objective fitness function value. The selected individuals are crossover and mutation processes are performed using an adaptive crossover and mutation operator. The best individual of the current generation is saved through an elite retention strategy to obtain the Pareto optimal solution set. Based on the optimal gene sequence decoding in the Pareto optimal solution set, virtual machine deployment scheme, network bandwidth allocation scheme, and storage priority configuration scheme are obtained, and dynamic resource allocation strategy is generated.

[0010] Furthermore, the resource allocation update unit specifically includes: Update the resource allocation parameters of each node in the cloud according to the dynamic resource allocation strategy, and obtain real-time resource usage data of all nodes in the cluster; Based on the resource load balancing index, the target nodes for resource allocation are marked. A two-phase commit protocol is used to send resource configuration update requests to the marked target nodes. The real traffic data of the current production environment is copied through traffic mirroring technology, and the mirrored traffic is imported into the test environment after the configuration is updated to simulate the actual business load scenario. Based on the results of the mirrored traffic in the test environment, determine whether the configuration update meets the performance requirements. If the performance verification is successful, start the canary release mechanism and gradually switch production traffic to the new configuration node according to the preset ratio, while monitoring the changes in service operation status indicators. A sliding window algorithm is used to continuously monitor service quality indicators during the canary release process. If all indicators remain stable, the service upgrade is completed and the resource configuration parameters are updated to the configuration management center.

[0011] Furthermore, the security and trust management module includes a log recording and storage unit, a trust credential generation and verification unit, and a security assessment and dynamic adjustment unit. The log recording and storage unit acquires resource allocation request data, uses a hash function to perform a summary calculation on the decision content, and if the integrity verification of the decision data passes, it generates an execution process instruction set containing allocation strategy, target node, and resource quota. By monitoring the running status of the instruction set during the execution process, performance index data such as resource utilization, response time, and throughput are obtained to determine whether the performance index exceeds the preset threshold range. Clustering algorithms are used to classify and identify abnormal patterns to determine the anomaly type and severity level. The system automatically executes compliance verification logic through smart contracts, generates alarm event records, uses a zero-knowledge proof protocol to protect the privacy of these records, and hides sensitive business data through a commitment scheme, resulting in proof data that is both privacy-protecting and audit-verifiable.

[0012] Furthermore, the trust credential generation and verification unit specifically includes: By acquiring basic information about the resource management entity, creating a unique digital identity through a distributed identity generation algorithm, and writing the digital certificate into a distributed ledger using a blockchain notarization mechanism, cross-domain identity verification credentials are obtained. By accessing historical resource usage behavior data through authentication credentials, a weighted average algorithm is used to calculate the subject's reputation score and determine dynamic access control policies. Resource allocation operations are performed according to access control policies, generating an immutable operation log data structure. The operation log is verified by multiple parties using the Byzantine fault-tolerant consensus algorithm, and each verification node votes on the legality of the operation. Determine the trustworthiness status of the operation log, generate a trust credential containing the operation hash value, verification node signature and timestamp based on the trustworthiness status, and link the trust credentials in chronological order to form a traceable trust credential chain to obtain a complete trust verification result; The entity's reputation score is updated based on the trust credential chain to obtain the updated reputation score data.

[0013] Furthermore, the security assessment and dynamic adjustment unit specifically includes: The Long Short-Term Memory Network algorithm is used to analyze historical access log data, establish a baseline model of user resource access behavior, and obtain the normal access frequency distribution and resource usage pattern feature vectors of various user roles in different time periods. The deviation between real-time access requests and normal patterns is calculated based on the behavioral baseline model to obtain abnormal access event identifiers and risk level scores. The trust credential verification process is triggered by the abnormal access event identifiers to determine the credibility assessment results of the current access subject. Based on the credibility assessment results and the anomaly risk level score, the decision engine dynamically calculates the access control adjustment parameters to obtain the updated dynamic access control list. Dynamic access control lists are used to manage subsequent resource access requests in a fine-grained manner. By matching the identity of the access subject and the type of the target resource, it is determined whether the access operation is within the allowed range. Then, the overall security status indicators of the system are statistically analyzed to obtain resource isolation instructions and traffic cleaning configuration parameters. By executing security response orchestration operations through resource isolation commands and traffic scrubbing configuration parameters, high-risk access sources and affected resource nodes are isolated, abnormal network traffic is filtered, and the execution status of system security protection measures and threat elimination effectiveness are determined.

[0014] This application provides a method for dynamic allocation and management of enterprise-level cloud computing resources, applicable to an enterprise-level cloud computing resource dynamic allocation and management system, including the following steps: The system collects resource usage data and load status information of each node in the cloud infrastructure in real time through a distributed sensor network, and then performs classification storage, feature extraction and noise reduction processing. In-depth analysis of the preprocessed resource status data is conducted to extract the trends and periodic patterns of load changes, and a business load feature map is constructed. Based on historical resource monitoring data and business load characteristic maps, predictive models are used to estimate resource demand in the future and identify potential bottleneck areas. Based on the predicted potential bottleneck areas, the resource allocation scheme is optimized using a genetic algorithm, and the resource allocation parameters of each node in the cloud are updated in real time according to the optimization scheme. The entire process of resource allocation is monitored for security and trust is verified. System security is then ensured by dynamically adjusting access permissions. By continuously monitoring changes in the resource status dataset, and cyclically executing the above data collection and processing flow according to the requirements of real-time status awareness, the latest dynamic information on resource management is obtained, and the system is optimized based on the feedback results.

[0015] The beneficial effects of this invention are as follows: This invention, through a resource status perception and preprocessing module, collects and processes resource usage data from cloud infrastructure in real time. Combined with a load analysis and periodic detection module, it can accurately extract the trends and periodic patterns of load changes. This real-time perception capability enables the system to quickly respond to sudden load fluctuations, achieve intelligent and adaptive adjustment of resource allocation, significantly improve resource utilization efficiency, reduce resource waste and service delays, thereby enhancing the operational efficiency and competitiveness of enterprises. It effectively solves the problem that traditional resource management solutions are difficult to adapt to dynamic environments, especially in terms of the accuracy and real-time performance of resource allocation, providing significant improvements. The security and trust management module ensures that every step of the resource allocation process is traceable and verifiable by recording operation logs, generating trust credentials, and dynamically adjusting access permissions. By adopting blockchain technology and zero-knowledge proof protocols, it not only protects data privacy but also enhances data credibility and transparency, effectively preventing data tampering and trust crises. It provides a solid foundation for multi-party collaboration, ensuring the security and reliability of the resource management process and solving the problem of lack of credibility and transparency in traditional centralized management models. In particular, it provides innovative solutions in terms of the reliability of data recording and the trust mechanism for multi-party collaboration. The optimization and continuous monitoring module continuously monitors changes in the resource status dataset, iteratively executes data acquisition and processing flows, and optimizes the system based on feedback results. Employing the Q-learning algorithm and the ARIMA time series forecasting model, the system can dynamically adjust resource allocation strategies, achieving continuous improvement in resource management. This closed-loop optimization mechanism not only enhances system performance and stability but also ensures that resource management strategies can adapt to ever-changing business needs, further strengthening the enterprise's digital transformation capabilities. It effectively addresses the shortcomings of traditional resource management solutions in terms of continuous optimization and adaptability, particularly providing significant improvements in system performance and the dynamic adjustment capabilities of resource management strategies. Attached Figure Description

[0016] To better understand and implement this application, the technical solution is described in detail below with reference to the accompanying drawings.

[0017] Figure 1 This is a schematic diagram of the structure of the enterprise-level cloud computing resource dynamic allocation and management system provided in Embodiment 1 of this application; Figure 2 A schematic diagram of the resource status perception and preprocessing module of the enterprise-level cloud computing resource dynamic allocation and management system provided in Embodiment 1 of this application; Figure 3 A schematic diagram of the security and trust management module of the enterprise-level cloud computing resource dynamic allocation and management system provided in Embodiment 1 of this application; Figure 4 This is a flowchart illustrating the method for dynamic allocation and management of enterprise-level cloud computing resources provided in Embodiment 2 of this application. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and systems consistent with some aspects of this application as detailed in the appended claims.

[0019] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0020] The following detailed description of the specific implementation methods, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided in detail. Example

[0021] Please see Figures 1-3 This embodiment provides an enterprise-level cloud computing resource dynamic allocation and management system, including: The resource status perception and preprocessing module collects resource usage data and load status information of each node in the cloud infrastructure in real time through a distributed sensor network. It classifies and stores the collected data, performs preliminary feature extraction, and performs noise reduction and standardization processing to provide basic data support for subsequent load feature analysis and resource demand prediction. Furthermore, the resource status perception and preprocessing module includes a data acquisition and classification unit, a feature extraction unit, and a data preprocessing unit. The data acquisition and classification unit, by deploying a distributed sensor network, collects resource usage data (such as CPU utilization, memory consumption, and network bandwidth usage) and load status information of each node in the cloud infrastructure in real time. Then, according to the preset sensor network protocol, it marks the node number of the data source and matches the preset business type label according to the node number, and stores the resource data into the time series database according to the label classification. Among them, the business type label is determined by classifying resource usage data based on business needs and historical data using clustering algorithms (such as K-means).

[0022] The feature extraction unit uses a sliding window algorithm to calculate the mean and variance of each type of business data to obtain resource usage fluctuation characteristics. If the fluctuation characteristics exceed a preset threshold, a dynamic sampling mechanism is triggered to increase the data collection frequency of the node. By merging high-frequency sampled data with regular data, a resource status feature matrix with timestamps is generated. The data preprocessing unit uses wavelet transform to decompose the resource status dataset into multiple scales, separating high-frequency noise components (detail coefficients) and low-frequency signal components (approximation coefficients). It removes abnormal fluctuations caused by sudden traffic through a threshold function to obtain the data after preliminary denoising. Then, it constructs the state equation of a Kalman filter and performs recursive filtering on the resource utilization rate sequence to obtain a smoothed load status sequence.

[0023] In this process, the wavelet transform uses the Daubechies wavelet (db4) as the wavelet basis function, and the parameters of the Kalman filter are adjusted based on the statistical characteristics of historical data.

[0024] Specifically, by collecting and processing cloud resource usage data in real time, the accuracy and reliability of the data are effectively improved, providing a solid foundation for subsequent load characteristic analysis and resource demand prediction. This not only ensures high data quality but also enhances the system's real-time response capability through dynamic sampling and advanced signal processing technology, significantly improving the efficiency and accuracy of resource management.

[0025] The load analysis and periodic detection module performs in-depth analysis on the preprocessed resource status data, extracts the trends and periodic patterns of load changes, and constructs a business load feature map to provide a basis for resource demand prediction and resource allocation. Furthermore, the load analysis and periodicity detection module includes a frequency domain feature extraction unit and a service load feature map construction unit. The frequency domain feature extraction unit uses a sliding window technique to segment and sample the smooth load sequence. The window length is set to a fixed time interval. Fourier transform is used to calculate the frequency domain distribution characteristics of the load data within each window, and the amplitude and phase information of the main frequency components are extracted. If there are periodic peaks in the frequency domain features, the load data of the corresponding time period is input into the LSTM network for training to learn the temporal dependency of resource consumption and identify the periodic patterns and trends of load changes. The Fourier transform window length is set to 1 hour, and the LSTM network structure consists of 2 layers with 128 neurons in each layer.

[0026] The business load feature graph construction unit acquires the topological structure information of the microservice call chain, establishes the mapping relationship between service nodes and resource consumption data, constructs an associated index table to record the correspondence between call paths and load peaks, and forms a business load feature graph. Based on the node weights and edge connection strengths in the feature graph, it calculates the resource impact factor of each microservice; combines time-series features to generate multi-dimensional resource state vectors, constructs a vector space model to represent the overall system load distribution, calculates the similarity of resource state vectors at different time points through a distance metric algorithm in the vector space, identifies key time nodes of load changes, and determines the triggering conditions and optimization strategies for resource scheduling.

[0027] The microservice call chain topology information is a graph structure, collected by service monitoring tools. It details the call relationships and dependencies between microservices. Specifically, it includes service nodes (each microservice is an independent business function module), directed edges (representing call relationships from one microservice to another, with edge weights indicating call frequency or intensity), and call paths (the complete call path from one service to another, which may include multiple intermediate services). The resource impact factor of each microservice reflects the degree of impact of each microservice on overall resource consumption, providing an important basis for resource scheduling and optimization.

[0028] Specifically, by extracting the trends and periodic patterns of load changes and constructing a business load feature map, the accuracy of resource demand forecasting and the efficiency of resource allocation have been significantly improved, enhancing the overall performance and stability of the system.

[0029] The resource demand forecasting and assessment module uses historical resource monitoring data and business load characteristic maps to predict resource demand over a future period using a forecasting model. It generates forecast results to meet the requirements of dynamic environment adaptation, identifies potential bottleneck areas in resource allocation, and provides a scientific basis for resource allocation optimization.

[0030] Furthermore, the resource demand forecasting and assessment module includes a multi-dimensional time series construction unit and a resource demand forecasting unit. The multidimensional time series construction unit acquires key indicators such as CPU utilization, memory usage, and network bandwidth consumption from historical resource monitoring datasets, constructs a multidimensional time series data matrix according to timestamp order, and performs normalization processing on the data to obtain a standardized resource status sequence. Among them, the multidimensional time series uses CPU utilization, memory usage, and network bandwidth consumption as the dimensions of the time series, and the normalization method adopts Min-Max normalization.

[0031] The resource demand forecasting unit specifically includes: A Prophet prediction model is constructed based on a standardized resource state sequence. Seasonal parameters are set to capture the resource consumption patterns of daily and weekly cycles. The trend prediction values ​​and confidence intervals of each resource indicator within the future time window are obtained through model training. A microservice dependency graph is constructed using service call logs. The resource propagation weights between nodes are learned through a graph neural network to obtain the resource dependency strength matrix between services. The prediction results of the Prophet model are randomly perturbed by the Monte Carlo sampling method to generate multiple resource demand scenarios. The probability density function of resource consumption in each scenario is calculated to form a probability distribution model of resource demand. Based on the abnormal probability values ​​in the probability distribution model, the Isolation Forest algorithm is used to detect abrupt change nodes in the resource demand sequence and mark the abrupt change time with the corresponding resource type. By combining Prophet predictions, graph neural network dependency weights, and mutation node information, a multidimensional resource demand prediction matrix is ​​constructed to identify time periods and service nodes where resource demand is concentrated. If the resource demand of a certain service node continuously exceeds the current allocated capacity within the prediction time window, the node is determined to be a potential bottleneck area, and a bottleneck risk assessment report is generated.

[0032] In the multidimensional resource demand forecasting matrix, rows represent the time dimension, columns represent different resource types, and element values ​​represent the expected demand for that resource at that moment.

[0033] Specifically, by constructing a multidimensional time series data matrix and the Prophet prediction model, combined with seasonal parameters and graph neural networks, it is possible to accurately predict future resource demand and identify potential bottleneck areas. This not only improves the accuracy of resource demand forecasting but also provides a scientific basis for resource allocation optimization by generating bottleneck risk assessment reports, significantly enhancing the system's adaptability in dynamic environments and the foresight of resource management.

[0034] The resource allocation optimization and execution module predicts potential bottleneck areas, optimizes resource allocation schemes through genetic algorithms, and updates the resource allocation parameters of each node in the cloud in real time according to the optimization scheme to ensure the reliability and effectiveness of resource allocation.

[0035] Furthermore, the resource allocation optimization and execution module includes a genetic algorithm optimization unit and a resource allocation update unit. The genetic algorithm optimization unit identifies potential bottleneck areas based on the bottleneck risk assessment report, constructs a resource allocation scheme using chromosome encoding, including gene segments such as virtual machine deployment location, network bandwidth allocation, and storage priority, initializes the population through the genetic algorithm, calculates the resource utilization balance fitness value, service quality assurance fitness value, and energy consumption cost fitness value for each individual, determines the multi-objective fitness function value, and obtains the optimized resource allocation scheme. Among them, the virtual machine deployment location is encoded as the location gene segment, the network bandwidth allocation is encoded as the bandwidth gene segment, and the storage priority is encoded as the priority gene segment; The selection operation is performed based on the multi-objective fitness function value. The selected individuals are cross-mutated using an adaptive crossover and mutation operator to generate a new generation of individuals. The best individual of the current generation is saved through an elite retention strategy. If the population convergence condition meets the constraints, the Pareto optimal solution set is obtained. Based on the optimal gene sequence decoding in the Pareto optimal solution set, virtual machine deployment scheme, network bandwidth allocation scheme, and storage priority configuration scheme are obtained, and dynamic resource allocation strategy is generated.

[0036] The fitness function is defined to include resource utilization equilibrium, service quality assurance rate, and energy consumption cost, specifically in the form f(x) = α⋅utilization equilibrium + β⋅service quality assurance rate − γ⋅energy consumption cost.

[0037] Furthermore, the resource allocation update unit specifically includes: Update the resource allocation parameters of each node in the cloud according to the dynamic resource allocation strategy, and establish a persistent connection with the Kubernetes orchestration platform through the bidirectional stream communication protocol to obtain real-time resource usage data of all nodes in the cluster; Based on the resource load balancing index, the target nodes for resource allocation are marked. A two-phase commit protocol is used to send resource configuration update requests to the marked target nodes, and cross-node resource reallocation operations are performed. The real traffic data of the current production environment is copied through traffic mirroring technology, and the mirrored traffic is imported into the test environment after the configuration is updated to simulate the actual business load scenario. Based on the results of mirrored traffic in the test environment, reinforcement learning algorithms are used to analyze system performance metrics such as response time, throughput, and error rate under the new configuration to determine whether the configuration update meets performance requirements. If the performance verification passes, the canary release mechanism will be initiated, and production traffic will be gradually switched to the newly configured node according to the preset ratio, while monitoring changes in service operation status indicators. A sliding window algorithm is used to continuously monitor service quality indicators during the canary release process. If all indicators remain stable, the service upgrade is completed and the resource configuration parameters are updated to the configuration management center.

[0038] The canary release is divided into 5 stages, with each stage monitoring metrics such as response time, throughput, and error rate. Specifically, by optimizing resource allocation schemes through genetic algorithms and updating resource allocation parameters of each node in the cloud in real time, the reliability and effectiveness of resource allocation are effectively solved. It can not only accurately identify potential bottleneck areas and generate optimization schemes, but also ensure the stability and performance of new configurations through canary release and reinforcement learning algorithms, significantly improving the intelligence level of resource management and the stability of system operation.

[0039] The security and trust management module monitors and verifies the entire resource allocation process to ensure transparency and trustworthiness. It safeguards system security by dynamically adjusting access permissions, and ensures the security and reliability of the resource allocation process by recording operation logs, generating trust credentials, and dynamically adjusting access permissions. It also provides security feedback for subsequent system optimization. Furthermore, the security and trust management module includes a log recording and storage unit, a trust credential generation and verification unit, and a security assessment and dynamic adjustment unit. The log recording and storage unit acquires resource allocation request data, parses the resource type, quantity, timestamp, and applicant identification information in the request, and uses a hash function to perform a summary calculation on the decision content. If the integrity verification of the decision data passes, it generates an execution process instruction set containing allocation strategy, target node, and resource quota. By monitoring the running status of the instruction set during the execution process, performance index data such as resource utilization, response time, and throughput are obtained to determine whether the performance index exceeds the preset threshold range. If the performance index exceeds the threshold range, the anomaly detection mechanism is triggered, and a clustering algorithm is used to classify and identify the anomaly pattern to determine the anomaly type and severity level. The compliance verification logic is automatically executed through smart contracts. If the verification result shows a violation, the early warning response process is activated to generate an alarm event record. The alarm event record is processed for privacy protection using a zero-knowledge proof protocol. Sensitive business data is hidden through a commitment scheme, resulting in proof data that both protects privacy and supports audit verification.

[0040] Based on the evidence data, the entire process of resource allocation information is encapsulated into transaction data through the blockchain consensus mechanism, and digital signature technology is used to ensure data integrity, thus forming an unalterable audit trail record.

[0041] Specifically, through full-process security monitoring and trust verification, the transparency and credibility of resource allocation operations are ensured, effectively safeguarding system security. It can not only record operation logs and generate trust credentials in real time, but also prevent unauthorized operations and protect data privacy through dynamic adjustment of access permissions and smart contract verification. By utilizing blockchain technology and zero-knowledge proof protocols, the immutability and audit traceability of data are further enhanced, providing reliable security feedback for continuous system optimization and significantly improving the security and reliability of resource management.

[0042] Furthermore, the trust credential generation and verification unit specifically includes: By acquiring basic information about the resource management entity, a unique digital identity is created using a distributed identity generation algorithm. This identity includes the entity's attribute hash value and timestamp information. A corresponding digital certificate is generated based on the identity data. The digital certificate is written into a distributed ledger using a blockchain storage mechanism. The validity of the certificate is verified through a smart contract, and cross-domain identity verification credentials are obtained. Accessing historical resource usage behavior data through authentication credentials, the system calculates the subject's reputation score using a weighted average algorithm. If the reputation score is higher than a preset threshold, advanced access permissions are granted; if the reputation score is lower than the preset threshold, basic access permissions are granted, thus determining a dynamic access control policy. Resource allocation operations are performed according to access control policies. The timestamp, operation type, resource identifier and execution result of each operation are recorded. An immutable operation log data structure is generated. The operation log is verified by multiple parties using the Byzantine fault-tolerant consensus algorithm. Each verification node votes on the legality of the operation. If more than two-thirds of the nodes vote in favor, the operation is verified successfully. The credibility status of the operation log is then determined. A trust credential containing the operation hash value, the verification node signature, and the timestamp is generated based on the credibility status. The trust credentials are then linked in chronological order to form a traceable trust credential chain, resulting in a complete and credible verification result. The entity's reputation score is updated based on the trust credential chain. If the number of successful operation verifications increases, the reputation score is improved; if the number of failed operation verifications increases, the reputation score is reduced, and the updated reputation score data is obtained.

[0043] The distributed identity generation algorithm uses a distributed identity (DID) generation algorithm, and the blockchain evidence storage mechanism adopts the Ethereum blockchain. Furthermore, the security assessment and dynamic adjustment unit specifically includes: The Long Short-Term Memory (LSTM) network algorithm is used to analyze historical access log data, establish a baseline model of user resource access behavior, and obtain the normal access frequency distribution and resource usage pattern feature vectors of various user roles in different time periods. The deviation value between the real-time access request and the normal mode is calculated based on the behavior baseline model. If the deviation value exceeds the preset anomaly detection threshold, it is determined that the current access behavior has an abnormal mode, and the abnormal access event identifier and risk level score are obtained. The trust credential verification process is triggered by the abnormal access event identifier. The validity and integrity of the trust credential in the access request are verified by the digital signature algorithm, and the credibility assessment result of the current access subject is determined. Based on the credibility assessment results and the anomaly risk level score, the decision engine dynamically calculates the access permission adjustment parameters. If the risk level exceeds the medium risk threshold, the resource access scope of the corresponding user is narrowed, and an updated dynamic access control list is obtained. Dynamic access control lists are used to manage subsequent resource access requests in a fine-grained manner. By matching the identity of the access subject and the type of the target resource, it is determined whether the access operation is within the allowed range, the resource access authorization decision is determined, and the overall security status indicators of the system are statistically analyzed. If the system security status indicators deviate from the preset security threshold range, the emergency response process is automatically triggered to obtain resource isolation instructions and traffic cleaning configuration parameters. By executing security response orchestration operations through resource isolation commands and traffic scrubbing configuration parameters, high-risk access sources and affected resource nodes are isolated, abnormal network traffic is filtered, and the execution status of system security protection measures and threat elimination effectiveness are determined.

[0044] The LSTM network consists of two layers, each with 128 neurons. The anomaly detection threshold is set based on the statistical characteristics of historical data.

[0045] Specifically, the trust credential generation and verification unit ensures the credibility of the resource management entity's identity by creating unique digital identity identifiers and digital certificates, utilizing blockchain notarization mechanisms and smart contracts to verify certificate validity. It also dynamically adjusts access permissions based on the entity's reputation score, generates immutable operation logs, and generates a trust credential chain through multi-party verification, further enhancing the transparency and credibility of operations. The security assessment and dynamic adjustment unit analyzes access behavior using the LSTM algorithm, identifies abnormal patterns, dynamically adjusts access permissions, and automatically triggers emergency response procedures to isolate high-risk access sources, effectively improving the system's security protection capabilities and threat response efficiency.

[0046] The optimization and continuous monitoring module continuously monitors changes in the resource status dataset, and cyclically executes the above data collection and processing flow according to the requirements of real-time status awareness to obtain the latest dynamic information on resource management. At the same time, it optimizes the system based on the feedback results to ensure continuous improvement in resource allocation and system performance.

[0047] Furthermore, the optimization and continuous monitoring module specifically includes: Acquire service quality index data during system operation, including key performance parameters such as response time, throughput, and error rate. Use time series analysis methods to backtrack historical QoS data to obtain quantitative evaluation results of resource allocation effectiveness. Historical QoS data refers to performance metrics recorded by the system during its past operation, including response time, throughput, error rate, and resource utilization, which are used to evaluate the system's service quality and guide optimization decisions.

[0048] The Q-learning algorithm is used to train the resource management agent. The agent selects resource allocation actions based on the current system state, obtains reward feedback signals by interacting with the environment, and continuously updates the action value function to obtain optimized resource allocation strategy parameters. By analyzing historical resource usage patterns using the ARIMA time series forecasting model and combining them with policy parameters learned by the agent, resource demand within the future time window is predicted and calculated to obtain the predicted value and confidence interval range of resource demand. If the prediction results show that the change in resource demand exceeds the preset threshold, the canary release mechanism is triggered to deploy a new resource allocation strategy on a small subset of the production environment. The actual execution effect of the new strategy is judged by monitoring the performance index changes of the deployment area in real time. The effectiveness score of the strategy is calculated based on the monitoring data of the canary deployment area. If the score reaches the preset standard, the deployment scope of the new strategy is gradually expanded, and the experience replay buffer of the agent is updated to obtain the strategy verification results and deployment decisions. The training sample library of the reinforcement learning agent is updated using the policy verification results. The model parameters are adjusted through incremental learning to form a closed-loop resource management optimization mechanism, thereby obtaining continuously improved resource allocation capabilities.

[0049] The learning rate of the Q-learning algorithm was set to 0.1, and the discount factor was set to 0.9. The parameters of the ARIMA model were determined based on the autocorrelation function and partial autocorrelation function of historical data. The canary deployment was divided into 5 stages, and each stage monitored indicators such as response time, throughput and error rate to gradually expand the deployment scope of the new strategy.

[0050] Specifically, by continuously monitoring changes in the resource status dataset, cyclically executing data collection and processing procedures, and combining service quality index data and historical QoS data for quantitative evaluation, the resource allocation strategy is dynamically adjusted using the Q-learning algorithm and ARIMA model. This not only ensures continuous improvement in resource allocation but also verifies the effectiveness of new strategies through a canary release mechanism and real-time monitoring, ultimately forming a closed-loop optimization mechanism that significantly improves system performance and the level of intelligence in resource management. Example

[0051] Please see Figure 4 This embodiment provides a method for dynamic allocation and management of enterprise-level cloud computing resources, applied to an enterprise-level cloud computing resource dynamic allocation and management system, including the following steps: S1. Real-time collection of resource usage data and load status information of each node in the cloud infrastructure through a distributed sensor network, followed by classification, storage, feature extraction, and noise reduction to provide basic data support for subsequent analysis; S2. Conduct in-depth analysis of the preprocessed resource status data, extract the trends and periodic patterns of load changes, and construct a business load characteristic map to provide a basis for resource demand forecasting and resource allocation. S3. Based on historical resource monitoring data and business load characteristic maps, predictive models are used to estimate resource demand in the future, identify potential bottleneck areas, and provide a scientific basis for resource allocation optimization. S4. Based on the predicted potential bottleneck areas, optimize the resource allocation scheme through genetic algorithms, and update the resource allocation parameters of each node in the cloud in real time according to the optimization scheme to ensure the reliability and effectiveness of resource allocation. S5. Conduct security monitoring and trust verification throughout the entire resource allocation process to ensure the transparency and credibility of operations, and safeguard system security by dynamically adjusting access permissions; S6. By continuously monitoring changes in the resource status dataset, the above data collection and processing process is executed cyclically according to the requirements of real-time status awareness to obtain the latest dynamic information on resource management. At the same time, the system is optimized based on the feedback results to ensure continuous improvement in resource allocation and system performance.

[0052] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An enterprise-level cloud computing resource dynamic allocation and management system, characterized by: include: The resource status perception and preprocessing module collects resource usage data and load status information of each node in the cloud infrastructure in real time through a distributed sensor network. It then classifies and stores the collected data, performs preliminary feature extraction, and performs noise reduction and standardization processing. The load analysis and periodic detection module performs in-depth analysis on the preprocessed resource status data, extracts the trends and periodic patterns of load changes, and constructs a business load feature map. The resource demand forecasting and assessment module uses a predictive model to estimate resource demand over a future period based on historical resource monitoring data and business load characteristic maps, and to identify potential bottleneck areas in resource allocation. The resource allocation optimization and execution module predicts potential bottleneck areas, optimizes resource allocation schemes through genetic algorithms, and updates resource allocation parameters of each node in the cloud in real time based on the optimization scheme. The security and trust management module monitors and verifies the trust of the entire resource allocation process, and then records operation logs, generates trust credentials, and dynamically adjusts access permissions. The optimization and continuous monitoring module continuously monitors changes in the resource status dataset, iterates through the data collection and processing flow according to the requirements of real-time status awareness, obtains the latest dynamic information on resource management, and then optimizes the system based on the feedback results.

2. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 1, characterized in that: The resource status awareness and preprocessing module includes a data acquisition and classification unit, a feature extraction unit, and a data preprocessing unit. The data acquisition and classification unit collects resource usage data and load status information of each node in the cloud infrastructure in real time by deploying a distributed sensor network. Then, according to the preset sensor network protocol, it marks the node number of the data source and matches the preset business type label according to the node number, and stores the resource data into the time series database according to the label classification. The feature extraction unit uses a sliding window algorithm to calculate the mean and variance of each type of business data to obtain resource usage fluctuation characteristics. By merging high-frequency sampled data with regular data, a resource status feature matrix with timestamps is generated. The data preprocessing unit uses wavelet transform to decompose the resource status dataset into multiple scales, separating high-frequency noise components and low-frequency signal components. It removes abnormal fluctuations caused by sudden traffic through a threshold function, and then constructs the state equation of a Kalman filter to perform recursive filtering on the resource utilization rate sequence to obtain a smoothed load status sequence.

3. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 1, characterized in that: The load analysis and periodicity detection module includes a frequency domain feature extraction unit and a service load feature map construction unit. The frequency domain feature extraction unit uses a sliding window technique to segment and sample the smooth load sequence. The window length is set to a fixed time interval. Fourier transform is used to calculate the frequency domain distribution characteristics of the load data within each window, and the amplitude and phase information of the main frequency components are extracted. The load data of the corresponding time period is input into the LSTM network for training to learn the temporal dependency of resource consumption and identify the periodic patterns and trends of load changes. The business load feature map construction unit acquires the topological structure information of the microservice call chain, establishes the mapping relationship between service nodes and resource consumption data, constructs an associated index table to record the correspondence between call paths and load peaks, calculates the resource impact factor of each microservice based on the node weights and edge connection strengths in the feature map, generates a multi-dimensional resource state vector by combining time-series features, constructs a vector space model to represent the overall system load distribution, calculates the similarity of resource state vectors at different time points through a distance metric algorithm in the vector space, identifies key time nodes of load changes, and determines the triggering conditions and optimization strategies for resource scheduling.

4. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 1, characterized in that: The resource demand forecasting and assessment module includes a multi-dimensional time series construction unit and a resource demand forecasting unit. The multidimensional time series construction unit acquires key indicators from historical resource monitoring datasets, constructs a multidimensional time series data matrix according to timestamp order, and performs normalization processing on the data to obtain a standardized resource status sequence. The resource demand forecasting unit specifically includes: A Prophet prediction model is constructed based on a standardized resource state sequence. The trend prediction values ​​and confidence intervals of each resource indicator within the future time window are obtained through model training. The resource propagation weights between nodes are learned through a graph neural network to obtain the resource dependency strength matrix between services. The prediction results of the Prophet model are randomly perturbed by the Monte Carlo sampling method to generate multiple resource demand scenarios. The probability density function of resource consumption in each scenario is calculated to form a probability distribution model of resource demand. Based on the abnormal probability values ​​in the probability distribution model, abrupt nodes in the resource demand sequence are detected, and the abrupt time and the corresponding resource type are marked. By combining Prophet predictions, graph neural network dependency weights, and mutation node information, a multidimensional resource demand prediction matrix is ​​constructed to identify time periods and service nodes with concentrated resource demand, and a bottleneck risk assessment report is generated.

5. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 1, characterized in that: The resource allocation optimization and execution module includes a genetic algorithm optimization unit and a resource allocation update unit. The genetic algorithm optimization unit determines potential bottleneck regions based on the bottleneck risk assessment report, constructs a resource allocation scheme using chromosome encoding, and determines the multi-objective fitness function value through a genetic algorithm to obtain an optimized resource allocation scheme. The selection operation is performed based on the multi-objective fitness function value. The selected individuals are crossover and mutation processes are performed using an adaptive crossover and mutation operator. The best individual of the current generation is saved through an elite retention strategy to obtain the Pareto optimal solution set. Based on the optimal gene sequence decoding in the Pareto optimal solution set, virtual machine deployment scheme, network bandwidth allocation scheme, and storage priority configuration scheme are obtained, and dynamic resource allocation strategy is generated.

6. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 5, characterized in that: The resource allocation update unit specifically includes: Update the resource allocation parameters of each node in the cloud according to the dynamic resource allocation strategy, and obtain real-time resource usage data of all nodes in the cluster; Based on the resource load balancing index, the target nodes for resource allocation are marked. A two-phase commit protocol is used to send resource configuration update requests to the marked target nodes. The real traffic data of the current production environment is copied through traffic mirroring technology, and the mirrored traffic is imported into the test environment after the configuration is updated to simulate the actual business load scenario. Based on the results of the mirrored traffic in the test environment, determine whether the configuration update meets the performance requirements. If the performance verification is successful, start the canary release mechanism and gradually switch production traffic to the new configuration node according to the preset ratio, while monitoring the changes in service operation status indicators. A sliding window algorithm is used to continuously monitor service quality indicators during the canary release process. If all indicators remain stable, the service upgrade is completed and the resource configuration parameters are updated to the configuration management center.

7. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 1, characterized in that: The security and trust management module includes a log recording and storage unit, a trust credential generation and verification unit, and a security assessment and dynamic adjustment unit. The log recording and storage unit acquires resource allocation request data, uses a hash function to perform a summary calculation on the decision content, and if the integrity verification of the decision data passes, it generates an execution process instruction set containing allocation strategy, target node, and resource quota. By monitoring the running status of the instruction set during the execution process, performance index data such as resource utilization, response time, and throughput are obtained to determine whether the performance index exceeds the preset threshold range. Clustering algorithms are used to classify and identify abnormal patterns to determine the anomaly type and severity level. The system automatically executes compliance verification logic through smart contracts, generates alarm event records, uses a zero-knowledge proof protocol to protect the privacy of these records, and hides sensitive business data through a commitment scheme, resulting in proof data that is both privacy-protecting and audit-verifiable.

8. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 7, characterized in that: The trust credential generation and verification unit specifically includes: By acquiring basic information about the resource management entity, creating a unique digital identity through a distributed identity generation algorithm, and writing the digital certificate into a distributed ledger using a blockchain notarization mechanism, cross-domain identity verification credentials are obtained. By accessing historical resource usage behavior data through authentication credentials, a weighted average algorithm is used to calculate the subject's reputation score and determine dynamic access control policies. Resource allocation operations are performed according to access control policies, generating an immutable operation log data structure. The operation log is verified by multiple parties using the Byzantine fault-tolerant consensus algorithm, and each verification node votes on the legality of the operation. Determine the trustworthiness status of the operation log, generate a trust credential containing the operation hash value, verification node signature and timestamp based on the trustworthiness status, and link the trust credentials in chronological order to form a traceable trust credential chain to obtain a complete trust verification result; The entity's reputation score is updated based on the trust credential chain to obtain the updated reputation score data.

9. The enterprise-level cloud computing resource dynamic allocation and management system according to claim 7, characterized in that: The security assessment and dynamic adjustment unit specifically includes: The Long Short-Term Memory Network algorithm is used to analyze historical access log data, establish a baseline model of user resource access behavior, and obtain the normal access frequency distribution and resource usage pattern feature vectors of various user roles in different time periods. The deviation between real-time access requests and normal patterns is calculated based on the behavioral baseline model to obtain abnormal access event identifiers and risk level scores. The trust credential verification process is triggered by the abnormal access event identifiers to determine the credibility assessment results of the current access subject. Based on the credibility assessment results and the anomaly risk level score, the decision engine dynamically calculates the access control adjustment parameters to obtain the updated dynamic access control list. Dynamic access control lists are used to perform fine-grained control over subsequent resource access requests. By matching the identity of the access subject and the type of the target resource, it is determined whether the access operation is within the allowed range. Then, the overall security status indicators of the system are statistically analyzed to obtain resource isolation instructions and traffic cleaning configuration parameters. By executing security response orchestration operations through resource isolation commands and traffic scrubbing configuration parameters, high-risk access sources and affected resource nodes are isolated, abnormal network traffic is filtered, and the execution status of system security protection measures and threat elimination effectiveness are determined.

10. A method for dynamic allocation and management of enterprise-level cloud computing resources, applied to the enterprise-level cloud computing resource dynamic allocation and management system as described in any one of claims 1-9, characterized in that: Includes the following steps: The system collects resource usage data and load status information of each node in the cloud infrastructure in real time through a distributed sensor network, and then performs classification storage, feature extraction and noise reduction processing. In-depth analysis of the preprocessed resource status data is conducted to extract the trends and periodic patterns of load changes, and a business load feature map is constructed. Based on historical resource monitoring data and business load characteristic maps, predictive models are used to estimate resource demand in the future and identify potential bottleneck areas. Based on the predicted potential bottleneck areas, the resource allocation scheme is optimized using a genetic algorithm, and the resource allocation parameters of each node in the cloud are updated in real time according to the optimization scheme. The entire process of resource allocation is monitored for security and trust is verified. System security is then ensured by dynamically adjusting access permissions. By continuously monitoring changes in the resource status dataset, and cyclically executing the above data collection and processing flow according to the requirements of real-time status awareness, the latest dynamic information on resource management is obtained, and the system is optimized based on the feedback results.

Images