Resource object relationship design method for computing network service scene

Abstract

The invention belongs to the technical field of computing network fusion, and particularly relates to a computing network business scene-oriented resource object relationship design method, which comprises the following steps of: generating real-time updated virtual resource mapping by using a digital twinning technology to predict a future resource state, and finding a potential association mode among resources by using a graph neural network algorithm GNN to predict the resource state. According to the method, through the digital twin technology and the real-time updating mechanism, the resource state can be monitored in real time, the future resource state can be predicted, the real-time performance and the prospective performance of resource management are improved, and compared with traditional manual monitoring or simple automatic scripts, the resource management efficiency is improved. The digital twin technology provides real-time data with higher precision, the prediction model can deal with resource changes in advance, and problems caused by resource shortage or excess are reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of algorithm and network integration, and particularly relates to a resource object relationship design method for algorithm and network service scenarios. BACKGROUND

[0002] With the development of information technology, algorithm and network service scenarios are becoming increasingly complex, and resource management is facing many challenges.

[0003] The existing resource management system has the problems of poor real-time performance, low data accuracy and lack of prediction ability in resource state monitoring and prediction; in the identification of interaction rules between resources, traditional methods are difficult to capture nonlinear interaction and the model lacks intelligence; in the extraction of business requirements, manual extraction is inefficient and prone to errors; in the generation of optimal resource allocation strategies, static strategies have poor adaptability and lack self-learning ability; in the evaluation of resource allocation strategy effect, there is a lack of effective quantitative evaluation mechanism and user feedback mechanism; in the design of user interface, the existing interface has poor interactivity and poor user experience; in the security protection, single security measures are difficult to comprehensively protect data privacy and network security, and there is a lack of effective security evaluation mechanism.

[0004] To solve the above problems, the application provides an intelligent resource object relationship design method for algorithm and network service scenarios. SUMMARY

[0005] To make up for the deficiencies of the prior art, the application solves the technical problems by adopting the technical scheme of a resource object relationship design method for algorithm and network service scenarios, which comprises the following steps:

[0006] S1: generating a real-time updated virtual resource mapping X using digital twin technology t , to predict future resource state X t+1 ;

[0007] S2: discovering potential association patterns between resources using a graph neural network algorithm GNN, and continuously optimizing the intelligent connection model between resources through a weight update formula ΔW;

[0008] S3: extracting business requirements using natural language processing technology NLP, and constructing a requirement knowledge graph V;

[0009] S4: combining the intelligent connection model GNN and the requirement knowledge graph V; generating an optimal resource allocation strategy Q(s, a) using a reinforcement learning algorithm RL to implement the resource allocation strategy and optimizing the scheduling algorithm through real-time monitoring;

[0010] Wherein, S1 to S4 are influenced by the closed-loop feedback mechanism, and the optimization of resource allocation is promoted together; and the resource object relationship design method for network service scene also contains a calculation formula: S=f(A,D,L), and is related to each other through a closed-loop feedback mechanism, wherein S represents the resource allocation score, f represents the comprehensive score function, A represents the resource availability, D represents the resource demand, and L represents the network delay.

[0011] Preferably, the digital twin technology in S1 is used to predict the future resource state through a prediction model, and serves as an input of the graph neural network algorithm in S2;

[0012] The prediction model formula is as follows:

[0013] X t+1 =g(X t ,E)

[0014] Wherein, X t+1 represents the resource state at the next moment, g represents the prediction function, X t represents the current resource state, and E represents the environmental factor.

[0015] Preferably, the graph neural network algorithm in S2 identifies the nonlinear interaction law between resources through weight update, and feeds back to S1 to improve the resource mapping accuracy;

[0016] The weight update formula is as follows:

[0017]

[0018] Wherein, ΔW represents the weight update value, η represents the learning rate, Y represents the target output, Y' represents the actual output, W represents the weight, represents the partial derivative of the target output with respect to the weight.

[0019] Preferably, the natural language processing technology in S3 extracts business intent from the document through intent extraction and converts it into an operable data format;

[0020] The intent extraction formula is as follows:

[0021] V=h(T)

[0022] Wherein, V represents the intent vector, h represents the intent extraction function, and T represents the document text.

[0023] Preferably, the reinforcement learning algorithm in S4 learns from history data through policy update to realize dynamic optimization of resource allocation;

[0024] The policy update formula is as follows:

[0025] Q(s, a) = Q(s, a) + a[R + gmax a′ Q(s', a') - Q(s, a)]

[0026] wherein Q denotes a state-action value function, s denotes a state, a denotes an action, a denotes a learning rate, R denotes an immediate reward, g denotes a discount factor, s' denotes a next state, and a' denotes an action in the next state.

[0027] Preferably, it further comprises evaluating the effect of the resource allocation strategy and adjusting the subsequent resource scheduling logic according to the evaluation result;

[0028] The effect evaluation formula is:

[0029]

[0030] wherein M denotes an effect evaluation score, T i denotes an actual completion time of task i, T i denotes an expected completion time of task i.

[0031] Preferably, the reinforcement learning algorithm used in S4 is adjusted according to the specific application scenario;

[0032] The algorithm selection formula is:

[0033] S A = S A x S C

[0034] wherein S A denotes algorithm applicability, and S C denotes application scenario applicability.

[0035] Preferably, it further comprises a user interface module, allowing users to visually view the resource state and receive feedback for improving the strategy;

[0036] The feedback processing formula is:

[0037] F = w1S + w2P

[0038] wherein F denotes a feedback score, w1 and w2 denote weights, S denotes user satisfaction, and P denotes user suggestions.

[0039] Preferably, the application scenario adaptability evaluation formula is applicable to multiple application scenarios and can improve the adaptability and flexibility of the system;

[0040] The application scenario adaptability evaluation formula is:

[0041]

[0042] wherein A Aindicates application scenario adaptability, D j indicates successful deployment times, T j indicates total deployment times.

[0043] Preferably, a security module is further included, which is used for ensuring data privacy and network security in resource management and scheduling processes through a security evaluation formula;

[0044] The security evaluation formula is:

[0045]

[0046] Wherein, L S indicates security level, K indicates total number of control measures, C k indicates the kth control measure E k indicates effectiveness of the kth control measure.

[0047] The beneficial effects of the present application are as follows:

[0048] 1. The present application can monitor resource status in real time and predict future resource status through digital twin technology and real-time updating mechanism, thereby improving real-time and foresight of resource management. Compared with traditional manual monitoring or simple automatic script, digital twin technology provides higher-precision real-time data, and the prediction model can respond to resource changes in advance to reduce problems caused by resource shortage or excess.

[0049] 2. The present application can automatically discover nonlinear interaction rules between resources through graph neural network and natural language processing technology, and extract business requirements from documents, thereby improving intelligence and accuracy of resource management.

[0050] 3. The present application generates dynamically optimized resource allocation strategies through reinforcement learning algorithm, and continuously adjusts strategies through real-time monitoring and quantitative evaluation mechanism, thereby having higher flexibility. Meanwhile, the security module ensures data privacy and network security in resource management and scheduling processes. BRIEF DESCRIPTION OF DRAWINGS

[0051] The present application will be further described below in conjunction with the accompanying drawings.

[0052] Figure 1 is a module schematic diagram of the present application. DETAILED DESCRIPTION

[0053] In order to make the technical means, creative features, purposes and effects realized by the present application easy to understand, the present application will be further described below in conjunction with specific embodiments.

[0054] For example, Figure 1As shown, the present application proposes a resource object relationship design method for computing network service scenarios, including the following steps:

[0055] Including the following steps:

[0056] S1: generating a real-time updated virtual resource mapping X by using digital twin technology t , to predict the future resource state X t+1 ;

[0057] S2: discovering the potential association pattern between resources by using a graph neural network algorithm GNN, and continuously optimizing the intelligent connection model between resources through a weight update formula AW;

[0058] S3: applying natural language processing technology NLP to extract business requirements and build a requirement knowledge graph V;

[0059] S4: combining the intelligent connection model GNN and the requirement knowledge graph V; using a reinforcement learning algorithm RL to generate an optimal resource allocation strategy Q(s, a) to implement the resource allocation strategy and optimize the scheduling algorithm through real-time monitoring;

[0060] Wherein, S1 to S4 are mutually influenced through a closed-loop feedback mechanism, and together promote the optimization of resource allocation; and the resource object relationship design method for computing network service scenarios also includes a calculation formula: S=f(A, D, L), and is related to each other through a closed-loop feedback mechanism, wherein S represents the resource allocation score, f represents the comprehensive score function, A represents resource availability, D represents resource demand, and L represents network latency;

[0061] By comprehensively considering resource availability, resource demand and network latency, the comprehensive score of resource allocation is calculated, so that the resource allocation strategy is more scientific and comprehensive, which helps to make the optimal decision in a complex computing network environment, and the comprehensive score mechanism considers the influence of multiple key factors, so that the resource allocation not only focuses on current demand, but also considers network performance, thereby improving the efficiency and quality of resource allocation.

[0062] As a specific embodiment of the present application, the digital twin technology in S1 is used to predict the future resource state through a prediction model, and serves as the input of the graph neural network algorithm in S2;

[0063] The prediction model formula is as follows:

[0064] X t+1 =g(X t ,E)

[0065] Wherein, X t+1 represents the resource state at the next moment, g represents the prediction function, X t represents the current resource state, and E represents environmental factors;

[0066] By predicting the future resource state, preparing in advance to cope with resource changes, reducing problems caused by resource shortages or surpluses, and combining digital twin technology and environmental factors in the dynamic prediction model, the future resource state is accurately predicted, and the foresight and adaptability of the system are enhanced.

[0067] As a specific embodiment of the present application, the graph neural network algorithm in S2 identifies the nonlinear interaction law between resources through weight updating and feeds back to S1 to improve resource mapping accuracy.

[0068] The weight update formula is as follows:

[0069]

[0070] where ΔW represents the weight update value, η represents the learning rate, Y represents the target output, Y' represents the actual output, W represents the weight, the partial derivative of the target output with respect to the weight;

[0071] Updating the weights of the graph neural network ensures that the model can accurately capture the nonlinear interaction law between resources, which enables the model to more accurately reflect the complex relationships in the real world. By learning the interaction between resources through the graph neural network, potential association patterns can be automatically discovered, improving the intelligence and accuracy of the resource connection model.

[0072] As a specific embodiment of the present application, the natural language processing technology in S3 extracts business intent from documents through intent extraction and converts it into an operable data format.

[0073] The intent extraction formula is as follows:

[0074] V = h(T)

[0075] where V represents the intent vector, h represents the intent extraction function, and T represents the document text.

[0076] Extracting business intent from document text and converting it into an operable data format helps to convert natural language descriptions of requirements into a form that the system can understand and execute. By using natural language processing technology to automatically extract business intent, the requirement analysis process is simplified, and it is ensured that requirements can be accurately communicated to the system.

[0077] As a specific embodiment of the present application, the reinforcement learning algorithm in S4 learns from historical data through policy updating to achieve dynamic optimization of resource allocation.

[0078] The policy update formula is as follows:

[0079] Q(s, a) = Q(s, a) + a[R + gmax a′ Q(s', a') - Q(s, a)]

[0080] Where Q represents the state-action value function, s represents the state, a represents the action, a represents the learning rate, R represents the immediate reward, g represents the discount factor, s' represents the next state, and a' represents the action in the next state.

[0081] The policy update in the reinforcement learning algorithm optimizes the resource allocation strategy by learning from historical data, enabling the system to learn and adjust itself according to actual conditions, achieving dynamic optimization. Combined with historical data and future predictions, the reinforcement learning algorithm can achieve self-learning and dynamic optimization of strategies, improving the flexibility and efficiency of resource allocation.

[0082] As a specific embodiment of the present application, it also includes evaluating the effect of the resource allocation strategy and adjusting the subsequent resource scheduling logic according to the evaluation results.

[0083] The effect evaluation formula is:

[0084]

[0085] Where M represents the effect evaluation score, T i represents the actual completion time of task i, T i represents the expected completion time of task i.

[0086] Evaluating the actual effect of the resource allocation strategy by comparing the actual completion time and the expected completion time helps to discover problems in a timely manner and adjust the strategy, improving the accuracy and reliability of resource allocation. By quantifying the effect evaluation, the resource allocation strategy can be continuously improved to ensure that the system is always in the best operating state.

[0087] As a specific embodiment of the present application, the reinforcement learning algorithm used in S4 is adjusted according to the specific application scenario.

[0088] The algorithm selection formula is:

[0089] S A = S A × S C

[0090] Where S A represents the algorithm suitability, S C represents the application scenario suitability.

[0091] The algorithm most suitable for a specific application scenario is selected, and the algorithm selection is more personalized by calculating the product of the algorithm applicability and the application scenario applicability, so that the algorithm selection can better adapt to different business requirements; the algorithm selection mechanism combines the algorithm characteristics and the application scenario characteristics, improves the pertinence and effectiveness of the algorithm selection.

[0092] As a specific embodiment of the application, a user interface module is further included, allowing users to visually view resource status and receive feedback for improving the strategy;

[0093] The feedback processing formula is:

[0094] F = ω1S + ω2P

[0095] Wherein, F represents the feedback score, ω1 and ω2 represent the weight, S represents the user satisfaction, and P represents the user suggestion;

[0096] The user satisfaction and the user suggestion are weighted. This makes the system better respond to user needs and improve user experience; the user feedback processing mechanism considers the importance of user satisfaction and suggestion, so that the resource allocation strategy can better meet user expectations.

[0097] As a specific embodiment of the application, the application scenario adaptability evaluation formula is applicable to a variety of application scenarios and can improve the adaptability and flexibility of the system;

[0098] The application scenario adaptability evaluation formula is:

[0099]

[0100] Wherein, A A represents the application scenario adaptability, D j represents the number of successful deployments, and T j represents the total number of deployments;

[0101] By calculating the ratio of the number of successful deployments to the total number of deployments, the performance of the system in different scenarios can be verified, and its adaptability and flexibility can be improved; the application scenario adaptability evaluation mechanism provides a standard for measuring the adaptability of the system by quantifying the deployment success rate, and enhances the robustness of the system.

[0102] As a specific embodiment of the application, a security module is further included, which is used to ensure data privacy and network security in the resource management and scheduling process through a security evaluation formula;

[0103] The security evaluation formula is:

[0104]

[0105] Wherein, L Swherein S represents the level of security, K represents the total number of control measures, C k wherein E represents the kth control measure k wherein E represents the effectiveness of the kth control measure

[0106] By calculating the sum of the products of security control measures and control measure effectiveness, data privacy and network security in resource management and scheduling processes are facilitated; and the security evaluation mechanism comprehensively considers various security control measures and their effectiveness, improving the system's ability to protect data privacy and network security.

[0107] In a data center management scenario, the allocation of server resources needs to be optimized. First, real-time state data of servers is collected by deploying sensors, and a virtual mapping is constructed using digital twin technology. Then, the interdependence between resources is analyzed using graph neural networks. Next, business requirement documents are parsed using natural language processing technology. Finally, an optimal resource allocation strategy is generated using reinforcement learning algorithms, and resource states are displayed through a user interface to collect feedback information and continuously optimize the strategy.

[0108] The present application uses digital twin technology to update resource states in real time and predict future resource states. Graph neural networks are used to discover the nonlinear interaction rules between resources. Natural language processing technology is used to automatically extract business requirements. Combining graph neural network models and requirement knowledge graphs, reinforcement learning algorithms are used to generate optimal resource allocation strategies. Real-time monitoring and quantitative evaluation mechanisms are used to continuously optimize the strategy. A user-friendly interface is designed to enhance interactivity and user experience. A security module is implemented to ensure data privacy and network security.

[0109] The basic principles, main features and advantages of the present application are shown and described above. Those skilled in the art should understand that the present application is not limited by the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A resource object relationship design method for computing network service scenarios, characterized in that, Includes the following steps: S1: Utilize digital twin technology to generate real-time updated virtual resource maps X t To predict future resource status X t+1 ; S2: Use the graph neural network algorithm (GNN) to discover potential correlation patterns between resources, and continuously optimize the intelligent connection model between resources through the weight update formula ΔW. S3: Apply Natural Language Processing (NLP) technology to extract business requirements and construct a requirement knowledge graph V; S4: Combine the intelligent connection model GNN and the demand knowledge graph V; use the reinforcement learning algorithm RL to generate the optimal resource allocation strategy Q(s,a), implement the resource allocation strategy, and optimize the scheduling algorithm through real-time monitoring; Among them, S1 to S4 influence each other through a closed-loop feedback mechanism to jointly promote the optimization of resource allocation; and the resource object relationship design method for computing network service scenarios also includes the calculation formula: S = f(A, D, L), which are interconnected through a closed-loop feedback mechanism, where S represents the resource allocation score, f represents the comprehensive scoring function, A represents resource availability, D represents resource demand, and L represents network latency.

2. The resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, The digital twin technology described in S1 is used to predict future resource status through a predictive model and serves as input to the graph neural network algorithm in S2. The prediction model formula is as follows: X t+1 =g(X t ,E) Among them, X t+1 Let g represent the resource state at the next time step, g represent the prediction function, and X represent the resource state at the next time step. t E represents the current resource status, and E represents environmental factors.

3. The resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, The graph neural network algorithm described in S2 identifies the nonlinear interaction patterns between resources through weight updates and feeds them back to S1 to improve the accuracy of resource mapping. The weight update formula is as follows: Where ΔW represents the weight update value, η represents the learning rate, Y represents the target output, Y′ represents the actual output, and W represents the weight. This represents the partial derivative of the target output with respect to the weights.

4. The resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, Natural language processing technology in S3 extracts business intent from documents through intent extraction and transforms it into an actionable data format; The intent extraction formula is as follows: V = h(T) Where V represents the intent vector, h represents the intent extraction function, and T represents the document text.

5. The resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, The reinforcement learning algorithm described in S4 learns from historical data through policy updates to achieve dynamic optimization of resource allocation; The strategy update formula is: Q(s,a)=Q(s,a)+α[R+γmax a′ Q(s′,a′)-Q(s,a)] Where Q represents the state-action value function, s represents the state, a represents the action, α represents the learning rate, R represents the immediate reward, γ represents the discount factor, s′ represents the next state, and a′ represents the action in the next state.

6. The resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, It also includes evaluating the effectiveness of resource allocation strategies and adjusting subsequent resource scheduling logic based on the evaluation results; The formula for evaluating the effect is: Where M represents the effect evaluation score, T i Indicates the actual completion time of task i, T i This indicates the expected completion time of task i.

7. A resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, The reinforcement learning algorithm used in S4 is adjusted according to the specific application scenario; The algorithm selection formula is: S A =S A ×S C Among them, S A Indicates algorithm applicability, S C Indicates the applicability of the application scenario.

8. A resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, It also includes a user interface module that allows users to visualize resource status and receive feedback for strategy improvement; The feedback processing formula is: F = ω1S + ω2P Where F represents the feedback score, ω1 and ω2 represent the weights, S represents user satisfaction, and P represents user suggestions.

9. A resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, The application scenario adaptability evaluation formula is applicable to a variety of application scenarios and can improve the system's adaptability and flexibility. The formula for evaluating the adaptability of an application scenario is: Among them, A A Indicates application scenario adaptability, D j Indicates the number of successful deployments, T j Indicates the total number of deployments.

10. A resource object relationship design method for computing network service scenarios according to claim 1, characterized in that, It also includes a security module, which uses a security assessment formula to ensure data privacy and network security during resource management and scheduling. The security assessment formula is: Among them, L S K represents the safety level, C represents the total number of control measures, and K represents the total number of control measures. k E represents the k-th control measure. k This indicates the effectiveness of the k-th control measure.

Images