A real-time performance evaluation method for computing power network based on analytic hierarchy process

Abstract

This application relates to the technical field of computer networks and discloses a real-time performance evaluation method for computing power networks based on the Analytic Hierarchy Process (AHP). The method includes: collecting node operating status and task requirement data through sensors, and combining it with an edge quantum algorithm to generate local performance indicators; simulating future network states using digital twins to generate a multidimensional performance dataset through integration; dynamically adjusting AHP weights based on resource deviations and geological classification models: high-frequency updates are enabled during high loads / faults, and the weight range is expanded during low loads; a risk assessment algorithm is introduced to quantify the level of performance-cost-carbon-efficiency conflicts, triggering resource recycling, optimization prompts, or single-metric recommendations; hierarchical scheduling strategies are triggered based on the evaluation results, and proactive intervention is initiated in conjunction with anomaly detection; and AHP weights are dynamically updated through reinforcement learning, with automatic optimization of quantum-classical hybrid algorithm parameters and blockchain verification weights. This application can improve the real-time response efficiency and resource utilization of computing power networks.

Description

Technical Field

[0001] The present application relates to the technical field of computer networks, and in particular to a method for evaluating the real-time performance of a computing network based on the hierarchical analysis method. Background Art

[0002] With the rapid expansion of computing networks and the increasing complexity of business needs, traditional static weighted analytic hierarchy processes (AHPs) face significant limitations in real-time performance evaluation. Existing technologies often rely on expert experience to set fixed weights, making them difficult to adapt to dynamically changing network conditions (such as sudden task loads and node failures). This results in delayed and inflexible evaluation results.

[0003] Furthermore, the lack of a unified data sharing and verification mechanism in cross-domain collaborative scheduling can easily lead to resource allocation conflicts and reduced scheduling efficiency. As can be seen from the above, improving the real-time response efficiency and resource utilization of the computing power network remains to be solved. Summary of the Invention

[0004] In order to improve the real-time response efficiency and resource utilization of the computing power network, the present application provides a computing power network real-time performance evaluation method and system based on hierarchical analysis method.

[0005] In the first aspect, the present application provides a method for evaluating the real-time performance of a computing network based on the hierarchical analysis method, which adopts the following technical solutions:

[0006] A real-time performance evaluation method for computing power network based on hierarchical analysis method, including:

[0007] The operating status data and task requirement data of computing network nodes are obtained in real time through distributed sensors and log systems. The operating status data includes CPU utilization, memory occupancy, network bandwidth, task delay, and energy consumption. The task requirement data includes task type, priority, and resource requirements. The quantum-inspired algorithm deployed on the edge node processes the operating status data and generates local performance evaluation indicators. Based on digital twin simulation, the future network status data includes bandwidth change trends and task load predictions. The local performance evaluation indicators are integrated with the future network status data to generate a multi-dimensional performance evaluation data set containing the current status and future predictions. The sensor includes a lightweight quantum computing module deployed at the edge of the node.

[0008] Based on the operating status data and the preset standard performance reference value, the corresponding resource deviation data is calculated. The current network status is classified in real time using the geological classification model to obtain the network status classification results. If a node failure or high load is detected, the weight adjustment threshold of the AHP judgment matrix is ​​lowered and the high-frequency dynamic weight update mode is enabled. If a low-load or idle node is detected, the weight allocation range is expanded and the global resource scheduling strategy is enabled. If a sudden influx of tasks is detected, the AHP sub-indicator weights are dynamically adjusted based on the task priority data.

[0009] Based on resource deviation data, network status classification results, and task priority data, a risk assessment algorithm is used to calculate the performance-cost-carbon efficiency conflict level during computing power scheduling. If the conflict level is high risk, an emergency resource recovery strategy is enabled and the scheduling of high-energy-consuming nodes is restricted. If the conflict level is medium risk, a multi-objective optimization prompt is triggered and a performance-cost-carbon efficiency trade-off solution is displayed. If the conflict level is low risk, single-indicator optimization suggestions are provided.

[0010] Based on resource deviation data and conflict assessment results, scheduling policy adjustment signals are triggered in a hierarchical manner. These signals include green resource allocation instructions, yellow warnings (dynamic weight corrections), and red alerts (global scheduling resets). If the pressure sensor detects a node anomaly and no conflict assessment is triggered, proactive scheduling intervention is initiated directly to limit resource allocation to high-load nodes.

[0011] Real-time performance evaluation data from a multidimensional performance evaluation dataset is obtained, compared with a historical environmental parameter database, and the AHP weight matrix is ​​dynamically updated through a reinforcement learning model based on the environmental comparison results. If an abnormal operating condition occurs, an attribution analysis is performed on the abnormal condition, and the quantum-classical hybrid algorithm parameters and blockchain verification weights are automatically optimized.

[0012] Optionally, when the multi-dimensional performance evaluation data set is integrated with the local performance evaluation indicator and the future network status data, the method further includes:

[0013] Synchronize local performance indicators with future state data in the time dimension through the timestamp alignment mechanism;

[0014] Obtain time deviation data between the local performance indicator and the future state data in the time dimension. If the time deviation data exceeds a preset time deviation threshold, perform interpolation compensation on the local performance indicator or downsample the future state data.

[0015] Optionally, in the process of triggering the scheduling policy adjustment signal in a hierarchical manner based on the resource deviation data and the conflict assessment result, if the pressure sensor detects a node abnormality and does not trigger the conflict assessment, then adding a constraint condition for active scheduling intervention, the method further includes:

[0016] When the pressure sensor detects that the node has an abnormal operating status, and the abnormal operating status does not reach any risk trigger threshold of the performance-cost-carbon efficiency conflict level, it determines whether the resource deviation data of the abnormal node continues to exceed the preset resource threshold;

[0017] If at least one indicator in the resource deviation data exceeds the set resource threshold for a continuous period of time that reaches the preset length, the active scheduling intervention mechanism is triggered;

[0018] When performing active scheduling intervention, priority is given to retaining the resource allocation rights corresponding to nodes whose energy consumption is lower than the benchmark value, and at the same time, resource freezing operations are implemented on high-load nodes whose resource usage exceeds the load limit.

[0019] Optionally, in the process of dynamically adjusting the weights of the AHP judgment matrix sub-indicators based on the resource deviation data, the network status classification results, and the task priority data, when a sudden influx of tasks is detected, the method further includes:

[0020] Identify the urgency of the current task flow based on the task priority data obtained from the log system;

[0021] If the task priority is judged to be high, the relative importance weight of the task delay indicator in the AHP judgment matrix is ​​increased by 15% to 20%, and the weight of the resource demand indicator is reduced by 5% to 10%;

[0022] After completing the weight adjustment, the consistency ratio test method is used to verify the consistency of the updated AHP judgment matrix. Only when the consistency ratio CR is less than 0.1, it is confirmed that the weight adjustment is logically reasonable and will be applied to subsequent performance evaluation and scheduling decisions.

[0023] Optionally, in the process of simulating future network state data based on the digital twin, the method further includes:

[0024] Build a 3D visual topology model that corresponds one-to-one with the actual computing power network, and synchronously update the connection status, bandwidth change trend and task load distribution between nodes through real-time data streams;

[0025] Deploy a virtual user behavior simulation module in a virtual environment corresponding to the 3D visualization topology model, retrieve the corresponding historical task pattern, obtain the corresponding user behavior characteristics, and generate dynamic task demand data based on the historical task pattern and the user behavior characteristics. The dynamic task demand data serves as an input parameter for future network state prediction.

[0026] The dynamic task demand data generated by simulation is integrated with the operation status data for analysis, and the future network status data including the extension of time and space dimensions is output.

[0027] Optionally, when an abnormal operating condition occurs and attribution analysis is required, the method further includes:

[0028] Based on historical operating status data and task scheduling logs, a causal reasoning graph is constructed between nodes in the computing power network. Nodes in the causal reasoning graph represent computing power resource units, and edges represent the causal strength between task delays, energy consumption changes, and bandwidth fluctuations. Task delays, energy consumption changes, and bandwidth fluctuations are all performance indicators.

[0029] Identify the corresponding anomaly propagation path based on the causal reasoning graph, and quantify the causal correlation between the anomaly source node and the downstream affected nodes based on the anomaly propagation path;

[0030] Based on the abnormal propagation path, a counterfactual virtual scenario is generated to simulate the expected operating state if no abnormality occurs at the key node, and the expected operating state is compared and analyzed with the actual operating state. Based on the comparison results, an attribution report is output to assist in generating root cause-oriented optimization measures, including priority reallocation, node isolation, or parameter adaptive adjustment.

[0031] In the second aspect, the present application provides a real-time performance evaluation system for computing power network based on hierarchical analysis method, which adopts the following technical solutions:

[0032] A real-time performance evaluation system for computing power network based on hierarchical analysis method, including:

[0033] A multi-dimensional performance evaluation data set generation module obtains the operating status data and task requirement data of computing power network nodes in real time through distributed sensors and log systems. The operating status data includes CPU utilization, memory occupancy, network bandwidth, task delay, and energy consumption, and the task requirement data includes task type, priority, and resource requirements. The operating status data is processed by a quantum-inspired algorithm deployed on the edge node to generate local performance evaluation indicators. Based on digital twin simulations, future network status data includes bandwidth change trends and task load predictions. The local performance evaluation indicators are integrated with the future network status data to generate a multi-dimensional performance evaluation data set containing current status and future predictions. The sensor includes a lightweight quantum computing module deployed at the edge of the node.

[0034] The network status classification result acquisition module calculates the corresponding resource deviation data based on the operating status data and the preset standard performance reference value, and classifies the current network status in real time using the geological classification model to obtain the network status classification results. If a node failure or high load is detected, the weight adjustment threshold of the AHP judgment matrix is ​​lowered and the high-frequency dynamic weight update mode is enabled. If a low-load or idle node is detected, the weight allocation range is expanded and the global resource scheduling strategy is enabled. If a sudden influx of tasks is detected, the AHP sub-indicator weights are dynamically adjusted in combination with the task priority data.

[0035] The conflict level calculation module uses a risk assessment algorithm to calculate the performance-cost-carbon efficiency conflict level in the computing power scheduling process based on resource deviation data, network status classification results, and task priority data. If the conflict level is high risk, an emergency resource recovery strategy is enabled and the scheduling of high-energy-consuming nodes is restricted. If the conflict level is medium risk, a multi-objective optimization prompt is triggered and a performance-cost-carbon efficiency trade-off solution is displayed. If the conflict level is low risk, a single indicator optimization suggestion is provided.

[0036] The strategy adjustment signal scheduling module triggers scheduling strategy adjustment signals in a hierarchical manner based on resource deviation data and conflict assessment results. Scheduling strategy adjustment signals include green resource allocation instructions, yellow warnings (dynamic weight correction), and red alerts (global scheduling reset). If the pressure sensor detects a node anomaly and no conflict assessment is triggered, active scheduling intervention is directly initiated to limit resource allocation to high-load nodes.

[0037] The weight matrix update module obtains real-time performance evaluation data from the multidimensional performance evaluation dataset, compares the real-time performance evaluation data with the historical environmental parameter database, and uses the reinforcement learning model to dynamically update the AHP weight matrix based on the environmental comparison results. If an abnormal operating condition occurs, the module performs attribution analysis on the abnormal condition and automatically optimizes the quantum-classical hybrid algorithm parameters and blockchain verification weights.

[0038] In a third aspect, the present application provides a real-time performance evaluation system for computing power networks based on the analytic hierarchy process, which adopts the following technical solutions:

[0039] A system for evaluating the real-time performance of a computing power network based on the hierarchical analysis method comprises a processor in which a program of any one of the above-mentioned methods for evaluating the real-time performance of a computing power network based on the hierarchical analysis method is run.

[0040] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution:

[0041] A storage medium storing a program for the real-time performance evaluation method of a computing power network based on the hierarchical analysis method as described above.

[0042] In summary, this application includes at least one of the following beneficial technical effects:

[0043] The integration of multi-dimensional technologies significantly improves the real-time response efficiency and resource utilization of the computing network. First, based on quantum-inspired algorithms and digital twin technology, the system achieves real-time dynamic modeling of node operating status and future load: lightweight quantum computing modules at edge nodes accelerate the generation of local performance evaluation indicators (such as task latency and energy consumption), while the 3D topology model constructed by digital twins, combined with virtual user behavior simulation, can predict bandwidth fluctuations and task load trends in advance. This dual-track mechanism of "real-time monitoring + future prediction" enables the system to quickly generate scheduling policy adjustment signals (such as green resource allocation and global scheduling reset) based on multi-dimensional performance evaluation datasets when tasks suddenly occur or nodes are abnormal, thereby shortening decision-making delays and reducing the risk of resource conflicts.

[0044] Secondly, through dynamic weight optimization and causal-driven attribution analysis, the system achieves refined resource allocation control. The AHP judgment matrix's weight adjustment mechanism adaptively adjusts based on task priority (e.g., increasing the delay weight of high-priority tasks by 15%-20%), network status (e.g., enabling high-frequency dynamic updates during high load), and abnormal operating conditions (e.g., using causal reasoning graphs to locate root causes), ensuring that critical resources prioritize high-value tasks. Furthermore, a reinforcement learning model dynamically updates the weight matrix by comparing historical environmental parameters. Combined with blockchain-based weight verification, this approach avoids the rigid resource allocation caused by static weights. This closed-loop control system of "prediction-assessment-optimization-verification" significantly improves the computing network's responsiveness and resource utilization in complex scenarios while maintaining a balance between performance, cost, and carbon efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The present invention is a flowchart of a method for evaluating the real-time performance of a computing power network based on the hierarchical analysis method according to an exemplary embodiment.

[0046] Figure 2 It is a structural block diagram of a computing power network real-time performance evaluation system based on hierarchical analysis method according to an exemplary embodiment. DETAILED DESCRIPTION

[0047] Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings.

[0048] Throughout this specification, reference to the terms "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0049] The present application embodiment discloses a method for evaluating the real-time performance of a computing network based on the hierarchical analysis method. Figure 1 ,include:

[0050] S100 obtains the operating status data and task requirement data of computing network nodes in real time through distributed sensors and log systems. The operating status data includes CPU utilization, memory occupancy, network bandwidth, task delay, and energy consumption. The task requirement data includes task type, priority, and resource requirements. The quantum-inspired algorithm deployed on the edge node processes the operating status data and generates local performance evaluation indicators. Based on digital twins, the future network status data is simulated. The future network status data includes bandwidth change trends and task load predictions. The local performance evaluation indicators are integrated with the future network status data to generate a multi-dimensional performance evaluation data set containing the current status and future predictions. Among them, the sensor includes a lightweight quantum computing module deployed at the edge of the node.

[0051] Specifically, the following steps are included:

[0052] Step 1: Collecting status data: Distributed sensors are deployed at the edge of network nodes to monitor hardware resource status in real time, including CPU utilization (reflecting computing power load), memory occupancy (reflecting the degree of memory resource shortage), network bandwidth (measuring data transmission capacity), task latency (characterizing task response efficiency), and energy consumption (related to energy cost and carbon efficiency). After the above data is collected by sensors, it is pre-processed by the lightweight quantum computing module of the edge node to ensure low-latency data transmission and preliminary feature extraction.

[0053] Task requirement data collection: The log system records metadata during the task scheduling process, including task type (such as compute-intensive or communication-intensive), priority (to distinguish urgent tasks from ordinary tasks), and resource requirements (such as the number of CPU cores required and memory size). It should be noted here that task requirement data and operating status data together constitute the multi-dimensional input of performance evaluation.

[0054] Through the collaborative work of distributed sensors and log systems, the system can capture the dynamic behavior of network nodes in real time, providing an accurate, low-latency data basis for subsequent performance evaluation and scheduling decisions, and avoiding decision-making deviations caused by data lags.

[0055] Step 2: Generate local performance evaluation indicators:

[0056] Quantum-inspired algorithm processing: Edge nodes have built-in lightweight quantum computing modules that use quantum annealing or variational quantum algorithms (such as VQE) to extract features and optimize operating status data. For example, by encoding the trade-off between task latency and energy consumption using quantum bits, the optimal local performance indicator (such as the "task latency-energy consumption balance index") can be quickly solved.

[0057] Generating local performance evaluation indicators. The indicators output by the quantum-inspired algorithm include but are not limited to: resource utilization balance (the coordinated load level of CPU and memory); task response efficiency (a comprehensive score of latency and bandwidth); and energy consumption sensitivity (the energy consumption fluctuation rate of a unit computing task).

[0058] It should be pointed out here that quantum-inspired algorithms can significantly shorten the calculation time of local performance evaluation (several times compared to traditional algorithms), while enhancing the robustness of indicators through quantized feature extraction, providing high-quality input for subsequent multidimensional data fusion.

[0059] Step 3: Use digital twin technology to build a virtual network model, predict future states, and integrate it with real-time data to form a multidimensional dataset that includes time and space dimensions. Specifically, it includes:

[0060] Digital twin simulation of future states: Build a 3D digital twin model consistent with the actual computing power network topology, and synchronize the connection status, bandwidth change trends and task load distribution between nodes in real time; through the virtual user behavior simulation module, generate dynamic task demand data based on historical task patterns and user behavior characteristics (such as task request frequency and resource preferences) as input for future state prediction; key parameters for simulating future network states include: bandwidth fluctuation prediction (based on time series analysis), task load peak prediction (based on machine learning model), and node failure probability (based on reliability analysis).

[0061] Multidimensional data fusion: Local performance evaluation indicators (such as resource utilization balance) are temporally and spatially aligned and weightedly fused with future state data generated by digital twins (such as bandwidth prediction and load peak) to generate a multidimensional performance evaluation dataset containing the following dimensions: the current state dimension, which includes real-time operating status and task demand data; the future prediction dimension, which includes predicted values ​​such as bandwidth trends, load peaks, and failure probabilities; and the spatiotemporal correlation dimension, which includes the resource competition relationship between nodes and the time window of the task migration path.

[0062] It should be pointed out here that the multidimensional dataset not only reflects the current network status, but also provides a forward-looking perspective through future predictions, enabling scheduling strategies to predict risks in advance before tasks burst or nodes become abnormal, thereby significantly improving the real-time response efficiency and resource utilization of the computing network.

[0063] A real-time computing network evaluation system is constructed through the collaborative efforts of lightweight quantum computing modules, digital twins, virtual user behavior simulation, and multidimensional data fusion. Lightweight quantum computing enables the generation of low-latency performance indicators at the edge; digital twins combine historical tasks with user behavior characteristics to improve future prediction accuracy; and multidimensional data fusion unifies real-time and predicted data through timestamp alignment and weighting algorithms. These three elements work together to form a closed loop from real-time data collection to forward-looking modeling, providing high-precision data support for AHP weight adjustment, conflict assessment, and scheduling optimization, significantly improving the response efficiency and resource utilization of computing networks in complex environments.

[0064] S200: Calculate corresponding resource deviation data based on the operating status data and preset standard performance reference values, classify the current network status in real time using a geological classification model, and obtain a network status classification result. If a node failure or high load is detected, lower the weight adjustment threshold of the AHP judgment matrix and enable a high-frequency dynamic weight update mode. If a low-load or idle node is detected, expand the weight allocation range and enable a global resource scheduling strategy. If a sudden influx of tasks is detected, dynamically adjust the AHP sub-indicator weights based on the task priority data.

[0065] Specifically, the following steps are included:

[0066] Step 1: Calculate resource deviation data based on operating status data and standard performance reference values:

[0067] Data input: The operating status data obtained from S100 (such as CPU utilization, memory usage, network bandwidth, task latency, and energy consumption) is used as input, combined with preset standard performance reference values ​​(such as a CPU utilization threshold of 85% and a memory usage threshold of 90%).

[0068] Deviation quantification: Generate a multi-dimensional resource deviation dataset by calculating the difference between the actual value and the standard value or the relative deviation rate (for example, CPU utilization deviation rate = (actual value - standard value) / standard value).

[0069] Abnormal determination: If the deviation rate of a certain indicator exceeds the preset threshold (such as the CPU utilization deviation rate > 20%), it is marked as an abnormal state, which serves as an important basis for subsequent network status classification.

[0070] It should be pointed out here that resource deviation data directly reflects the health status of network nodes and the efficiency of resource allocation. By quantifying the deviation, the system can quickly identify high-load, low-load or resource waste scenarios, provide an objective basis for dynamically adjusting the AHP weights, and avoid decision-making bias caused by subjective judgment.

[0071] Step 2: Classify the current network status in real time using the geological classification model:

[0072] Model selection: The LVQ neural network (Learning Vector Quantization) is used as the geological classification model because of its simple structure, fast training speed and good classification effect. It is suitable for processing high-dimensional resource deviation data.

[0073] Classification logic: Input resource deviation data sets (such as CPU utilization deviation rate, memory usage deviation rate, etc.) and output network status classification labels (such as "high load", "low load", "idle", and "abnormal"). The model uses a competitive learning mechanism to map the input data to preset category centers to achieve real-time classification.

[0074] Dynamic update: The model weight vector is regularly iteratively optimized using historical resource deviation data to ensure the accuracy and adaptability of the classification results.

[0075] By classifying the network status in real time through the geological classification model, the system can quickly identify key scenarios (such as sudden task influx and node failure), thereby triggering the corresponding AHP weight adjustment strategy and improving the pertinence and real-time nature of scheduling decisions.

[0076] Step 3: Dynamically adjust the AHP judgment matrix weights based on the network status classification results. The weight adjustment strategy specifically includes:

[0077] High-load / node failure scenarios: Lower the weight adjustment threshold of the AHP judgment matrix (for example, from the default 10% to 5%) and enable high-frequency dynamic weight update mode; prioritize increasing the weights of indicators such as "task delay" and "resource demand" to quickly respond to resource competition issues under high load.

[0078] Low-load / idle node scenarios: Expand the weight distribution range (e.g., introduce more sub-indicators, such as "energy sensitivity" and "task priority"), and enable a global resource scheduling strategy; reduce the weight of "task delay" and increase the weight of "carbon efficiency" to optimize energy utilization efficiency.

[0079] In the scenario of a sudden influx of tasks: Combined with task priority data (such as the task type and priority obtained in S100), the weights of the "task delay" and "resource demand" indicators are dynamically adjusted. If the task priority is high, the "task delay" weight is increased by 15% to 20%, and the "resource demand" weight is reduced by 5% to 10%.

[0080] By dynamically adjusting AHP weights, the system can flexibly balance performance, cost, and carbon efficiency priorities based on different network states. For example, it prioritizes task response speed in high-load scenarios and optimizes energy efficiency in low-load scenarios, thereby achieving dynamic resource allocation and conflict avoidance in the computing network.

[0081] The S200 achieves adaptive optimization of computing network resource allocation strategies by quantifying resource deviations and classifying network status in real time, combined with a dynamic AHP weight adjustment mechanism. Specifically, the system first calculates resource deviations based on operational status data and standard reference values ​​to identify network load anomalies. It then uses geological classification models (such as the LVQ neural network) to classify the current status in real time (e.g., high load, idle, or bursty tasks). Finally, based on the classification results, it dynamically adjusts the sub-indicator weights of the AHP judgment matrix (e.g., prioritizing task delays and reducing resource demand weights), thereby balancing performance-cost-efficiency conflicts in different network scenarios. This step ensures the flexibility and targeted nature of resource scheduling strategies, significantly improving the computing network's responsiveness and resource utilization in complex and dynamic environments.

[0082] S300, based on resource deviation data, network status classification results and task priority data, calculates the performance-cost-carbon efficiency conflict level in the computing power scheduling process through a risk assessment algorithm; if the conflict level is high risk, the emergency resource recovery strategy is enabled and the scheduling of high-energy consumption nodes is restricted; if the conflict level is medium risk, a multi-objective optimization prompt is triggered and a performance-cost-carbon efficiency trade-off solution is displayed; if the conflict level is low risk, a single indicator optimization suggestion is provided.

[0083] Specifically, the following steps are included:

[0084] Step 1: Integrate resource deviation data, network status classification results, and task priority data:

[0085] Resource deviation data: The resource deviation data (such as CPU utilization deviation rate and memory usage deviation rate) obtained from S200 reflects the health status of the current network nodes and the resource allocation efficiency.

[0086] Network status classification results: Based on the classification labels (such as "high load", "low load", and "burst task") output by the geological classification model in S200, it provides macro characteristics of the network operation scenario.

[0087] Task priority data: Task requirement data (such as task type, priority, and resource requirements) obtained from S100 to clarify the task's sensitivity to performance, cost, and carbon efficiency.

[0088] By integrating resource status, network scenarios, and mission characteristics, the system can build a multi-dimensional conflict assessment input framework, providing a basis for subsequent risk quantification and avoiding the one-sidedness of single indicator evaluation.

[0089] Step 2: Calculate the performance-cost-carbon efficiency conflict level using a risk assessment algorithm:

[0090] Risk assessment algorithm logic:

[0091] Weighted indicator calculation: The improved analytic hierarchy process (AHP) is used to perform weighted scoring on performance (such as task delay), cost (such as energy consumption), and carbon efficiency (such as carbon emissions per unit task). The weights are provided by the AHP judgment matrix dynamically adjusted by S200.

[0092] Conflict quantification formula: Conflict level = (performance requirement - actual performance) + (cost budget - actual cost) + (carbon efficiency target - actual carbon efficiency). If the deviation in any dimension exceeds the threshold, the conflict escalation is triggered.

[0093] Risk grading rules: High risk: The sum of the three conflicts ≥ the upper threshold (e.g., the sum > 15%), indicating that the system is on the verge of performance collapse or resource waste; Medium risk: The lower threshold < the sum < the upper threshold (e.g., 5% < the sum < 15%), requiring trade-off optimization; Low risk: The sum ≤ the lower threshold (e.g., ≤ 5%), allowing for local optimization.

[0094] By quantifying the conflict level, the system can accurately identify high-risk scenarios (such as resource contention under high load) and provide a decision-making basis for subsequent hierarchical scheduling strategies to avoid rigid resource allocation.

[0095] Step 3: Trigger the corresponding scheduling strategy based on the conflict level:

[0096] High-risk level processing: Enable emergency resource recovery strategies to forcibly reclaim resources from high-energy-consuming nodes (such as shutting down non-critical tasks), giving priority to ensuring the performance requirements of high-priority tasks; restrict the scheduling of high-energy-consuming nodes, and prohibit high-energy-consuming nodes from participating in new task allocation through blockchain verification weight mechanism to reduce carbon efficiency conflicts.

[0097] Medium-risk level processing: Trigger multi-objective optimization prompts and display the performance-cost-carbon efficiency trade-off solution to the scheduling system (such as "increasing task delay tolerance to reduce energy consumption"); dynamically adjust weight distribution, combine task priority data, temporarily increase carbon efficiency weight (such as increasing it by 5%), and guide resource allocation towards green.

[0098] Low-risk level processing: Provide single indicator optimization suggestions and propose local optimization measures (such as adjusting the task scheduling order) for the weakest dimension (such as insufficient performance); maintain global stable scheduling, keep the current AHP weight unchanged, and only implement minor adjustments to abnormal nodes (such as resource freezing).

[0099] The tiered strategy ensures that the system takes differentiated responses in different risk scenarios: rapid intervention in high-risk scenarios to avoid collapse, balancing the needs of multiple parties in medium-risk scenarios, and maintaining efficient operation in low-risk scenarios, thereby achieving dynamic adaptive scheduling of the computing power network.

[0100] S300 accurately evaluates the performance-cost-carbon efficiency conflict level by combining conflict quantification with AHP dynamic weights, uses blockchain verification linkage to ensure the trusted execution of resource recovery strategies in high-risk scenarios, and achieves real-time multi-objective optimization prompts based on task priorities, effectively solving the resource allocation rigidity problem caused by static weight deviation, decision lag and lack of trust in traditional scheduling systems.

[0101] S400, based on resource deviation data and conflict assessment results, triggers scheduling strategy adjustment signals in a hierarchical manner, where scheduling strategy adjustment signals include green resource allocation instructions, yellow warning - dynamic weight correction, and red alert - global scheduling reset; if the pressure sensor detects a node abnormality and does not trigger a conflict assessment, it directly initiates active scheduling intervention and limits resource allocation to high-load nodes.

[0102] Specifically, the following steps are included:

[0103] Step 1: trigger scheduling strategy adjustment signals based on resource deviation and conflict assessment results:

[0104] Scheduling policy hierarchical triggering logic:

[0105] High-risk level processing: Enable emergency resource recovery strategies to forcibly reclaim resources from high-energy-consuming nodes (such as shutting down non-critical tasks), giving priority to ensuring the performance requirements of high-priority tasks; restrict the scheduling of high-energy-consuming nodes, and prohibit high-energy-consuming nodes from participating in new task allocation through blockchain verification weight mechanism to reduce carbon efficiency conflicts.

[0106] Medium-risk level processing: Trigger multi-objective optimization prompts and display the performance-cost-carbon efficiency trade-off solution to the scheduling system (such as "increasing task delay tolerance to reduce energy consumption"); dynamically adjust weight distribution, combine task priority data, temporarily increase carbon efficiency weight (such as increasing it by 5%), and guide resource allocation towards green.

[0107] Low-risk level processing: Provide single indicator optimization suggestions and propose local optimization measures (such as adjusting the task scheduling order) for the weakest dimension (such as insufficient performance); maintain global stable scheduling, keep the current AHP weight unchanged, and only implement minor adjustments to abnormal nodes (such as resource freezing).

[0108] A hierarchical strategy ensures that the system takes differentiated responses in different risk scenarios: rapid intervention in high-risk scenarios to avoid collapse, balancing the needs of multiple parties in medium-risk scenarios, and maintaining efficient operation in low-risk scenarios, thereby achieving dynamic adaptive scheduling of the computing power network.

[0109] Step 2: The pressure sensor detects node anomalies and triggers active scheduling intervention:

[0110] Abnormal operation status determination: Pressure sensors continuously monitor the node's CPU temperature, power consumption, load fluctuations, and other indicators. If a certain indicator (such as CPU temperature) continuously exceeds the set threshold (such as 90°C) and lasts for more than a preset period of time (such as 5 minutes), it is determined to be abnormal.

[0111] Active scheduling intervention mechanism: reserves resource allocation rights for nodes whose energy consumption is lower than the baseline value (such as nodes in energy-saving mode) to ensure that critical tasks are not affected; implements resource freezing operations (such as pausing task scheduling) for nodes whose resource usage exceeds the load limit (such as CPU utilization >95%) to prevent overload crashes.

[0112] Through real-time monitoring and active intervention of pressure sensors, the system can avoid potential risks in advance before the conflict level trigger conditions are reached, and prevent global resource allocation imbalances caused by local node anomalies.

[0113] Step 3: Dynamically update the AHP weight matrix based on historical environmental parameter comparison:

[0114] Comparison of real-time performance evaluation data with historical environment parameters: Compare and analyze the multi-dimensional performance evaluation data set generated by the S100 (including current status and future forecast data) with the historical environment parameter database (such as load peaks and task type distribution over the past 24 hours); identify differences between the current and historical environments (such as sudden task influxes and bandwidth drops).

[0115] Weight updates driven by reinforcement learning models: Reinforcement learning models (such as deep Q networks) dynamically adjust the weights of each sub-indicator in the AHP judgment matrix based on comparison results (such as increasing the weight of "task delay" to cope with sudden tasks); weight updates must meet consistency checks (such as CR < 0.1) to ensure that the adjusted weights are logically reasonable.

[0116] Quantum-classical hybrid algorithm and blockchain verification weight optimization: If an abnormal operating condition occurs (such as a node failure), the quantum-classical hybrid algorithm parameters are automatically optimized (such as adjusting the number of quantum annealing iterations) to improve the accuracy of local performance evaluation; the credibility and immutability of weight adjustments are ensured through blockchain verification weight mechanisms (such as smart contracts).

[0117] By comparing historical data and dynamically updating the weight matrix through reinforcement learning, the system can adapt to complex and changing operating environments and avoid the rigid resource allocation caused by static weights. At the same time, the blockchain verification mechanism ensures the credibility of strategy adjustments.

[0118] Step 4: Abnormal operating condition attribution analysis and parameter optimization:

[0119] Causal reasoning graph construction: Based on historical operating status data and task scheduling logs, a causal graph is constructed between computing network nodes (nodes represent resource units, and edges represent the causal strength of performance indicators). This identifies abnormal propagation paths (e.g., "high latency at node A → bandwidth congestion at node B") and quantifies the causal relationship between the abnormal source node and the affected nodes.

[0120] Counterfactual scenario generation: This simulates the expected operating state of "what if no abnormality occurred at the critical node" (e.g., bandwidth utilization when node A is operating normally) and compares and analyzes it with the actual operating state. Based on the comparison results, an attribution report is generated (e.g., "The failure of node A caused a 30% increase in task delays").

[0121] Parameter and strategy optimization: Adjust system parameters based on the attribution report (such as increasing redundant resource allocation for node A); generate optimization measures (such as "priority reallocation," "node isolation," and "parameter adaptive adjustment") and automatically deploy them to the scheduling strategy.

[0122] Through causal reasoning and counterfactual analysis, the system can accurately locate the root cause of anomalies and generate targeted optimization solutions, significantly improving the fault recovery capability and long-term stability of the computing network.

[0123] S500 obtains real-time performance evaluation data from a multi-dimensional performance evaluation dataset, compares the real-time performance evaluation data with the historical environmental parameter database, and dynamically updates the AHP weight matrix through a reinforcement learning model based on the environmental comparison results. If an abnormal operating condition occurs, it performs attribution analysis on the abnormal condition and automatically optimizes the quantum-classical hybrid algorithm parameters and blockchain verification weights.

[0124] Specifically, the following steps are included:

[0125] Step 1: Monitor the execution effect of the scheduling strategy in real time and generate feedback data:

[0126] Performance indicator monitoring: Real-time collection of key performance indicators (KPIs) such as task completion time, resource utilization, and task delay, and comparison with the target values ​​set by the scheduling strategy.

[0127] Cost and carbon efficiency tracking: Record energy consumption and carbon emissions per unit task after the scheduling strategy is executed, and verify whether the cost and carbon efficiency constraints are met.

[0128] Abnormal behavior detection: Identify abnormal behaviors in scheduling policy execution (such as frequent task migration and node overload) through log analysis and traffic monitoring.

[0129] Through comprehensive data feedback, the system can quantify the actual effect of the scheduling strategy, locate execution deviations, and avoid resource waste or performance degradation caused by strategy design flaws or environmental changes.

[0130] Step 2: Dynamically adjust scheduling policy parameters based on feedback data:

[0131] Weight matrix fine-tuning: If feedback shows that performance indicators (such as task delay) are not up to standard, dynamically increase the weight of "task priority" in the AHP weight matrix to prioritize resource allocation for high-priority tasks.

[0132] Resource allocation threshold correction: If feedback shows that node loads frequently exceed limits, adjust the resource allocation threshold (for example, lower the maximum number of tasks per node) to avoid local overloads.

[0133] Conflict handling rule updates: Optimize conflict handling rules (such as introducing a migration cost penalty factor) to address abnormal behaviors (such as frequent task migration) and reduce invalid scheduling operations.

[0134] Through dynamic parameter adjustment, the system can quickly respond to environmental changes (such as sudden task surges or node failures), ensuring the robustness and efficiency of the scheduling strategy in complex scenarios.

[0135] Step 3: Introduce reinforcement learning model to optimize the global scheduling strategy:

[0136] State space definition: Encode the current network state (such as node load, task queue length, resource deviation rate) into a state vector for reinforcement learning.

[0137] Reward function design: Define a multi-objective reward function that comprehensively considers performance (+10 / task completion), cost (-5 / unit energy consumption), and carbon efficiency (-8 / unit carbon emissions) to guide the model to learn the optimal strategy.

[0138] Strategy training and deployment: Through the Deep Q Network (DQN) training model, the optimal scheduling action (such as "migrate tasks to low-energy nodes") is output and deployed to the actual scheduling system.

[0139] Based on the above steps, it is possible to autonomously learn the optimal scheduling rules in complex scenarios from historical feedback, solve the problem that traditional static strategies are difficult to adapt to dynamic environments, and significantly improve resource utilization and scheduling efficiency.

[0140] Step 4: Blockchain verification and policy credibility assurance:

[0141] On-chain policy changes: Each adjustment record of the scheduling policy (such as weight matrix update, resource allocation threshold modification) is written to the blockchain as a transaction, generating an unalterable audit log.

[0142] Multi-node consensus verification: Consensus verification of policy changes is performed through multiple nodes in the alliance chain (such as network management nodes and security audit nodes) to prevent malicious tampering.

[0143] Smart contract automatic execution: Deploy smart contract rules (such as "If performance indicators do not improve after policy adjustment, roll back to the previous version") to ensure that policy adjustments comply with preset security boundaries.

[0144] Blockchain technology provides a transparent and reliable audit mechanism for the adjustment of scheduling strategies, avoiding strategy failure caused by human intervention or malicious attacks, and ensuring the safe operation of the computing power network.

[0145] Through multi-dimensional data fusion and dynamic decision-making mechanisms, the resource scheduling efficiency and stability of the computing power network have been significantly improved. S100 combines distributed sensors with quantum-inspired algorithms to achieve low-latency real-time performance evaluation; S200 dynamically adjusts AHP weights based on resource deviation and geological classification models to accurately balance the priorities of performance, cost, and carbon efficiency; S300 ensures the timeliness and credibility of resource recovery in high-risk scenarios through conflict quantification and hierarchical strategy triggering; S400 and S500 introduce reinforcement learning and blockchain verification to dynamically optimize global scheduling strategies and ensure the transparency of strategy adjustments. The overall process achieves closed-loop control from real-time monitoring to forward-looking prediction, solving the problem of resource allocation rigidity caused by static weight deviation, decision lag, and lack of trust in traditional scheduling systems, making the computing power network more adaptable and stable in complex dynamic environments.

[0146] Through the synergy of quantum computing acceleration, digital twin prediction, and blockchain trusted verification, an intelligent scheduling system for computing power networks has been constructed. Lightweight quantum computing modules rapidly generate robustness performance indicators at the edge, reducing data processing latency. Digital twins combine historical tasks with user behavior simulations to improve the accuracy of future state predictions, providing a forward-looking basis for scheduling strategies. Reinforcement learning models dynamically optimize global strategies to adapt to sudden tasks and environmental changes. Blockchain verification mechanisms ensure the immutability of weight adjustments and policy changes, enhancing system security. This technology combination not only breaks through the static and localized limitations of traditional scheduling systems, but also significantly improves resource utilization, fault recovery capabilities, and green levels through multi-objective trade-offs and real-time feedback mechanisms, providing core support for the efficient operation of complex computing power networks.

[0147] In an embodiment of the present application, when the multi-dimensional performance evaluation data set is integrated with the local performance evaluation indicators and the future network status data, the method further includes:

[0148] In step 1, the system selects a high-precision time source from among all sensors and digital twin modules as the master time base, such as sensors deployed at core nodes. This master time base must have a high sampling frequency (e.g., 100 times per second) and low clock deviation (less than 1 millisecond). It is calibrated using the Network Time Protocol or Precision Time Protocol to ensure time synchronization with other devices. The system then uses linear or Lagrangian interpolation algorithms to map local performance metrics collected by other sensors (e.g., resource utilization balance) or future state data predicted by the digital twin (e.g., bandwidth trends) onto the master time base. For example, if a sensor at an edge node has a low sampling frequency (50 times per second), its data is interpolated to high-frequency time points on the master time base, ensuring temporal alignment of all data. This process is similar to adjusting videos with different frame rates to the same playback rate to ensure synchronization of video and audio.

[0149] Step 2: After timestamp alignment, the system detects the time offset between the local performance metrics and future state data. Time offset refers to the time difference between the aligned data and the original data. For example, if a sensor's data collection is delayed due to network latency, the system initiates a compensation mechanism. For local performance metrics, the system uses interpolation to infer the performance value at the missing time point using data from known time points. For example, if a sensor on a node fails to return data at a certain time point, the system calculates an intermediate value based on performance metrics (such as CPU utilization) from two preceding and subsequent time points to fill in the missing data. For future state data (such as bandwidth prediction), the system downsamples the data density to match the temporal resolution of the local performance metric. For example, if the digital twin model predicts a high-frequency bandwidth trend (100 times per second), but the local performance metric only supports an update frequency of 50 times per second, the system retains a predicted value every two time points to ensure temporal consistency. This process is similar to compressing high-definition video into a low-resolution version to accommodate the performance limitations of the playback device while preserving key information.

[0150] In an embodiment of the present application, in the process of hierarchically triggering a scheduling policy adjustment signal based on resource deviation data and conflict assessment results, if the pressure sensor detects a node abnormality and does not trigger a conflict assessment, then a constraint condition for active scheduling intervention is added, and the method further includes:

[0151] In step 1, if the pressure sensor detects an abnormal node operating status but does not trigger the performance-cost-carbon efficiency conflict level, the system will further analyze the node's resource deviation data. Resource deviation data includes the difference between the actual value and the standard value of indicators such as CPU utilization, memory usage, network bandwidth, and task latency.

[0152] The system checks whether these metrics consistently exceed preset resource thresholds (e.g., CPU utilization > 95%, memory usage > 90%) within a preset time window (e.g., 5 minutes). If at least one metric consistently exceeds the threshold for the preset duration, the node is deemed to be at risk and requires proactive scheduling intervention. For example, if a node's CPU utilization exceeds 95% for 5 consecutive minutes, the system triggers intervention to prevent local overload from leading to imbalanced global resource allocation. This process is similar to the "threshold warning" mechanism used in industrial equipment, which continuously monitors the changing trends of key metrics to identify anomalies and take action in advance.

[0153] Step 2: When the resource deviation data meets the continuous exceeding condition, the system will start the active scheduling intervention mechanism. At this time, the scheduling strategy will prioritize the allocation of resources for critical tasks while limiting the resource usage of high-load nodes. Specifically:

[0154] Prioritizing resource allocation for low-energy nodes: The system selects nodes with energy consumption below a baseline (e.g., nodes with per-task energy consumption below the industry average) to ensure their resource allocation privileges remain intact. This measure aims to maintain the stable operation of critical tasks while optimizing energy efficiency. For example, during periods of power shortage, the system prioritizes resource allocation for low-energy nodes, avoiding energy waste and carbon-efficiency conflicts.

[0155] Freezing resource allocation for heavily loaded nodes: For nodes whose resource usage exceeds the upper limit (e.g., CPU utilization > 95%), the system freezes resources. This means suspending new task scheduling for the node and gradually releasing its allocated resources. For example, if a node's CPU is overloaded due to a sudden influx of tasks, the system freezes its resource allocation to prevent task accumulation and node crashes. This strategy is similar to the node resource isolation mechanism in Kubernetes, which dynamically adjusts resource allocation to prevent local failures from spreading across the entire network.

[0156] Step 3: The execution of active scheduling intervention must comply with the following constraints:

[0157] Gradual resource freezing: When freezing resources on heavily loaded nodes, the system prioritizes reclaiming resources for non-critical tasks (such as low-priority tasks) rather than immediately terminating all tasks. For example, if a node is running both high- and low-priority tasks, the system will retain resources for the high-priority task and only freeze the allocation for the low-priority task.

[0158] Dynamic adjustment of low-energy nodes: The system regularly updates the baseline values ​​for low-energy nodes (e.g., based on real-time electricity prices or energy availability) to ensure resource allocation strategies align with current environmental conditions. For example, during nighttime periods of low electricity consumption, the system might relax the baseline values ​​for low-energy nodes to take advantage of cheaper electricity.

[0159] Recovery mechanism after intervention: When the resource deviation data of an abnormal node returns to normal range (for example, CPU utilization drops below 85%), the system will gradually unfreeze its resource allocation rights and reintegrate it into the scheduling pool. For example, if a node's resources are frozen due to a short-term overload, once its load returns to normal, the system will re-evaluate its performance indicators and restore scheduling rights based on the AHP weight adjustment strategy.

[0160] This mechanism uses dynamic threshold judgment and resource freezing / recovery strategies to achieve active intervention on abnormal nodes without triggering conflict assessment, effectively balancing system stability and resource utilization efficiency.

[0161] In an embodiment of the present application, in the process of dynamically adjusting the weights of the AHP judgment matrix sub-indicators based on the resource deviation data, the network status classification results, and the task priority data, when a sudden influx of tasks is detected, the method further includes:

[0162] Step 1: When a sudden influx of tasks is detected, the system will obtain the priority data of the current task flow through the log system to determine the urgency of the task. The log system will record the label of each task (such as "high priority", "medium priority", "low priority") or dynamic score (such as a value generated based on the task deadline, resource demand intensity, etc.). If the task priority is determined to be high (for example, exceeding the preset threshold, such as a priority score > 8), the system will trigger the weight adjustment process. This process is similar to the factory production line where sensors are used to detect product types (such as standard parts or customized parts) and adjust the processing priority based on the type to ensure that high-value products are processed first.

[0163] Step 2: After confirming that the task priority is high, the system will dynamically adjust the weights of the sub-indicators in the AHP judgment matrix:

[0164] Increase the weight of task delay indicators: Increase the relative importance of task delay indicators (such as task response time and waiting time) by 15% to 20%. For example, if the original weight of the task delay indicator in the judgment matrix was 0.3, the adjusted weight will be 0.345-0.36 (the specific value depends on the adjustment). This adjustment reflects the strong need for rapid response for high-priority tasks, similar to creating a green lane for emergency vehicles in traffic scheduling to prioritize their passage efficiency.

[0165] Reduce the weight of resource demand indicators: Reduce the weight of resource demand indicators (such as CPU utilization and memory consumption) by 5% to 10%. For example, if the original weight was 0.25, it will be adjusted to 0.225-0.2375. This adjustment reflects that in high-priority task scenarios, the system is more likely to sacrifice some resource efficiency to meet task timeliness requirements, similar to reserving capacity for critical equipment in power dispatch, even if overall resource utilization is slightly reduced.

[0166] Step 3: After completing the weight adjustment, the system will use the consistency ratio test method to verify whether the updated AHP judgment matrix is ​​logically reasonable. The specific process is as follows:

[0167] Calculating the consistency ratio: The system first determines the maximum eigenvalue of the adjusted judgment matrix and then calculates the consistency index based on the matrix order (i.e., matrix dimension). For example, if the matrix order is 4, the system derives the specific value of the consistency index based on the relationship between the maximum eigenvalue and the order. The system then uses a table lookup to obtain the random consistency index corresponding to the matrix order (this index is a precalculated and stored benchmark value) and further calculates the consistency ratio. If the consistency ratio is less than 0.1, the logical consistency of the adjusted judgment matrix is ​​acceptable; if it is greater than or equal to 0.1, the weight assignment needs to be revised.

[0168] Confirmation and application of logical rationality: Only when the consistency ratio is less than 0.1 will the system confirm that the weight adjustment is logically rational and apply the adjusted weights to subsequent performance evaluations and scheduling decisions. For example, if the adjusted consistency ratio is 0.08, it indicates that the weight distribution conforms to expert judgment logic and can be used to guide resource scheduling. If the consistency ratio is 0.12, the original weights must be rolled back or readjusted to avoid decision-making deviations due to logical contradictions. This process is similar to when engineers design a bridge, they verify the rationality of the structural stress distribution through finite element analysis. The construction plan will only be approved if the simulation results meet the safety factor.

[0169] In an embodiment of the present application, in the process of simulating future network state data based on digital twins, the method further includes:

[0170] In step 1, when simulating future network state data based on digital twins, the system first constructs a 3D visual topology model that corresponds exactly to the actual computing network. This model dynamically captures the physical network's topology (e.g., node locations and connectivity), device configuration information (e.g., bandwidth limits and computing power), and real-time operational status (e.g., current workload and link utilization). For example, in a campus network or enterprise data center scenario, the system maps physical devices such as servers, switches, and routers as nodes in the virtual model, visually displaying their geographic locations and connectivity through a three-dimensional spatial layout.

[0171] To synchronize the model with the real network, the system continuously receives real-time data streams from sensors, network management platforms, or telemetry protocols. Based on this data, the system updates node connectivity status (e.g., link failure or recovery), bandwidth trends (e.g., congestion caused by bursts), and task load distribution (e.g., a node's CPU utilization spikes due to high concurrent requests). This process is similar to synchronizing real-time vibration data in bridge health monitoring systems, where model parameters are continuously updated to ensure dynamic consistency with the physical entity.

[0172] Step 2: Based on the 3D visual topology model, the system deploys a virtual user behavior simulation module to generate dynamic task demand data required for future network status prediction. The core logic of this module includes:

[0173] Retrieving historical task patterns: The system extracts historical task data (such as user access frequency, task type distribution, and peak time periods) from the log system or database and identifies typical task patterns based on statistical analysis. For example, in an e-commerce scenario, the system may discover periodic peaks in shopping cart checkout requests during the "Double 11" shopping festival.

[0174] Generating User Behavior Features: Combining collaborative filtering algorithms or machine learning models (such as LSTM), the system extracts user behavior features (such as preferred product categories, operation duration, and payment method choices) from historical task data. For example, for a particular user group, the system may discover that they tend to complete a purchase after viewing three product pages.

[0175] Simulating dynamic task demands: Based on historical task patterns and user behavior characteristics, the virtual user behavior simulation module generates dynamic task demand data (such as simulated visit volume, task type ratio, and resource request intensity within the next hour). For example, when predicting future network conditions, the system might simulate a scenario where 500 users simultaneously initiate video conference requests at 10:00 AM.

[0176] This process is similar to the simulation test of a factory production line. By presetting historical production data and equipment behavior patterns, production capacity bottlenecks or failure risks can be predicted in advance.

[0177] Step 3: The system integrates and analyzes the dynamic task demand data generated by simulation with the real-time operation status data to output future network status data with extended time and space dimensions. Specifically:

[0178] Spatiotemporal modeling: The system combines the temporal characteristics of dynamic task demand data (such as the periodicity of task initiation) with spatial characteristics (such as the geographic distribution of task requests) to construct a multidimensional spatiotemporal model. For example, if a sudden influx of tasks occurs in a specific area, the system can predict the load diffusion effect on surrounding nodes.

[0179] Data Fusion and Prediction: Through the digital twin platform's simulation engine, the system inputs dynamic task demand data (such as simulated user behavior) and real-time operational status data (such as current node load) into the prediction model to generate future network status data (such as bandwidth utilization, task latency, and node overload risk). For example, when predicting the network status for the next five minutes, the system might output the conclusion that "Node A's CPU utilization will increase from 70% to 95%, triggering resource scheduling intervention."

[0180] Ultimately, the system presents the prediction results in a spatiotemporal format. For example, it can visualize network bottlenecks at a specific point in the future in a 3D topology visualization (e.g., highlighting links with insufficient bandwidth in red) or generate a timeline view showing trends in key performance indicators (e.g., task latency). This process is similar to typhoon path prediction in weather forecasting, using spatiotemporal data modeling to provide early warning of potential risks.

[0181] In an embodiment of the present application, when an abnormal operating condition occurs and attribution analysis is required, the method further includes:

[0182] In step 1, the system constructs a causal inference graph between nodes in the computing network based on historical operating status data and task scheduling logs. Nodes in the causal inference graph represent computing resource units (e.g., servers, storage devices), and edges represent causal relationships between performance indicators (e.g., task latency, bandwidth fluctuations). For example, if a CPU overload on node A causes increased latency on node B, a causal edge is formed between the two, with the weight reflecting the strength of the causal relationship (e.g., quantified using a statistical model). The causal inference graph is automatically generated through time series analysis and task dependencies to describe the dynamic relationships between nodes.

[0183] In step 2, the system locates the anomaly propagation path based on the causal inference graph and quantifies the causal correlation between the anomaly source node and downstream nodes. For example, if an abnormal bandwidth fluctuation is detected at node C, the system traces back to its upstream nodes (such as resource contention at node B) and downstream nodes (such as latency at node D), forming a complete propagation chain. Integrating statistical models (such as linear regression) or physical simulation, the system quantifies the impact of the anomaly source on downstream nodes along each path (for example, if a CPU overload at node A causes a 20% increase in latency at node B, the causal correlation is 0.8).

[0184] In step 3, the system generates a counterfactual scenario based on the anomaly propagation path, simulating the expected operating state of the critical node if the anomaly had not occurred. For example, assuming that Node A's CPU overload had not occurred, the system adjusts its resource allocation strategy and predicts task latency and energy consumption distribution. The system then compares the simulation results with the actual operating state across multiple dimensions (such as latency differences and resource utilization changes), identifies the root cause of the anomaly (e.g., resource contention at Node A causing a 30% increase in latency), and generates an attribution report.

[0185] Based on the attribution report, the system proposes optimization measures for the root cause:

[0186] Priority reallocation: adjust task scheduling priorities to prioritize the execution of critical tasks (such as reducing the resource allocation ratio of low-priority tasks on node A); node isolation: isolate abnormal source nodes with high causal correlation (such as migrating high-volatility tasks on node B to redundant nodes); parameter adaptive adjustment: dynamically adjust network parameters (such as optimizing bandwidth allocation strategies or increasing cache capacity) to mitigate the impact of abnormalities.

[0187] By driving dynamic analysis through causal reasoning graphs and combining them with counterfactual simulations to accurately locate root causes, and linking multi-dimensional optimization measures (task scheduling, resource management, and network layers), the fault response efficiency and stability of the computing network can be significantly improved.

[0188] The present application embodiment discloses a real-time performance evaluation system for computing network based on hierarchical analysis method. Figure 2 ,include:

[0189] Multi-dimensional performance evaluation data set generation module 001, which obtains the operating status data and task demand data of computing network nodes in real time through distributed sensors and log systems. The operating status data includes CPU utilization, memory occupancy, network bandwidth, task delay, and energy consumption, and the task demand data includes task type, priority, and resource requirements; the operating status data is processed and local performance evaluation indicators are generated through quantum-inspired algorithms deployed on edge nodes; based on digital twins, future network status data is simulated. The future network status data includes bandwidth change trends and task load predictions. The local performance evaluation indicators are integrated with the future network status data and used to generate a multi-dimensional performance evaluation data set containing current status and future predictions. The sensors include lightweight quantum computing modules deployed at the edge of the nodes;

[0190] The network status classification result acquisition module 002 calculates the corresponding resource deviation data based on the operating status data and the preset standard performance reference value, and classifies the current network status in real time using the geological classification model to obtain the network status classification result. If a node failure or high load is detected, the weight adjustment threshold of the AHP judgment matrix is ​​lowered and a high-frequency dynamic weight update mode is enabled. If a low-load or idle node is detected, the weight allocation range is expanded and a global resource scheduling strategy is enabled. If a sudden influx of tasks is detected, the AHP sub-indicator weights are dynamically adjusted in combination with the task priority data.

[0191] Conflict level calculation module 003, based on resource deviation data, network status classification results and task priority data, uses a risk assessment algorithm to calculate the performance-cost-carbon efficiency conflict level in the computing power scheduling process. If the conflict level is high risk, an emergency resource recovery strategy is enabled and the scheduling of high-energy consumption nodes is restricted. If the conflict level is medium risk, a multi-objective optimization prompt is triggered and a performance-cost-carbon efficiency trade-off solution is displayed. If the conflict level is low risk, a single indicator optimization suggestion is provided.

[0192] The strategy adjustment signal scheduling module 004 is used to trigger scheduling strategy adjustment signals in a hierarchical manner based on resource deviation data and conflict assessment results. Scheduling strategy adjustment signals include green resource allocation instructions, yellow warning - dynamic weight correction, and red alert - global scheduling reset. If the pressure sensor detects a node anomaly and no conflict assessment is triggered, active scheduling intervention is directly initiated and resource allocation to high-load nodes is restricted.

[0193] The weight matrix update module 005 obtains real-time performance evaluation data from the multidimensional performance evaluation dataset, compares the real-time performance evaluation data with the historical environmental parameter database, and dynamically updates the AHP weight matrix through a reinforcement learning model based on the environmental comparison results. If an abnormal working condition occurs, the attribution analysis of the abnormal condition is performed, and the quantum-classical hybrid algorithm parameters and blockchain verification weights are automatically optimized.

[0194] An embodiment of the present application also discloses a real-time performance evaluation system for a computing power network based on the hierarchical analysis method, including a processor in which a program of any one of the above-mentioned real-time performance evaluation methods for a computing power network based on the hierarchical analysis method is running.

[0195] An embodiment of the present application also discloses a storage medium storing a program of any one of the above-mentioned methods for evaluating the real-time performance of a computing power network based on the hierarchical analysis method.

[0196] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A real-time performance evaluation method for computing power network based on hierarchical analysis method, characterized in that: include: The operating status data and task requirement data of computing network nodes are obtained in real time through distributed sensors and log systems. The operating status data includes CPU utilization, memory occupancy, network bandwidth, task delay, and energy consumption. The task requirement data includes task type, priority, and resource requirements. The quantum-inspired algorithm deployed on the edge node processes the operating status data and generates local performance evaluation indicators. Based on digital twin simulation, the future network status data includes bandwidth change trends and task load predictions. The local performance evaluation indicators are integrated with the future network status data to generate a multi-dimensional performance evaluation data set containing the current status and future predictions. The sensor includes a lightweight quantum computing module deployed at the edge of the node. Based on the operating status data and the preset standard performance reference value, the corresponding resource deviation data is calculated. The current network status is classified in real time using the geological classification model to obtain the network status classification results. If a node failure or high load is detected, the weight adjustment threshold of the AHP judgment matrix is ​​lowered and the high-frequency dynamic weight update mode is enabled. If a low-load or idle node is detected, the weight allocation range is expanded and the global resource scheduling strategy is enabled. If a sudden influx of tasks is detected, the AHP sub-indicator weights are dynamically adjusted based on the task priority data. Based on resource deviation data, network status classification results, and task priority data, a risk assessment algorithm is used to calculate the performance-cost-carbon efficiency conflict level during computing power scheduling. If the conflict level is high risk, an emergency resource recovery strategy is enabled and the scheduling of high-energy-consuming nodes is restricted. If the conflict level is medium risk, a multi-objective optimization prompt is triggered and a performance-cost-carbon efficiency trade-off solution is displayed. If the conflict level is low risk, single-indicator optimization suggestions are provided. Based on resource deviation data and conflict assessment results, scheduling policy adjustment signals are triggered in a hierarchical manner. These signals include green resource allocation instructions, yellow warnings (dynamic weight corrections), and red alerts (global scheduling resets). If the pressure sensor detects a node anomaly and no conflict assessment is triggered, proactive scheduling intervention is initiated directly to limit resource allocation to high-load nodes. Real-time performance evaluation data from a multidimensional performance evaluation dataset is obtained, compared with a historical environmental parameter database, and the AHP weight matrix is ​​dynamically updated through a reinforcement learning model based on the environmental comparison results. If an abnormal operating condition occurs, an attribution analysis is performed on the abnormal condition, and the quantum-classical hybrid algorithm parameters and blockchain verification weights are automatically optimized.

2. The method for evaluating computing power network real-time performance based on the analytic hierarchy process according to claim 1 is characterized in that: When the multi-dimensional performance evaluation data set is integrated with the local performance evaluation indicators and the future network status data, the method further includes: Synchronize local performance indicators with future state data in the time dimension through the timestamp alignment mechanism; Obtain time deviation data between the local performance indicator and the future state data in the time dimension. If the time deviation data exceeds a preset time deviation threshold, perform interpolation compensation on the local performance indicator or downsample the future state data.

3. The method for evaluating computing power network real-time performance based on the analytic hierarchy process according to claim 2 is characterized in that: In the process of hierarchically triggering a scheduling policy adjustment signal based on resource deviation data and conflict assessment results, if the pressure sensor detects a node anomaly and no conflict assessment is triggered, then a constraint condition for active scheduling intervention is added. The method further includes: When the pressure sensor detects that the node has an abnormal operating status, and the abnormal operating status does not reach any risk trigger threshold of the performance-cost-carbon efficiency conflict level, it determines whether the resource deviation data of the abnormal node continues to exceed the preset resource threshold; If at least one indicator in the resource deviation data exceeds the set resource threshold for a continuous period of time that reaches the preset length, the active scheduling intervention mechanism is triggered; When performing active scheduling intervention, priority is given to retaining the resource allocation rights corresponding to nodes whose energy consumption is lower than the benchmark value, and at the same time, resource freezing operations are implemented on high-load nodes whose resource usage exceeds the load limit.

4. The method for evaluating computing power network real-time performance based on the analytic hierarchy process according to claim 1 is characterized in that: In the process of dynamically adjusting the weights of the AHP judgment matrix sub-indicators based on the resource deviation data, the network status classification results, and the task priority data, when a sudden influx of tasks is detected, the method further includes: Identify the urgency of the current task flow based on the task priority data obtained from the log system; If the task priority is judged to be high, the relative importance weight of the task delay indicator in the AHP judgment matrix is ​​increased by 15% to 20%, and the weight of the resource demand indicator is reduced by 5% to 10%; After completing the weight adjustment, the consistency ratio test method is used to verify the consistency of the updated AHP judgment matrix. Only when the consistency ratio CR is less than 0.1, it is confirmed that the weight adjustment is logically reasonable and will be applied to subsequent performance evaluation and scheduling decisions.

5. The method for evaluating computing power network real-time performance based on the analytic hierarchy process according to claim 1 is characterized in that: In the process of simulating future network status data based on digital twins, the method further includes: Build a 3D visual topology model that corresponds one-to-one with the actual computing power network, and synchronously update the connection status, bandwidth change trend and task load distribution between nodes through real-time data streams; Deploy a virtual user behavior simulation module in a virtual environment corresponding to the 3D visualization topology model, retrieve the corresponding historical task pattern, obtain the corresponding user behavior characteristics, and generate dynamic task demand data based on the historical task pattern and the user behavior characteristics. The dynamic task demand data serves as an input parameter for future network state prediction. The dynamic task demand data generated by simulation is integrated with the operation status data for analysis, and the future network status data including the extension of time and space dimensions is output.

6. The method for evaluating computing power network real-time performance based on the analytic hierarchy process according to claim 1 is characterized in that: When abnormal operating conditions occur and attribution analysis is required, the method also includes: Based on historical operating status data and task scheduling logs, a causal reasoning graph is constructed between nodes in the computing power network. Nodes in the causal reasoning graph represent computing power resource units, and edges represent the causal strength between task delays, energy consumption changes, and bandwidth fluctuations. Task delays, energy consumption changes, and bandwidth fluctuations are all performance indicators. Identify the corresponding anomaly propagation path based on the causal reasoning graph, and quantify the causal correlation between the anomaly source node and the downstream affected nodes based on the anomaly propagation path; Based on the abnormal propagation path, a counterfactual virtual scenario is generated to simulate the expected operating state if no abnormality occurs at the key node, and the expected operating state is compared and analyzed with the actual operating state. Based on the comparison results, an attribution report is output to assist in generating root cause-oriented optimization measures, including priority reallocation, node isolation, or parameter adaptive adjustment.

7. A real-time performance evaluation system for computing power network based on hierarchical analysis method, characterized in that: include: A multi-dimensional performance evaluation data set generation module obtains the operating status data and task requirement data of computing power network nodes in real time through distributed sensors and log systems. The operating status data includes CPU utilization, memory occupancy, network bandwidth, task delay, and energy consumption, and the task requirement data includes task type, priority, and resource requirements. The operating status data is processed by a quantum-inspired algorithm deployed on the edge node to generate local performance evaluation indicators. Based on digital twin simulations, future network status data includes bandwidth change trends and task load predictions. The local performance evaluation indicators are integrated with the future network status data to generate a multi-dimensional performance evaluation data set containing current status and future predictions. The sensor includes a lightweight quantum computing module deployed at the edge of the node. The network status classification result acquisition module calculates the corresponding resource deviation data based on the operating status data and the preset standard performance reference value, and classifies the current network status in real time using the geological classification model to obtain the network status classification results. If a node failure or high load is detected, the weight adjustment threshold of the AHP judgment matrix is ​​lowered and the high-frequency dynamic weight update mode is enabled. If a low-load or idle node is detected, the weight allocation range is expanded and the global resource scheduling strategy is enabled. If a sudden influx of tasks is detected, the AHP sub-indicator weights are dynamically adjusted in combination with the task priority data. The conflict level calculation module uses a risk assessment algorithm to calculate the performance-cost-carbon efficiency conflict level in the computing power scheduling process based on resource deviation data, network status classification results, and task priority data. If the conflict level is high risk, an emergency resource recovery strategy is enabled and the scheduling of high-energy-consuming nodes is restricted. If the conflict level is medium risk, a multi-objective optimization prompt is triggered and a performance-cost-carbon efficiency trade-off solution is displayed. If the conflict level is low risk, a single indicator optimization suggestion is provided. The strategy adjustment signal scheduling module triggers scheduling strategy adjustment signals in a hierarchical manner based on resource deviation data and conflict assessment results. Scheduling strategy adjustment signals include green resource allocation instructions, yellow warnings (dynamic weight correction), and red alerts (global scheduling reset). If the pressure sensor detects a node anomaly and no conflict assessment is triggered, active scheduling intervention is directly initiated to limit resource allocation to high-load nodes. The weight matrix update module obtains real-time performance evaluation data from the multidimensional performance evaluation dataset, compares the real-time performance evaluation data with the historical environmental parameter database, and uses the reinforcement learning model to dynamically update the AHP weight matrix based on the environmental comparison results. If an abnormal operating condition occurs, the module performs attribution analysis on the abnormal condition and automatically optimizes the quantum-classical hybrid algorithm parameters and blockchain verification weights.

8. A real-time performance evaluation system for computing power network based on hierarchical analysis method, characterized in that: The method comprises a processor running a program of a real-time performance evaluation method for a computing power network based on the hierarchical analysis method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: A program is stored for the real-time performance evaluation method of a computing power network based on the hierarchical analysis method as described in any one of claims 1 to 6.

Images