CN118590430A - A network integrated dynamic resource routing management system and method - Google Patents

Abstract

The invention relates to the technical field of network resource management, in particular to a network integrated dynamic resource route management system and a method, wherein the system comprises a resource monitoring module, a data acquisition fusion module, an intelligent prediction analysis module, a decision support optimization module, a dynamic resource scheduling module and a feedback optimization module; wherein: and a resource monitoring module: monitoring and collecting the service condition of network resources; and a data acquisition fusion module: preprocessing the collected network resource data; intelligent predictive analysis module: predicting resource demand and flow variation trend; decision support optimization module: the resource allocation policy and routing path are dynamically adjusted. According to the invention, through comprehensive application of real-time monitoring, intelligent predictive analysis and dynamic resource scheduling, the utilization efficiency and response speed of network resources are obviously improved, the high quality of network service and the stability of a system are ensured, and the problems of static configuration and slow response in the prior art are solved.

Description

A network integrated dynamic resource routing management system and method

Technical Field

The present invention relates to the technical field of network resource management, and in particular to a network integrated dynamic resource routing management system and method.

Background Art

With the rapid development of information technology, the complexity of network resource management has increased significantly, especially in large-scale data centers and cloud computing environments. The challenges faced by modern network systems include the dynamic nature of resource configuration, service quality assurance, and the ability to respond quickly to emergencies. Traditional network resource management methods rely on static configuration strategies, which often cannot effectively adapt to the rapid changes in network conditions and business needs. In addition, traditional methods are insufficient in resource monitoring and prediction accuracy, which often leads to resource waste or service interruption. For example, when network traffic suddenly increases, a statically configured network may not be able to adjust resources in a timely manner, affecting user experience and business continuity.

In order to solve these problems, the present invention proposes a network integrated dynamic resource routing management system and method thereof, aiming to improve the flexibility and efficiency of network resource management through dynamic monitoring and real-time data analysis. The shortcomings of the prior art include slow response speed of resource allocation schemes, lack of effective resource optimization feedback mechanism and insufficient routing adjustment strategy, which limit the performance of network systems in high demand and changing environments.

Therefore, developing a method that can adjust and optimize network resource configuration in real time to adapt to changes in network conditions and business needs has become the key to improving network management efficiency and service quality.

Summary of the invention

Based on the above objectives, the present invention provides a network integrated dynamic resource routing management system and method.

A network integrated dynamic resource routing management system, comprising a resource monitoring module, a data acquisition and fusion module, an intelligent prediction and analysis module, a decision support and optimization module, a dynamic resource scheduling module and a feedback optimization module; wherein:

Resource monitoring module: used to monitor and collect network resource usage in real time, including bandwidth, storage and computing power, and generate resource usage reports;

Data acquisition and fusion module: pre-processes the collected network resource data to eliminate noise and redundant information, and simultaneously performs multi-source data fusion to improve data accuracy and consistency;

Intelligent prediction and analysis module: Based on the pre-processed data provided by the data acquisition and fusion module, it performs multi-dimensional analysis, adopts a hybrid model combining convolutional neural network and long short-term memory network, predicts resource demand and traffic change trends, generates resource optimization plans, and passes the optimization plans to the decision support optimization module;

Decision support optimization module: Use the resource optimization solution provided by the intelligent prediction analysis module to dynamically adjust resource allocation strategies and routing paths in combination with current network conditions and business needs;

Dynamic resource scheduling module: responsible for executing the resource allocation and routing adjustment schemes generated by the decision support optimization module, adjusting network resource configuration in real time through the dynamic resource scheduling algorithm, and monitoring the execution effect to ensure network stability and service quality during the adjustment process;

Feedback optimization module: Continuously collects the adjusted network operation data provided by the dynamic resource scheduling module, analyzes the execution effect, and adjusts and optimizes the hybrid model of the intelligent prediction analysis module based on the analysis results to form a closed-loop optimization mechanism; and optimizes resource scheduling and routing strategies through continuous trial and error and learning to improve the overall system performance.

Further, the resource monitoring module includes a bandwidth monitoring unit, a storage monitoring unit and a computing capacity monitoring unit; wherein:

Bandwidth monitoring unit: used to monitor the usage of network bandwidth in real time, and generate bandwidth usage reports by monitoring the data transmission rate, packet loss rate and delay parameters of the network interface;

Storage monitoring unit: used to monitor the usage of network storage resources in real time, and generate storage usage reports by monitoring the capacity utilization, read and write rates, and I/O operation frequency parameters of storage devices;

Computing capacity monitoring unit: used to monitor the usage of network computing resources in real time, and generate computing capacity usage reports by monitoring the CPU load, memory usage and computing task queue length parameters of servers and computing nodes;

Resource usage report generation unit: responsible for integrating the data provided by the bandwidth monitoring unit, storage monitoring unit and computing power monitoring unit to generate a comprehensive resource usage report; the report content includes the current usage, historical usage trends and predicted usage of various resources.

Furthermore, the data acquisition and fusion module includes a data preprocessing unit and a multi-source data fusion unit; wherein:

Data preprocessing unit: used to preprocess the real-time data collected from different network nodes, and eliminate the noise and redundant information in the data through the steps of noise filtering, data cleaning and duplicate data deletion; the noise filtering uses a signal processing algorithm to filter the data noise and remove abnormal values and mutation data; data cleaning uses data cleaning technology to correct missing values and erroneous values in the data; duplicate data deletion uses a hash algorithm to identify and delete duplicate items in the data;

Multi-source data fusion unit: used to perform multi-source data fusion on pre-processed data, and improve data accuracy and consistency through data alignment, data transformation and data integration steps. Among them, data alignment is specifically based on timestamps to align data from different network nodes to ensure the consistency of data in the same time dimension; data transformation is to transform and standardize data according to the characteristics of different data sources, using a unified data format and unit; data integration uses data fusion algorithms to integrate data from different sources into a unified data set to generate a comprehensive data view.

Furthermore, the intelligent prediction and analysis module includes a feature extraction unit, a prediction and analysis unit, and a resource optimization solution generation unit; wherein:

Feature extraction unit: used to perform multi-dimensional analysis based on the pre-processed data provided by the data acquisition and fusion module, and use convolutional neural network to extract features from the comprehensive data view. Specifically, the pre-processed data provided by the data acquisition and fusion module is first received and input into the convolutional neural network; then the local features of the data are extracted through multiple convolutional layers to capture the spatial relationship and pattern in the data; finally, the convolutional features are reduced in dimension and screened through the pooling layer to retain the predetermined features;

Prediction and analysis unit: Based on the features extracted by the feature extraction unit, the long short-term memory network is used to perform time series prediction, and the attention mechanism is combined to improve the accuracy and response speed of the prediction, predict resource demand and traffic change trends, and generate prediction results including future resource demand and traffic change trends;

Resource optimization solution generation unit: Generates a resource optimization solution based on the prediction results provided by the prediction analysis unit. Specifically, the output data of the prediction analysis unit is first integrated to form a complete prediction view; then an optimization algorithm is used to calculate the optimal resource allocation and routing strategy based on the prediction view to generate a resource optimization solution.

Furthermore, the prediction analysis unit includes:

Feature input: Receive the features output by the feature extraction unit, represent the feature data as a time series and input it into the long short-term memory network, which is specifically expressed as: Among them, _Xt is the feature vector at time point t, and _xti is the i-th feature data at time point t;

Sequence learning: Learn the temporal dependencies of time series data through the LSTM layer to capture the long-term and short-term trends of the data;

Introducing the attention mechanism: The attention mechanism is used to enhance the model's attention to the predetermined features in the time series data, automatically identifying and focusing on the time steps or features that have a greater impact on the prediction results. The calculation formula of the attention mechanism is:

e _t,j = v ^T ·tanh(W _e ·[h _t ;h _j ]+b _e );

Among them, e _t,j is the attention score of time point t and time point j, α _t,j is the attention weight of time point t and time point j, is the weighted hidden state at time point t, v is the attention weight vector, _We is the attention weight matrix, _be is the attention bias, _ht is the hidden state at time point t, and _hj is the hidden state at time point j;

Prediction output: hidden state based on attention weighting The prediction results are generated through the fully connected layer, including future resource requirements and traffic change trends. The formula is: in, is the predicted value at time point t+1, W _y is the weight matrix of the fully connected layer, and _by is the bias of the fully connected layer;

Result transmission: The generated prediction results, including future resource demand and traffic change trends, are transmitted to the resource optimization solution generation unit.

Furthermore, the resource optimization solution generating unit includes:

Data integration: Integrate the output data of the prediction analysis unit to form a complete prediction view, expressed as: Where _Pt is the predicted view at time point t, is the predicted value at time point t+i;

Objective function definition: Define the objective function of resource optimization, including maximizing resource utilization and optimizing service quality. The objective function is expressed as: Among them, U is the optimization objective function, R _t+i is the resource utilization at time point t+i, Q _t+i is the service quality loss at time point t+i, α and β are weight coefficients;

Constraint setting: Set the resource allocation constraints to ensure that the resource allocation is within a reasonable range. The constraint formula is: R _min ≤ _{R t+i} ≤ R _max , where R _min is the lower limit of resource utilization and R _max is the upper limit of resource utilization;

Optimization calculation: Genetic algorithm is used for optimization calculation to generate the best resource allocation and routing strategy;

Solution output: The optimal resource allocation and routing strategy is output as a resource optimization solution and transmitted to the decision support optimization module.

Further, the decision support optimization module includes a network status assessment unit, a service demand assessment unit and a strategy adjustment unit; wherein:

Network status evaluation unit: used to evaluate the current network status in real time, including parameters of bandwidth utilization, storage utilization, and computing power utilization. The network status evaluation formula is: Wherein, _Nt is the network status evaluation result at time point t, _Bt is the bandwidth utilization, _St is the storage utilization, and _Ct is the computing power utilization;

Business demand assessment unit: used to assess current business needs, including the priority of each business application, real-time traffic demand and service quality requirements; the business demand assessment formula is: Where _Dt is the service demand assessment result at time point t, _Pt is the priority of each service application, _Tt is the real-time traffic demand, and _Qt is the service quality requirement;

Policy adjustment unit: used to dynamically adjust resource allocation strategy and routing path based on resource optimization scheme provided by intelligent prediction analysis module in combination with network status assessment results and business demand assessment results; specifically includes the following steps:

First, integrate the network status assessment results and business demand assessment results to form a comprehensive assessment view, which is specifically expressed as follows: Where E _t is the comprehensive evaluation view at time point t;

Then, based on the comprehensive evaluation view and resource optimization scheme, the optimal resource allocation strategy and routing path are calculated. The strategy calculation formula is: S _t = argmax(α· _UR,t +β· _UQ,t -γ· _Ct ), where S _t is the resource allocation strategy at time point t, _UR,t is the resource utilization benefit, UQ _,t is the service quality benefit, C _t is the adjustment cost, and α, β and γ are weight coefficients;

Finally, the calculated optimal resource allocation strategy and routing path are output as adjustment solutions and transmitted to the dynamic resource scheduling module for executing resource allocation and routing adjustment.

Furthermore, the dynamic resource scheduling module includes a resource configuration unit, a routing adjustment unit and an execution monitoring unit; wherein:

Resource configuration unit: used to receive the resource allocation plan generated by the decision support optimization module, and adjust the network resource configuration in real time according to the plan. The resource configuration formula is: R _t+1 =R _t +ΔR _t , where R _t+1 is the resource configuration at time point t+1, R _t is the resource configuration at time point t, and ΔR _t is the resource adjustment amount calculated according to the resource allocation plan;

Route adjustment unit: used to receive the route adjustment plan generated by the decision support optimization module and dynamically adjust the network routing path according to the plan. The route adjustment formula is:

P _t+1 = argmin( _CP,t+1 + β·DP _,t+1 ), where P _t+1 is the routing path at time point t+1, _CP,t+1 is the routing cost at time point t+1, _DP,t+1 is the routing delay at time point t+1, and β is the weight coefficient;

Execution monitoring unit: used to monitor the execution effect of resource configuration and routing adjustment to ensure network stability and service quality during the adjustment process. The execution monitoring formula is: Wherein, M _t is the execution monitoring index at time point t, Qi _,t is the i-th service quality parameter at time point t, and n is the number of service quality parameters.

Furthermore, the feedback optimization module includes a data collection unit, a model adjustment unit and a strategy optimization unit; wherein:

Data collection unit: used to continuously collect the adjusted network operation data provided by the dynamic resource scheduling module, including resource utilization, service quality and network performance parameters;

Model adjustment unit: used to adjust and optimize the hybrid model of the intelligent prediction analysis module based on the operation data provided by the data collection unit. Specifically, the error between the prediction result and the actual operation data is calculated first. The formula is: Where _{Et is} the prediction error at time point t, is the predicted value at time point t+i, y _t+i is the actual value at time point t+i, and n is the number of predicted data points; then based on the calculated error, the parameters of the hybrid model are adjusted through the optimization algorithm, and the model is updated to improve the prediction accuracy;

Policy optimization unit: used to optimize resource scheduling and routing strategies through continuous trial and error and learning. Specifically, the reinforcement learning algorithm is first applied to learn the optimal resource scheduling and routing strategies through trial and error. The reward function of reinforcement learning is expressed as: R _t = α· _UR,t + β·U _Q,t -γ·C _t , where R _t is the reward value at time point t, _UR,t is the resource utilization benefit, U _Q,t is the service quality benefit, C _t is the adjustment cost, and α, β and γ are weight coefficients; then, based on the feedback of the reward function, the resource scheduling and routing strategies are adjusted to improve the overall system performance.

A network integrated dynamic resource routing management method comprises the following steps:

S1: Monitor the usage of network resources in real time, including bandwidth, storage, and computing power, and generate resource usage reports;

S2: Collect real-time data from different network nodes, pre-process the collected data to eliminate noise and redundant information, and perform multi-source data fusion to improve data quality;

S3: A hybrid model combining convolutional neural networks and long short-term memory networks, combined with an attention mechanism, performs multi-dimensional analysis on the data preprocessed by S2 to predict resource demand and traffic change trends and generate resource optimization plans;

S4: Use the resource optimization solution of S3 to dynamically adjust resource allocation strategies and routing paths based on current network conditions and business needs;

S5: Execute the resource allocation and routing adjustment plan generated in S4, adjust the network resource configuration in real time through the dynamic resource scheduling algorithm, and monitor the execution effect;

S6: Continue to collect network operation data adjusted by S5 and analyze the execution effect. Adjust and optimize the hybrid model in S3 based on the analysis results to form a closed-loop optimization mechanism. Optimize resource scheduling and routing strategies through continuous trial and error and learning.

Beneficial effects of the present invention:

The present invention significantly improves the utilization efficiency and response speed of network resources by implementing an intelligent resource management method. First, by real-time monitoring of the use of network resources and combining advanced data acquisition and fusion technology, the high accuracy and consistency of data are ensured, thereby making resource prediction and analysis more accurate. The intelligent prediction and analysis module uses a hybrid convolutional neural network and long short-term memory network model to effectively predict resource demand and traffic change trends, especially when dealing with emergencies and abnormal traffic, showing excellent prediction capabilities. In addition, the system also continuously optimizes resource allocation and routing strategies through continuous learning and trial and error, further improving the adaptability and flexibility of the system.

The present invention, through the collaborative work of the decision support optimization module and the dynamic resource scheduling module, timely adjusts the resource allocation and routing path according to the output of the intelligent prediction and analysis module. This dynamic adjustment mechanism not only improves the service quality of the network, but also greatly enhances the resilience of the network in the face of demand fluctuations and uncertain factors. By realizing closed-loop optimization, the feedback optimization module can adjust the prediction model based on actual operation results to ensure continuous improvement of performance and efficiency in long-term operation. In summary, the present invention can effectively solve the problems of static and slow response of network resource management in the prior art, and provides an innovative and practical solution for resource management in large-scale data centers and complex network environments.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a schematic diagram of a dynamic resource routing management system according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a network integrated dynamic resource routing management method according to an embodiment of the present invention.

DETAILED DESCRIPTION

In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. The words "include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

As shown in FIG1-2, a network integrated dynamic resource routing management system includes a resource monitoring module, a data acquisition and fusion module, an intelligent prediction and analysis module, a decision support and optimization module, a dynamic resource scheduling module, and a feedback optimization module; wherein:

Resource monitoring module: used to monitor and collect network resource usage in real time, including bandwidth, storage and computing power, and generate resource usage reports;

Data acquisition and fusion module: pre-processes the collected network resource data to eliminate noise and redundant information, and simultaneously performs multi-source data fusion to improve data accuracy and consistency;

Intelligent prediction and analysis module: Based on the pre-processed data provided by the data acquisition and fusion module, it performs multi-dimensional analysis and adopts a hybrid model combining convolutional neural network (CNN) and long short-term memory network (LSTM) to predict resource demand and traffic change trends and generate resource optimization plans. This hybrid model pays special attention to the prediction of sudden traffic and abnormal events, and passes the optimization plan to the decision support optimization module;

Decision support optimization module: Utilizes the resource optimization solution provided by the intelligent prediction analysis module, combines the current network status and business needs, and dynamically adjusts the resource allocation strategy and routing path to ensure efficient use of resources;

Dynamic resource scheduling module: responsible for executing the resource allocation and routing adjustment schemes generated by the decision support optimization module, adjusting network resource configuration in real time through the dynamic resource scheduling algorithm, and monitoring the execution effect to ensure network stability and service quality during the adjustment process;

Feedback optimization module: Continuously collects the adjusted network operation data provided by the dynamic resource scheduling module, analyzes the execution effect, and adjusts and optimizes the hybrid model of the intelligent prediction analysis module based on the analysis results to form a closed-loop optimization mechanism; and optimizes resource scheduling and routing strategies through continuous trial and error and learning to improve the overall system performance.

The resource monitoring module includes a bandwidth monitoring unit, a storage monitoring unit and a computing power monitoring unit; wherein:

Bandwidth monitoring unit: used to monitor the usage of network bandwidth in real time, and generate bandwidth usage reports by monitoring the data transmission rate, packet loss rate and delay parameters of the network interface to ensure the accuracy and timeliness of bandwidth usage;

Storage monitoring unit: used to monitor the usage of network storage resources in real time, and generate storage usage reports by monitoring the capacity utilization, read and write rates, and I/O operation frequency parameters of storage devices;

Computing capacity monitoring unit: used to monitor the usage of network computing resources in real time, and generate computing capacity usage reports by monitoring the CPU load, memory usage and computing task queue length parameters of servers and computing nodes;

Resource usage report generation unit: responsible for integrating the data provided by the bandwidth monitoring unit, storage monitoring unit and computing power monitoring unit to generate a comprehensive resource usage report; the report content includes the current usage, historical usage trends and predicted usage of various resources, and the report is transmitted to the data acquisition and fusion module; the resource monitoring module realizes comprehensive monitoring of network resource usage through the collaborative work of the bandwidth monitoring unit, storage monitoring unit and computing power monitoring unit, ensures the accuracy and timeliness of resource usage data, and provides comprehensive resource usage analysis through the resource usage report generation unit, providing a reliable data foundation for the intelligent prediction analysis module and the decision support optimization module.

The data acquisition and fusion module includes a data preprocessing unit and a multi-source data fusion unit; wherein:

Data preprocessing unit: used to preprocess the real-time data collected from different network nodes, and eliminate the noise and redundant information in the data through the steps of noise filtering, data cleaning and deduplication. Noise filtering uses signal processing algorithms to filter the noise of data and remove abnormal values and mutation data. Data cleaning uses data cleaning technology to correct missing values and error values in the data to ensure the integrity and accuracy of the data. Deduplication uses hash algorithms to identify and delete duplicate items in the data to ensure the uniqueness of the data.

Multi-source data fusion unit: used to fuse multi-source data after preprocessing, and improve the accuracy and consistency of data through data alignment, data transformation and data integration steps. Among them, data alignment is specifically based on timestamps to align data from different network nodes to ensure the consistency of data in the same time dimension; data transformation is to transform and standardize data according to the characteristics of different data sources, using a unified data format and unit; data integration uses a data fusion algorithm to integrate data from different sources into a unified data set, generate a comprehensive data view, and transmit the results to the intelligent prediction analysis module for further analysis and optimization; through the detailed design and functional division of the above data acquisition and fusion module, efficient preprocessing and multi-source data fusion of network resource data can be achieved. The various units of this module work together to eliminate noise and redundant information in the data through steps such as noise filtering, data cleaning and deduplication, and improve the accuracy and consistency of data through steps such as data alignment, data transformation and data integration, providing a reliable data foundation for intelligent prediction analysis and decision support, and enhancing the overall performance and stability of the system.

The intelligent prediction and analysis module includes a feature extraction unit, a prediction and analysis unit, and a resource optimization solution generation unit; wherein:

Feature extraction unit: used to perform multi-dimensional analysis based on the pre-processed data provided by the data acquisition and fusion module, and use convolutional neural network to extract features from the comprehensive data view. Specifically, the pre-processed data provided by the data acquisition and fusion module is first received and input into the convolutional neural network; then the local features of the data are extracted through multiple convolutional layers to capture the spatial relationship and pattern in the data; finally, the convolutional features are reduced in dimension and screened through the pooling layer to retain the predetermined features;

Prediction and analysis unit: Based on the features extracted by the feature extraction unit, a long short-term memory network is used to perform time series prediction, and an attention mechanism is combined to improve the accuracy and response speed of the prediction, and to predict resource demand and traffic change trends. The specific steps include receiving the features output by the feature extraction unit and inputting them into the long short-term memory network, learning the time dependency of the data through multiple LSTM layers, capturing the long-term and short-term trends of the data, and introducing an attention mechanism to automatically identify and focus on the time steps or features that have a greater impact on the prediction results, thereby improving the accuracy of the prediction and generating prediction results including future resource demand and traffic change trends.

Resource optimization solution generation unit: Generates resource optimization solutions based on the prediction results provided by the prediction analysis unit. Specifically, the output data of the prediction analysis unit is first integrated to form a complete prediction view. Then, the optimization algorithm is used to calculate the optimal resource allocation and routing strategy according to the prediction view, generate a resource optimization solution, and transmit the solution to the decision support optimization module for further resource allocation and routing adjustment. Through the detailed design and functional division of the above-mentioned intelligent prediction analysis module, efficient multi-dimensional analysis can be performed based on pre-processed data. A hybrid model combining convolutional neural network (CNN) and long short-term memory network (LSTM) is adopted, and an attention mechanism is introduced to perform feature extraction and time series prediction on the comprehensive data view, accurately predict resource demand and traffic change trends, and generate resource optimization solutions to ensure efficient resource utilization and network performance optimization.

The predictive analysis unit includes:

Feature input: Receive the features output by the feature extraction unit, represent the feature data as a time series and input it into the long short-term memory network (LSTM), which is specifically expressed as: Among them, _Xt is the feature vector at time point t, and _xti is the i-th feature data at time point t;

Sequence learning: The time dependency of time series data is learned through the LSTM layer to capture the long-term and short-term trends of the data; the calculation formula of LSTM is:

i _t =σ(W _ix ·x _t +W _ih ·h _t-1 +b _i );

f _t =σ(W _fx ·x _t +W _fh ·h _t-1 +b _f );

o _t =σ(W _ox ·x _t +W _oh ·h _t-1 +b _o );

c _t = f _t ·c _t-1 +i _t ·tanh(W _cx ·x _t +W _ch ·h _t-1 +b _c );

h _t = o _t ·tanh( _ct ); where _it is the input gate activation value, f _is the forget gate activation value, o _is the output gate activation value, _c is the memory unit state, _h is the hidden state, σ is the sigmoid function, tanh is the hyperbolic tangent function, _Wix is the weight matrix input to the input gate, _Wih is the weight matrix from the hidden state to the input gate, _Wfx is the weight matrix input to the forget gate, _Wfh is the weight matrix from the hidden state to the forget gate, _Wox is the weight matrix input to the output gate, _Woh is the weight matrix from the hidden state to the output gate, _Wcx is the weight matrix input to the memory unit, _Wch is the weight matrix from the hidden state to the memory unit, _bi is the bias of the input gate, _bf is the bias of the forget gate, _bo is the bias of the output gate, and _bc is the bias of the memory unit;

Introducing the attention mechanism: The attention mechanism is used to enhance the model's attention to the predetermined features in the time series data, automatically identifying and focusing on the time steps or features that have a greater impact on the prediction results. The calculation formula of the attention mechanism is:

e _t,j = v ^T ·tanh(W _e ·[h _t ;h _j ]+b _e );

Among them, e _t,j is the attention score of time point t and time point j, α _t,j is the attention weight of time point t and time point j, is the weighted hidden state at time point t, v is the attention weight vector, _We is the attention weight matrix, _be is the attention bias, _ht is the hidden state at time point t, and _hj is the hidden state at time point j;

Prediction output: hidden state based on attention weighting The prediction results are generated through the fully connected layer, including future resource requirements and traffic change trends. The formula is: in, is the predicted value at time point t+1, W _y is the weight matrix of the fully connected layer, and _by is the bias of the fully connected layer;

Result transmission: The generated prediction results, including future resource requirements and traffic change trends, are transmitted to the resource optimization plan generation unit for further resource allocation and routing adjustment. Through the detailed design and functional division of the above-mentioned prediction and analysis unit, it is possible to use the long short-term memory network (LSTM) for efficient time series prediction based on the features extracted by the feature extraction unit, and combine the attention mechanism to improve the accuracy and response speed of the prediction. By generating prediction results including future resource requirements and traffic change trends, it provides accurate and reliable data support for the resource optimization plan, improves the prediction accuracy and response capability of the system, and enhances the overall system performance and stability.

The resource optimization solution generation unit includes:

Data integration: Integrate the output data of the prediction analysis unit to form a complete prediction view, expressed as: Where _Pt is the predicted view at time point t, is the predicted value at time point t+i;

Objective function definition: Define the objective function of resource optimization, including maximizing resource utilization and optimizing service quality. The objective function is expressed as: Among them, U is the optimization objective function, R _t+i is the resource utilization at time point t+i, Q _t+i is the service quality loss at time point t+i, α and β are weight coefficients;

Constraint setting: Set the resource allocation constraints to ensure that the resource allocation is within a reasonable range. The constraint formula is: R _min ≤ _{R t+i} ≤ R _max , where R _min is the lower limit of resource utilization and R _max is the upper limit of resource utilization;

Optimization calculation: Genetic algorithm is used for optimization calculation to generate the optimal resource allocation and routing strategy; the calculation process of genetic algorithm includes the following steps:

Step 1: Randomly generate a set of initial resource allocation strategies as the population, expressed as: Among them, S ₀ is the initial population, s _i is the i-th resource allocation strategy, and m is the population size;

Step 2: Calculate the fitness of each resource allocation strategy based on the objective function. The formula is: Where F(s _i ) is the fitness value of the i-th strategy, R _si,t+j and Q _si,t+j are the resource utilization and service quality loss of the i-th strategy at time point t+j, respectively;

Step 3: Select a strategy with high fitness according to the fitness value for reproduction to generate the next generation population;

Step 4: Perform cross-operation on the selected strategies to generate a new resource allocation strategy. The formula is: Among them, s _p and s _q are parent strategies, and s _new is the new resource allocation strategy;

Step 5: Perform mutation operation on the newly generated strategy to increase the diversity of the population, expressed as: s _mutated = mutate(s _new ), where s _mutated is the mutated strategy;

Step 6: Repeat steps 2 to 5 until the termination condition is met and the optimal resource allocation strategy is obtained.

Solution output: The optimal resource allocation and routing strategy is output as a resource optimization solution and transmitted to the decision support optimization module for further resource allocation and routing adjustment; the resource optimization solution generation unit generates the optimal resource allocation and routing strategy through data integration, objective function definition, constraint setting and genetic algorithm optimization calculation, providing a reliable resource optimization solution for the decision support optimization module.

The decision support optimization module includes a network status assessment unit, a business demand assessment unit and a strategy adjustment unit; among which:

Network status evaluation unit: used to evaluate the current network status in real time, including parameters of bandwidth utilization, storage utilization, and computing power utilization. The network status evaluation formula is: Wherein, _Nt is the network status evaluation result at time point t, _Bt is the bandwidth utilization, _St is the storage utilization, and _Ct is the computing power utilization;

Business demand assessment unit: used to assess current business needs, including the priority of each business application, real-time traffic demand and service quality requirements; the business demand assessment formula is: Where _Dt is the service demand assessment result at time point t, _Pt is the priority of each service application, _Tt is the real-time traffic demand, and _Qt is the service quality requirement;

Policy adjustment unit: used to dynamically adjust resource allocation strategy and routing path based on resource optimization scheme provided by intelligent prediction analysis module in combination with network status assessment results and business demand assessment results; specifically includes the following steps:

First, integrate the network status assessment results and business demand assessment results to form a comprehensive assessment view, which is specifically expressed as follows: Where, E _t is the comprehensive evaluation view at time point t;

Then, based on the comprehensive evaluation view and resource optimization scheme, the optimal resource allocation strategy and routing path are calculated. The strategy calculation formula is: S _t = argmax(α· _UR,t +β· _UQ,t -γ· _Ct ), where S _t is the resource allocation strategy at time point t, _UR,t is the resource utilization benefit, UQ _,t is the service quality benefit, C _t is the adjustment cost, and α, β and γ are weight coefficients;

Finally, the calculated optimal resource allocation strategy and routing path are output as adjustment plans and transmitted to the dynamic resource scheduling module for executing resource allocation and routing adjustments. The decision support optimization module uses the resource optimization plan provided by the intelligent prediction analysis module through the collaborative work of the network status assessment unit, the business demand assessment unit and the policy adjustment unit, and combines the current network status and business needs to dynamically adjust the resource allocation strategy and routing path, ensuring efficient utilization of resources and optimization of service quality.

The dynamic resource scheduling module includes a resource configuration unit, a routing adjustment unit and an execution monitoring unit; wherein:

Resource configuration unit: used to receive the resource allocation plan generated by the decision support optimization module, and adjust the network resource configuration in real time according to the plan. The resource configuration formula is: R _t+1 =R _t +ΔR _t , where R _t+1 is the resource configuration at time point t+1, R _t is the resource configuration at time point t, and ΔR _t is the resource adjustment amount calculated according to the resource allocation plan;

Route adjustment unit: used to receive the route adjustment plan generated by the decision support optimization module and dynamically adjust the network routing path according to the plan. The route adjustment formula is:

P _t+1 = argmin( _CP,t+1 + β·DP _,t+1 ), where P _t+1 is the routing path at time point t+1, _CP,t+1 is the routing cost at time point t+1, _DP,t+1 is the routing delay at time point t+1, and β is the weight coefficient;

Execution monitoring unit: used to monitor the execution effect of resource configuration and routing adjustment to ensure network stability and service quality during the adjustment process. The execution monitoring formula is: Among them, _Mt is the execution monitoring indicator at time point t, Qi _,t is the i-th service quality parameter at time point t, and n is the number of service quality parameters. Through the detailed design and functional division of the above dynamic resource scheduling module, the resource allocation and routing adjustment schemes generated by the decision support optimization module can be executed. Specifically, the network resource configuration is adjusted in real time through the dynamic resource scheduling algorithm. Through the collaborative work of resource configuration, routing adjustment and execution monitoring, the efficient utilization of resources and the stability of the network are ensured, which provides accurate and reliable execution support for dynamic resource management and improves the overall performance and stability of the system.

The feedback optimization module includes a data collection unit, a model adjustment unit and a strategy optimization unit; wherein:

Data collection unit: used to continuously collect the adjusted network operation data provided by the dynamic resource scheduling module, including resource utilization, service quality and network performance parameters. The collected data is used for subsequent analysis and optimization;

Model adjustment unit: used to adjust and optimize the hybrid model of the intelligent prediction analysis module based on the operation data provided by the data collection unit. Specifically, the error between the prediction result and the actual operation data is calculated first. The formula is: Where _{Et is} the prediction error at time point t, is the predicted value at time point t+i, y _t+i is the actual value at time point t+i, and n is the number of predicted data points; then, based on the calculated error, the parameters of the hybrid model are adjusted through an optimization algorithm (such as a gradient descent algorithm) to update the model to improve the prediction accuracy;

Policy optimization unit: used to optimize resource scheduling and routing strategies through continuous trial and error and learning. Specifically, the reinforcement learning algorithm is first applied to learn the optimal resource scheduling and routing strategies through the trial and error process. The reward function of reinforcement learning is expressed as: R _t = α· _UR,t + β·U _Q,t -γ·C _t , where R _t is the reward value at time point t, _UR,t is the resource utilization benefit, U _Q,t is the service quality benefit, C _t is the adjustment cost, and α, β and γ are weight coefficients; then, based on the feedback of the reward function, the resource scheduling and routing strategies are adjusted to improve the overall system performance; through the detailed design and functional division of the above feedback optimization module, the hybrid model of the intelligent prediction analysis module can be efficiently adjusted and optimized based on the analysis results to form a closed-loop optimization mechanism. Through the collaborative work of data collection, error calculation, model updating and trial and error learning, the resource scheduling and routing strategies are optimized to ensure the long-term stability and efficiency of the system. The collaborative work of each unit improves the system's adaptability and overall performance, and provides continuous optimization support for dynamic resource management.

A network integrated dynamic resource routing management method comprises the following steps:

S1: Monitor the usage of network resources in real time, including bandwidth, storage, and computing power, and generate resource usage reports;

S2: Collect real-time data from different network nodes, pre-process the collected data to eliminate noise and redundant information, and perform multi-source data fusion to improve data quality;

S3: A hybrid model combining convolutional neural networks and long short-term memory networks, combined with an attention mechanism, performs multi-dimensional analysis on the data pre-processed by S2 to predict resource demand and traffic change trends and generate resource optimization plans;

S4: Use the resource optimization solution of S3 to dynamically adjust resource allocation strategies and routing paths based on current network conditions and business needs;

S5: Execute the resource allocation and routing adjustment plan generated in S4, adjust the network resource configuration in real time through the dynamic resource scheduling algorithm, and monitor the execution effect to ensure network stability and service quality during the adjustment process;

S6: Continue to collect network operation data adjusted by S5 and analyze the execution effect. Adjust and optimize the hybrid model in S3 based on the analysis results to form a closed-loop optimization mechanism. Optimize resource scheduling and routing strategies through continuous trial and error and learning.

The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims (10)

1. The network integrated dynamic resource route management system is characterized by comprising a resource monitoring module, a data acquisition fusion module, an intelligent prediction analysis module, a decision support optimization module, a dynamic resource scheduling module and a feedback optimization module; wherein:

And a resource monitoring module: the method is used for monitoring and collecting the use condition of network resources in real time, including bandwidth, storage and computing capacity, and generating a resource use report;

And a data acquisition fusion module: preprocessing the acquired network resource data to eliminate noise and redundant information, and simultaneously fusing multi-source data to improve the accuracy and consistency of the data;

Intelligent predictive analysis module: based on the preprocessing data provided by the data acquisition fusion module, multidimensional analysis is carried out, a mixed model combining a convolutional neural network and a long-short-term memory network is adopted to predict resource requirements and flow change trend, a resource optimization scheme is generated, and the optimization scheme is transmitted to a decision support optimization module;

decision support optimization module: the resource optimization scheme provided by the intelligent predictive analysis module is utilized to dynamically adjust the resource allocation strategy and the routing path by combining the current network condition and the service requirement;

Dynamic resource scheduling module: the resource allocation and route adjustment scheme generated by the decision support optimization module is executed, the network resource allocation is adjusted in real time through a dynamic resource scheduling algorithm, the execution effect is monitored, and the network stability and the service quality in the adjustment process are ensured;

And a feedback optimization module: continuously collecting the adjusted network operation data provided by the dynamic resource scheduling module, analyzing the execution effect, and adjusting and optimizing the hybrid model of the intelligent prediction analysis module according to the analysis result to form a closed-loop optimization mechanism; and the resource scheduling and routing strategies are optimized through continuous trial and error and learning so as to improve the performance of the whole system.

2. The network integrated dynamic resource routing management system of claim 1, wherein the resource monitoring module comprises a bandwidth monitoring unit, a storage monitoring unit, and a computing power monitoring unit; wherein:

Bandwidth monitoring unit: the method comprises the steps of monitoring the use condition of network bandwidth in real time, and generating a bandwidth use report by monitoring parameters of data transmission rate, packet loss rate and delay of a network interface;

and a storage monitoring unit: the method comprises the steps of monitoring the use condition of network storage resources in real time, and generating a storage use report by monitoring the capacity utilization rate, the read-write rate and the parameters of I/O operation frequency of storage equipment;

Computing power monitoring unit: the method comprises the steps of monitoring the use condition of network computing resources in real time, and generating a computing capacity use report by monitoring CPU loads, memory use rates and computing task queue length parameters of a server and a computing node;

resource usage report generation unit: the system is responsible for integrating data provided by a bandwidth monitoring unit, a storage monitoring unit and a computing capacity monitoring unit to generate a comprehensive resource use report; the report content includes current usage, historical usage trends, and predicted usage of various types of resources.

3. The network integrated dynamic resource routing management system of claim 1, wherein the data acquisition fusion module comprises a data preprocessing unit and a multi-source data fusion unit; wherein:

A data preprocessing unit: the method comprises the steps of preprocessing real-time data acquired from different network nodes, and eliminating noise and redundant information in the data through noise filtering, data cleaning and repeated data deleting; the noise filtering uses a signal processing algorithm to perform noise filtering on the data, and abnormal values and abrupt change data are removed; data cleaning utilizes a data cleaning technology to correct missing values and error values in data; the repeated data deletion adopts a hash algorithm to identify and delete repeated items in the data;

A multi-source data fusion unit: the method comprises the steps of carrying out multi-source data fusion on preprocessed data, and improving the accuracy and consistency of the data through data alignment, data transformation and data integration, wherein the data alignment is specifically based on time stamps, the data from different network nodes are aligned, and the consistency of the data in the same time dimension is ensured; the data transformation is to transform and normalize the data by adopting a unified data format and unit according to the characteristics of different data sources; and integrating the data from different sources into a unified data set by using a data fusion algorithm to generate a comprehensive data view.

4. The network integrated dynamic resource routing management system of claim 3, wherein the intelligent predictive analysis module comprises a feature extraction unit, a predictive analysis unit and a resource optimization scheme generation unit; wherein:

feature extraction unit: the method comprises the steps of carrying out multidimensional analysis on the preprocessed data provided by a data acquisition fusion module, carrying out feature extraction on a comprehensive data view by adopting a convolutional neural network, specifically, firstly receiving the preprocessed data provided by the data acquisition fusion module, and inputting the preprocessed data into the convolutional neural network; then extracting local features of the data through a plurality of convolution layers, and capturing spatial relations and modes in the data; finally, performing dimension reduction and screening on the convolved characteristics through a pooling layer, and reserving preset characteristics;

Prediction analysis unit: based on the features extracted by the feature extraction unit, a long-short-period memory network is adopted to conduct time sequence prediction, and an attention mechanism is combined to promote the accuracy and response speed of the prediction, predict the resource demand and the flow change trend, and generate a prediction result comprising the future resource demand and the flow change trend;

Resource optimization scheme generation unit: generating a resource optimization scheme based on the prediction result provided by the prediction analysis unit, and specifically integrating output data of the prediction analysis unit to form a complete prediction view; and then, calculating the optimal resource allocation and routing strategy according to the predicted view by adopting an optimization algorithm, and generating a resource optimization scheme.

5. The network integrated dynamic resource routing management system of claim 4, wherein said predictive analysis unit comprises:

Feature input: the method comprises the steps of receiving the characteristics output by a characteristic extraction unit, and representing characteristic data as a time sequence input long-short-period memory network, wherein the method specifically comprises the following steps: x _t＝{x_t1 ^,x_t2 ^,…^,x_tn }, wherein X _t is the feature vector at time point t, and X _ti is the ith feature data at time point t;

sequence learning: the LSTM layer is used for learning the time dependency relationship of the time sequence data, and capturing long-term and short-term trends of the data;

attention introducing mechanism: and (3) automatically identifying and focusing on time steps or features with larger influence on a prediction result by utilizing the attention mechanism enhancement model to focus on the preset features in the time sequence data, wherein a calculation formula of the attention mechanism is as follows:

e_t,j＝v^T·tanh(W_e·[h_t;h_j]+b_e)；

Wherein e _t,j is the attention score at time t and time j, alpha _t,j is the attention weight at time t and time j, For the weighted hidden state at time t, v is the attention weight vector, W _e is the attention weight matrix, b _e is the attention bias, h _t is the hidden state at time t, and h _j is the hidden state at time j;

prediction output: hiding state weighted based on attention Generating a prediction result through the full connection layer, wherein the prediction result comprises future resource demand and flow change trend, and the formula is as follows: wherein, As the predicted value of the time point t+1, W _y is the weight matrix of the full-connection layer, and b _y is the bias of the full-connection layer;

And (3) transmitting results: and transmitting the generated prediction result, including future resource demand and flow change trend, to a resource optimization scheme generating unit.

6. The network integrated dynamic resource route management system of claim 5, wherein said resource optimization scheme generation unit comprises:

data integration: integrating the output data of the predictive analysis unit to form a complete predictive view, expressed as: Wherein P _t is the predicted view at time point t, A predicted value for time t+i;

The objective function definition: defining an objective function of resource optimization, including resource utilization maximization and quality of service optimization, the objective function expressed as: Wherein U is an optimization objective function, R _t+i is the resource utilization rate of a time point t+i, Q _t+i is the service quality loss of the time point t+i, and alpha and beta are weight coefficients;

Constraint condition setting: setting constraint conditions of resource allocation, and ensuring that the resource allocation is within a reasonable range, wherein the constraint condition formula is as follows: r _min≤R_t+i≤R_max, wherein R _min is a lower limit of the resource utilization rate, and R _max is an upper limit of the resource utilization rate;

And (3) optimizing and calculating: adopting a genetic algorithm to perform optimization calculation to generate an optimal resource allocation and routing strategy;

scheme output: and outputting the optimal resource allocation and routing strategy as a resource optimization scheme, and transmitting the optimal resource allocation and routing strategy to a decision support optimization module.

7. The network integrated dynamic resource routing management system of claim 1, wherein the decision support optimization module comprises a network condition evaluation unit, a traffic demand evaluation unit, and a policy adjustment unit; wherein:

Network condition evaluation unit: parameters for real-time assessment of current network conditions, including bandwidth utilization, storage utilization, and computing power utilization, the network condition assessment formula is: Wherein N _t is the network condition evaluation result at time t, B _t is the bandwidth utilization, S _t is the storage utilization, and C _t is the computing power utilization;

business demand assessment unit: the method is used for evaluating the current service demands, including the priority of each service application, the real-time flow demand and the service quality demand; the business requirement evaluation formula is: Wherein, D _t is the service demand evaluation result at time T, P _t is the priority of each service application, T _t is the real-time traffic demand, and Q _t is the quality of service requirement;

Policy adjustment unit: the system is used for combining the network condition evaluation result and the service demand evaluation result, and dynamically adjusting a resource allocation strategy and a routing path based on a resource optimization scheme provided by the intelligent prediction analysis module; the method specifically comprises the following steps:

firstly, integrating a network condition evaluation result and a service demand evaluation result to form a comprehensive evaluation view, which is specifically expressed as: Wherein E _t is a comprehensive evaluation view of the time point t;

Then, based on the comprehensive evaluation view and the resource optimization scheme, calculating an optimal resource allocation strategy and a routing path, wherein a strategy calculation formula is as follows: s _t＝argmax(α·U_R,t+β·U_Q,t-γ·C_t), wherein S _t is a resource allocation policy at a time point t, U _R,t is a resource utilization benefit, U _Q,t is a quality of service benefit, C _t is an adjustment cost, and α, β, and γ are weight coefficients;

and finally, outputting the calculated optimal resource allocation strategy and the route path as an adjustment scheme, and transmitting the adjustment scheme to a dynamic resource scheduling module for executing resource allocation and route adjustment.

8. The network integrated dynamic resource routing management system of claim 1, wherein the dynamic resource scheduling module comprises a resource configuration unit, a route adjustment unit, and an execution monitoring unit; wherein:

Resource allocation unit: the resource allocation scheme is used for receiving the resource allocation scheme generated by the decision support optimization module, and adjusting the network resource allocation in real time according to the scheme, wherein the resource allocation formula is as follows: r _t+1＝R_t+ΔR_t, wherein R _t+1 is the resource configuration of the time point t+1, R _t is the resource configuration of the time point t, and DeltaR _t is the resource adjustment amount calculated according to the resource allocation scheme;

route adjustment unit: the routing adjustment scheme is used for receiving the routing adjustment scheme generated by the decision support optimization module, dynamically adjusting the network routing path according to the scheme, and the routing adjustment formula is as follows:

P _t+1＝argmin(C_P,t+1+β·D_P,t+1), wherein P _t+1 is a routing path at time point t+1, C _P,t+1 is a routing cost at time point t+1, D _P,t+1 is a routing delay at time point t+1, and β is a weight coefficient;

And an execution monitoring unit: the method is used for monitoring the execution effect of resource allocation and route adjustment, ensuring the stability and the service quality of the network in the adjustment process, and executing a monitoring formula as follows: Wherein M _t is an execution monitoring index at a time point t, Q _i,t is an i-th qos parameter at the time point t, and n is the number of qos parameters.

9. The network integrated dynamic resource routing management system of claim 1, wherein the feedback optimization module comprises a data collection unit, a model adjustment unit, and a policy optimization unit; wherein:

a data collection unit: the system comprises a dynamic resource scheduling module, a network performance monitoring module and a network performance monitoring module, wherein the dynamic resource scheduling module is used for continuously collecting adjusted network operation data provided by the dynamic resource scheduling module, including parameters of resource utilization rate, service quality and network performance;

model adjustment unit: the method is used for adjusting and optimizing the hybrid model of the intelligent prediction analysis module based on the operation data provided by the data collection unit, specifically, firstly calculating the error between the prediction result and the actual operation data, wherein the formula is as follows: Wherein E _t is the prediction error at time t, For the predicted value of time point t+i, y _t+i is the actual value of time point t+i, n is the number of predicted data points; then, based on the calculated error, adjusting parameters of the hybrid model through an optimization algorithm, and updating the model to improve the prediction precision;

Policy optimization unit: the method is used for optimizing the resource scheduling and routing strategy by continuously trial and error and learning, specifically, a reinforcement learning algorithm is firstly applied, and the optimal resource scheduling and routing strategy is learned by the trial and error process, and the reinforcement learning reward function is expressed as: r _t＝α·U_R,t+β·U_Q,t-γ·C_t, wherein R _t is a rewarding value of a time point t, U _R,t is a resource utilization rate benefit, U _Q,t is a service quality benefit, C _t is an adjustment cost, and alpha, beta and gamma are weight coefficients; and then adjusting the resource scheduling and routing strategy based on feedback of the reward function so as to improve the overall system performance.

10. A method of network integrated dynamic resource route management according to any of claims 1-9, comprising the steps of:

s1: monitoring the use condition of network resources in real time, including bandwidth, storage and computing capacity, and generating a resource use report;

s2: collecting real-time data from different network nodes, preprocessing the collected data to eliminate noise and redundant information, and simultaneously carrying out multi-source data fusion to improve data quality;

S3: adopting a mixed model combining a convolutional neural network and a long-term and short-term memory network and combining an attention mechanism to carry out multidimensional analysis on the data preprocessed by the S2 so as to predict the resource demand and the flow change trend and generate a resource optimization scheme;

s4: dynamically adjusting a resource allocation strategy and a routing path by utilizing the resource optimization scheme of the S3 and combining the current network condition and the service requirement;

S5: executing the resource allocation and route adjustment scheme generated in the step S4, adjusting the network resource allocation in real time through a dynamic resource scheduling algorithm, and monitoring the execution effect;

S6: continuously collecting the network operation data after the adjustment of S5, analyzing the execution effect, adjusting and optimizing the mixed model in S3 according to the analysis result to form a closed-loop optimization mechanism, and optimizing the resource scheduling and routing strategy through continuous trial and error and learning.