Resource scheduling method and device, terminal equipment and storage medium

Abstract

The invention discloses a resource scheduling method and device, terminal equipment and a storage medium, and relates to the technical field of power systems, and the method comprises the steps: converting collected resource data into a resource sensing state value; identifying a feature parameter data set of each task, analyzing a matching relationship between the feature parameter data set and a task demand, screening out a parameter combination matched with the current resource sensing state value from the feature data set according to the matching relationship, and generating a demand modeling parameter set according to the parameter combination; collecting sample data related to the demand modeling parameter set, performing time sequence analysis on the sample data to obtain sample analysis data, comparing difference conditions of task completion rates under different sample analysis data, and adjusting parameters of the modeling parameter set according to the difference conditions to obtain a parameter docking result; and screening out a parameter combination matched with the current sample condition from a parameter docking result, and converting the parameter combination into a resource scheduling strategy. According to the invention, the resource scheduling effect can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of power systems, and more particularly to a resource scheduling method, apparatus, terminal equipment, and storage medium. Background Technology

[0002] In modern power systems, the power grid faces a structural contradiction: peak-hour power shortages coexist with off-peak wind and solar power curtailment. Therefore, strengthening unified scheduling and optimized allocation of resources is particularly important and urgent for mitigating fluctuations and alleviating supply and demand pressures.

[0003] Existing resource scheduling methods lack in-depth analysis of task requirements and resource adaptation characteristics, making it difficult to determine parameter combinations based on task requirements and resource adaptation characteristics to generate resource scheduling strategies, resulting in poor resource scheduling performance. Summary of the Invention

[0004] This invention provides a resource scheduling method, apparatus, terminal device, and storage medium, which can solve the technical problem in the prior art that the lack of in-depth analysis of task requirements and resource adaptation characteristics makes it difficult to determine parameter combinations based on task requirements and resource adaptation characteristics to generate resource scheduling strategies, resulting in poor resource scheduling performance.

[0005] This invention provides a resource scheduling method, comprising: Based on the preset range of dynamic parameters for edge node resources, the resource data of the edge nodes is collected in a regionalized manner, and the collected resource data is converted into resource perception status values. For each task, resource requirement features are identified, a feature parameter dataset is generated based on the identification results, the matching relationship between the feature parameter dataset and task requirements is analyzed, parameter combinations that match the current resource awareness state value are selected from the feature dataset based on the matching relationship, and a requirement modeling parameter set is generated based on the parameter combinations. Collect sample data related to the required modeling parameter set from the training library, perform time series analysis on the sample data to obtain sample analysis data of sample feature distribution in each interval, compare the differences in task completion rate under different sample analysis data, adjust the parameters of the modeling parameter set according to the differences, and obtain parameter docking results. Parameter combinations that match the current sample conditions are selected from the parameter matching results, and the parameter combinations are transformed into resource scheduling strategies.

[0006] Furthermore, the step of regionally collecting resource data of edge nodes according to a preset edge node resource dynamic parameter collection range, and converting the collected resource data into resource perception status values, includes: The edge nodes are divided into different collection areas according to their physical deployment location, logical functional partitions, or resource types; An independent acquisition frequency is set for each acquisition area. Based on the preset acquisition range of dynamic parameters of edge node resources and the acquisition frequency, the resource data of the edge node is acquired in a regionalized manner to obtain resource data. Perform a difference analysis on the resource data, and integrate the resource data into a resource perception status value based on the difference analysis results.

[0007] Furthermore, the step of generating a requirement modeling parameter set based on the parameter combination includes: The matching degree of the parameter combination is evaluated according to the preset selection. When there are multiple parameter combinations with similar matching degrees, the optimal parameter combination is selected from the multiple parameter combinations based on the overall resource utilization rate. The pre-constructed demand modeling parameters are adjusted according to the optimal parameter combination to obtain the adjusted demand modeling parameters, and a demand modeling parameter set is generated based on the adjusted demand modeling parameters.

[0008] Furthermore, the comparison of task completion rates under different sample analysis data, and the adjustment of the parameters of the modeling parameter set based on the differences to obtain parameter alignment results, include: Furthermore, the step of filtering parameter combinations that match the current sample conditions from the parameter matching results and converting the parameter combinations into a resource scheduling strategy includes: Furthermore, after converting the parameter combination into a resource scheduling strategy, the following is also included: Based on the resource scheduling strategy, the resource feature monitoring area of ​​the monitoring edge node is located, and the feature change value of the resource feature monitoring area is extracted. Analyze the influence relationship between the feature change values ​​and real-time task data, and determine the feature resource matching structure based on the influence relationship; Resource allocation is performed based on the aforementioned feature-based resource matching structure.

[0009] Furthermore, after allocating resources based on the aforementioned feature resource matching structure, the process also includes: A resource allocation threshold is determined, and a dynamic task algorithm is applied to perform conflict detection on resource allocation based on the resource allocation threshold.

[0010] The present invention also provides a resource scheduling device, comprising: The resource data conversion module is used to collect resource data of edge nodes in a regionalized manner according to the preset collection range of dynamic parameters of edge node resources, and convert the collected resource data into resource perception status values. The requirement modeling parameter generation module is used to identify resource requirement features for each task, generate a feature parameter dataset based on the identification results, analyze the matching relationship between the feature parameter dataset and task requirements, select parameter combinations that match the current resource awareness state value from the feature dataset based on the matching relationship, and generate a requirement modeling parameter set based on the parameter combinations. The parameter docking result determination module is used to collect sample data related to the requirement modeling parameter set from the training library, perform time series analysis on the sample data to obtain sample analysis data of sample feature distribution in each interval, compare the differences in task completion rate under different sample analysis data, adjust the parameters of the modeling parameter set according to the differences, and obtain the parameter docking result. The resource scheduling strategy generation module is used to filter out parameter combinations that match the current sample conditions from the parameter docking results and convert the parameter combinations into resource scheduling strategies.

[0011] The present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the resource scheduling method described above.

[0012] The present invention also provides a computer-readable storage medium, comprising: a stored computer program, wherein, when the computer program is executed, it controls the device in which the computer-readable storage medium is located to perform the resource scheduling method described above.

[0013] The following benefits can be obtained by implementing the present invention: This invention generates a feature parameter dataset based on the resource requirement feature identification results of each task, and determines the parameters matching the current resource perception state value based on the matching relationship between the feature parameter dataset and the task requirements, so as to generate a modeling parameter set and determine the final parameter combination, and transform it into a resource scheduling strategy. It can deeply consider the correlation between task requirements and resource adaptation features, accurately determine the parameter combination, thereby effectively improving the accuracy of resource scheduling strategy generation, and thus effectively improving the resource scheduling effect.

[0014] Furthermore, this invention, through matching degree evaluation, can accurately screen parameter combinations that match task requirements. When multiple parameter combinations with similar matching degrees exist, further screening is performed based on the overall resource utilization rate to ensure that the selected parameter combinations not only meet task requirements but also maximize resource utilization efficiency. This effectively avoids over-allocation or under-allocation of resources due to inaccurate parameter selection, significantly improving the rationality of resource allocation and the success rate of task execution. Attached Figure Description

[0015] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a resource scheduling method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a resource scheduling device provided in one embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0019] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0022] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0023] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0024] See Figure 1 To address the technical problem in existing technologies that lack in-depth analysis of task requirements and resource adaptation characteristics, making it difficult to determine parameter combinations based on these characteristics to generate resource scheduling strategies, resulting in poor resource scheduling performance, an embodiment of the present invention provides a resource scheduling method, comprising: S1. Based on the preset range of dynamic parameters for edge node resources, regionalize the collection of resource data for edge nodes and convert the collected resource data into resource perception status values. S2. For each task, identify the resource requirement characteristics, generate a feature parameter dataset based on the identification results, analyze the matching relationship between the feature parameter dataset and the task requirements, select parameter combinations that match the current resource awareness state value from the feature dataset based on the matching relationship, and generate a requirement modeling parameter set based on the parameter combinations. S3. Collect sample data related to the requirement modeling parameter set from the training library, perform time series analysis on the sample data to obtain sample analysis data of sample feature distribution in each interval, compare the differences in task completion rate under different sample analysis data, adjust the parameters of the modeling parameter set according to the differences, and obtain the parameter docking results. S4. Filter out parameter combinations that match the current sample conditions from the parameter docking results, and convert the parameter combinations into resource scheduling strategies.

[0025] This invention generates a feature parameter dataset based on the resource requirement feature identification results of each task, and determines the parameters matching the current resource perception state value based on the matching relationship between the feature parameter dataset and the task requirements, so as to generate a modeling parameter set and determine the final parameter combination, and transform it into a resource scheduling strategy. It can deeply consider the correlation between task requirements and resource adaptation features, accurately determine the parameter combination, thereby effectively improving the accuracy of resource scheduling strategy generation, and thus effectively improving the resource scheduling effect.

[0026] In one embodiment, step S1, collecting resource data of edge nodes in a regionalized manner according to a preset edge node resource dynamic parameter collection range, and converting the collected resource data into resource awareness status values, includes: S11. Divide edge nodes into different collection areas according to their physical deployment location, logical functional partitions, or resource types; In this embodiment of the invention, the edge nodes of the industrial site can be divided into sensor access areas, data preprocessing areas, etc., and independent acquisition frequencies and accuracies can be set for each area to ensure that the data acquisition frequency of high-load areas is higher in order to capture subtle load changes.

[0027] S12. Set an independent acquisition frequency for each acquisition area, and based on the preset acquisition range and acquisition frequency of edge node resource dynamic parameters, perform regional acquisition of resource data of edge nodes to obtain resource data. In this embodiment of the invention, the range of dynamic parameters for edge node resources is the real-time change range of core parameters such as CPU utilization, memory usage, storage bandwidth, and network latency.

[0028] In this embodiment of the invention, invalid resource data can be located during the data acquisition process, such as abnormal fluctuations caused by sensor malfunctions or missing data resulting from communication interruptions. These abnormal load data are automatically removed using a preset filtering algorithm and outlier detection mechanism. For example, for CPU utilization data, if the acquired value suddenly exceeds 100% at a certain moment and the duration is less than one sampling period, it is determined to be invalid data and discarded. Then, the extracted stability values ​​(such as the average CPU utilization with fluctuations less than 5% over multiple consecutive sampling periods) and load values ​​(such as the peak memory usage at the current moment) are arranged in regional order to generate a structured resource dataset. This dataset is indexed by timestamps and contains historical sequences of resource parameters for each region, providing raw data support for subsequent analysis.

[0029] S13. Perform differential analysis on the resource data, and integrate the resource data into resource perception status values ​​based on the differential analysis results.

[0030] In this embodiment of the invention, the stability and load of different resource nodes within the same region can be compared horizontally to calculate the ratio of resource parameter changes. For example, the ratio of the difference in CPU processing power between node A and node B can be calculated, and a quantitative indicator can be obtained by (CPU frequency of node A / CPU frequency of node B). Simultaneously, the resource parameters of a single node can be analyzed vertically, such as calculating the rate of change of a node's CPU utilization over two adjacent time periods to reflect the dynamic fluctuations in resource status. Next, the differences between resources are sorted according to a preset weighting system. The weights are set based on the importance of resources in different task scenarios; for example, in real-time data processing scenarios, network latency has a higher weight than storage capacity. Through weighted sorting, areas with excessively large fluctuations are marked. For example, when the network latency difference between nodes in a certain region exceeds 30% of the average latency, that region is marked as a high-risk difference region. Finally, detailed resource-aware difference data is obtained, including the specific values ​​of differences in each dimension, regional distribution, and weighted sorting results.

[0031] In this embodiment of the invention, resource-aware difference values ​​are aggregated across multiple dimensions. These dimensions include resource type (computing resources, storage resources, network resources), regional attributes (geographical location, functional zones), and time span (minutes, hours, days). For example, the CPU utilization differences of all nodes under the computing resource dimension are aggregated hourly to obtain the overall difference trend under this dimension. During the aggregation process, resource-aware numerical differences are screened, and extreme outliers are identified and removed using statistical methods (such as the interquartile range) to ensure data reliability. Then, the difference data are classified according to their numerical magnitude, into categories such as slight differences, moderate differences, and significant differences, each corresponding to a different processing strategy. Finally, the resource-aware values ​​are arranged in an ordered manner to form multi-dimensional resource-aware status values ​​organized by region, resource type, and time series. These status values ​​are stored in matrix form, with each row representing an edge node and each column representing the perception status of a resource parameter, thus providing comprehensive and structured resource status input.

[0032] The embodiments of the present invention can effectively avoid indiscriminate high-frequency acquisition by dividing the area according to physical location, function and other factors and setting independent acquisition frequencies. For non-critical areas, the frequency can be reduced, thereby reducing the computation, storage and network bandwidth occupation of edge nodes and improving the operational stability of nodes.

[0033] In one embodiment, step S2, generating a requirement modeling parameter set based on parameter combinations, includes: S21. Based on the preset selection, accurately evaluate the matching degree of the parameter combination. When there are multiple parameter combinations with similar matching degrees, select the optimal parameter combination from the multiple parameter combinations based on the overall resource utilization rate. In this embodiment of the invention, each task is first identified with fine-grained resource requirement features. The types of tasks can cover a variety of scenarios, such as real-time data acquisition tasks in the Industrial Internet of Things and video analysis tasks in smart cities. The feature dimensions extracted differ for different task types: real-time tasks focus more on the processing speed of computing resources and the low latency of network transmission, while storage tasks focus on the capacity and reliability of storage resources.

[0034] In this embodiment of the invention, during the identification process, adaptation data for the task across various resource dimensions is recorded. For example, a video analysis task may require specific CPU floating-point processing capabilities, GPU image processing capabilities, and minimum network bandwidth thresholds. To ensure comparability of resource data with different dimensions and ranges, this embodiment of the invention normalizes the recorded data, mapping it to the interval [0,1]. For instance, when converting CPU clock speed from GHz to a normalized value, a linear transformation is performed using the maximum CPU clock speed of all edge nodes in the system as a benchmark. After normalization, the data is categorized and organized according to resource type (computing resources, storage resources, network resources) and requirement type (the specific requirements of the task for each type of resource), generating a feature parameter dataset. This dataset, indexed by task ID, contains normalized requirement features for each task across different resource dimensions, providing structured data support for subsequent feature optimization.

[0035] Furthermore, this embodiment of the invention analyzes the matching relationship between resource parameters in the dataset and task requirements. This matching relationship considers not only the adaptability of a single resource dimension but also the synergistic effect of multiple resource dimensions. For example, when determining whether an edge node is suitable for handling real-time data acquisition tasks, it considers not only its CPU processing power but also the memory response speed and network transmission latency for a comprehensive evaluation. This invention can set a matching degree calculation model to select parameter combinations with a high degree of matching with the current resource awareness state. The matching degree calculation model can assign different weights to each resource dimension according to different task scenarios; for example, in industrial control scenarios, the weight of network latency is higher than the weight of storage capacity.

[0036] In this embodiment of the invention, the result of each match can be recorded using a feature matching algorithm, including the matched resource nodes, the corresponding task requirements, and the matching score. Based on these matching results, the parameter combinations are dynamically adjusted. For example, when the matching score of a certain type of task is generally low, the requirement weight of that task in the storage resource dimension is adjusted to improve the overall matching score. Through multiple iterative optimizations, an optimized parameter combination result is generated. This result includes several sets of adjusted resource and requirement parameter combinations, as well as a matching score evaluation index for each combination.

[0037] In this embodiment of the invention, all optimized parameter combinations can be retrieved and comprehensively evaluated according to preset selection criteria (such as matching degree priority, balanced resource utilization, etc.). For example, when there are multiple combinations with similar matching degrees, the combination that can make the overall resource utilization more balanced is selected first to avoid the situation where some nodes are overloaded while some nodes are idle.

[0038] S22. Adjust the pre-built demand modeling parameters according to the optimal parameter combination to obtain the adjusted demand modeling parameters, and generate a demand modeling parameter set based on the adjusted demand modeling parameters.

[0039] In this embodiment of the invention, the demand modeling parameters can be adjusted based on the optimal parameter combination. The adjustment includes the demand threshold and weight coefficient of each resource dimension.

[0040] In this embodiment of the invention, the adjusted requirement modeling parameters are input into a preset generation configuration module. This module simulates the operation in a real-world scenario to verify the stability of the parameter set. The verification process includes stress testing (e.g., testing the adaptability of parameters under high load) and fault tolerance testing (e.g., simulating the robustness of parameters when some resource nodes fail). After successful verification, a final requirement modeling parameter set is generated. This parameter set contains information such as resource requirement characteristics for different task types, matching degree calculation rules, and optimal resource combination strategies, providing accurate requirements modeling basis.

[0041] This invention, through matching degree evaluation, can accurately screen parameter combinations that match task requirements. When multiple parameter combinations with similar matching degrees exist, further screening is performed based on overall resource utilization to ensure that the selected parameter combinations not only meet task requirements but also maximize resource utilization efficiency. This effectively avoids over-allocation or under-allocation of resources due to inaccurate parameter selection, significantly improving the rationality of resource allocation and the success rate of task execution.

[0042] In one embodiment, step S3, comparing the differences in task completion rates under different sample analysis data, adjusting the parameters of the modeling parameter set according to the differences, and obtaining the parameter alignment results, includes: S31. Based on the differences, determine the parameter adjustment weights for the current task; S32. Adjust the weights according to each parameter, and adjust the corresponding parameter combinations in the modeling parameter set; S33. Integrate the adjusted parameter combinations to obtain the parameter docking results.

[0043] In this embodiment of the invention, taking network resources as an example, when a 20% increase in network latency leads to a prolonged task response time in a sample, the influence of this sample characteristic (the change in network latency) on parameter adjustment is calculated. For example, by comparing the differences in task completion rates under different network latencies, the adjustment weight of the network latency parameter is determined. When determining the optimal matching parameter settings based on the influence score, a multi-dimensional parameter adjustment matrix is ​​established. For example, to address the problem of insufficient network bandwidth, parameters such as bandwidth allocation threshold, task priority weight, and data compression rate are adjusted simultaneously. Through iterative iteration, a set of parameter combinations is fine-tuned each time, and its improvement on task scheduling effect is evaluated. For example, in a certain iteration, the bandwidth allocation threshold is increased from 50Mbps to 60Mbps, and the priority weight of real-time tasks is increased by 10%. The average completion time of tasks under this combination is recorded. After dozens of such parameter adjustment cycles, a combination that makes task scheduling efficiency relatively better is captured, and the parameter matching result is obtained. For example, when the network latency exceeds 50ms, the parameter configuration prioritizes scheduling real-time tasks to the backup network channel.

[0044] In one embodiment, step S4, selecting parameter combinations that match the current sample conditions from the parameter matching results and converting the parameter combinations into a resource scheduling strategy, includes: S41. Filter out parameter combinations that match the current sample conditions from the parameter matching results; S42. In the simulation task, different parameter combinations are applied to schedule the task, and the resource utilization and task completion rate are recorded under each task scheduling. S43. Determine the parameter combination of task scheduling corresponding to the resource utilization rate and task completion rate that meet the preset conditions as the final parameter combination; S44. Convert the parameter combination into a resource scheduling strategy.

[0045] In this embodiment of the invention, taking a smart factory scenario as an example, if there is a batch of real-time equipment status monitoring tasks, and the required modeling parameters stipulate that the network latency is less than 30ms and the CPU utilization of computing resources needs to be kept below 70%, then parameter combinations that perform well within the range of 25-30ms network latency and 60-70% CPU utilization are selected from the parameter matching results for parameter training experiments. During the experiment, the equipment data traffic of the smart factory is simulated in the test environment, and different parameter combinations are applied for task scheduling. For example, the first experiment uses a parameter combination that prioritizes network bandwidth allocation, and the second experiment uses a parameter combination that balances computing resource load. Each experiment lasts for 24 hours, and indicators such as resource utilization and task completion rate of each edge node are recorded.

[0046] This invention, through multiple training and verification processes, such as conducting 5-8 rounds of comparative experiments with different parameter combinations, analyzes the performance of each combination at different times (e.g., peak production periods and off-peak periods) to optimize parameter settings. For example, it is found that when network traffic increases during peak production periods, the parameter combination of "dynamic adjustment of network bandwidth threshold + real-time redistribution of task priority" can more effectively ensure the transmission quality of real-time monitoring tasks. Therefore, the parameter settings of this combination are determined and solidified into a policy standard, generating a resource scheduling parameter set that includes network scheduling strategies, computing resource allocation rules, and storage resource reservation thresholds. For instance, it is stipulated that during peak production periods from 8:00 to 18:00 on weekdays, 40% of the network bandwidth is reserved as a minimum for real-time monitoring tasks, and 30% of the core computing power of the CPU is allocated to such tasks.

[0047] This invention verifies the resource utilization and task completion rate of different parameter combinations by simulating tasks, and selects the optimal parameter combination that meets preset conditions to ensure that the resource scheduling strategy can accurately meet the task requirements, thereby effectively improving resource utilization efficiency and task execution success rate.

[0048] In one embodiment, after step S4, which converts the parameter combination into a resource scheduling policy, the method further includes: S5. Based on the resource scheduling strategy, locate the resource feature monitoring area of ​​the monitoring edge node and extract the feature change value of the resource feature monitoring area. In this embodiment of the invention, taking traffic monitoring video analysis in a smart city as an example, the resource feature monitoring area of ​​the traffic monitoring edge node can be located first based on the resource scheduling parameter set. These resource feature monitoring areas include the CPU's floating-point unit utilization, the GPU's image processing core load, memory bandwidth utilization, and the network buffer status of video stream transmission, etc.

[0049] When extracting feature change values ​​in the monitoring area, the feature increase / decrease rate can be continuously recorded at 50-millisecond intervals. For example, when the GPU image processing core load of a certain edge node increases from 30% to 70% within 1 second, its increase / decrease rate is calculated to be 40 percentage points per second. Simultaneously, multiple key change nodes corresponding to the change rate are extracted, such as nodes at 30%, 50%, and 70% during the load increase process. These node values ​​are sorted in chronological order to obtain the current feature change value, forming a feature sequence such as "GPU load reaches 50% at 14:30:20 and 70% at 14:30:21".

[0050] S6. Analyze the influence relationship between feature change values ​​and real-time task data, and determine the feature resource matching structure based on the influence relationship; In this embodiment of the invention, in a traffic monitoring scenario, real-time task data includes the current traffic flow at the intersection, video resolution, and the complexity of the analysis algorithm. For example, when a traffic accident at an intersection causes a surge in traffic flow, and the video analysis task needs to upgrade the resolution from 1080P to 4K, calibration can be performed according to a predetermined matching criterion. This criterion might stipulate that when the video resolution is increased by one level, the GPU computing power requirement increases by 50%, and the memory bandwidth requirement increases by 30%. The matching criterion is invoked to perform feature reallocation within the resource range. For example, 20% of the GPU computing power and 15% of the memory bandwidth originally allocated to other non-real-time tasks are reallocated to this high-definition video analysis task, resulting in a feature resource matching structure that clarifies the reallocation ratio and time window for each resource dimension.

[0051] S7. Resource allocation based on feature-based resource matching structure.

[0052] In this embodiment of the invention, a dynamic resource rule adjustment method can be used to calculate the distribution of features among changing nodes. Taking the process of GPU load increasing from 30% to 70% as an example, the distribution density of this load change on the time axis is calculated to determine the steepness of the load increase. Upper and lower limits are set for node rules, such as specifying a safe upper limit of 85% for GPU load. When the load reaches 70%, an early warning rule is activated, and when it reaches 80%, a resource migration rule is triggered. The rules are applied to regional feature values. For example, when the GPU load of an edge node reaches 75%, some non-real-time video analysis tasks are automatically migrated to adjacent nodes for resource allocation, and a resource allocation threshold is obtained, such as "GPU load threshold is 80%, and a secondary resource scheduling strategy is activated when this threshold is reached."

[0053] In this embodiment of the invention, the resource rule dynamic adjustment method includes obtaining the allocation rule value of matching feature change nodes, and setting a lower limit for node rules through feature weights, resource weights, and dynamic adjustment weights to control minimum matching. In a traffic monitoring scenario, the feature weights can be set as follows: video resolution change accounts for 40%, traffic flow change accounts for 30%, and analysis algorithm complexity accounts for 30%; the resource weights can be set as follows: GPU computing power accounts for 50%, memory bandwidth accounts for 30%, and network transmission accounts for 20%; the dynamic adjustment weights change dynamically according to the current system load, for example, when the overall system load is low, the dynamic adjustment weight is 20%, and when the load is high, it increases to 40%. Through the comprehensive calculation of these three types of weights, a lower limit for node rules is set. For example, when the video resolution is increased to 4K and the traffic flow exceeds the threshold, regardless of the overall system load, at least 30% of the GPU computing power is reserved for this task to ensure that the minimum resource requirements of the critical task are met.

[0054] Taking a specific traffic monitoring edge node as an example, this node was initially allocated 40% of its GPU computing power for video analytics tasks. When the afternoon rush hour arrived at 5 PM, traffic flow at the intersection suddenly increased, requiring the video analytics task to upgrade the resolution from 1080P to 4K. Real-time monitoring showed the GPU load rising from 40% to 65% within 2 minutes, and 65% was identified as the critical change node. Based on real-time task data, it was calculated that an additional 30% of GPU computing power was needed after the resolution upgrade. Therefore, 25% of the GPU computing power allocated to non-real-time data storage tasks at this node was reclaimed, and 5% was borrowed from neighboring nodes, completing the resource reallocation. Based on feature weights, resource weights, and dynamic adjustment weights, the lower limit of the node rule was calculated to be 50% of the GPU computing power. This means that regardless of adjustments, the video analytics task should retain at least 50% of the GPU computing power. Therefore, in this allocation, it was ensured that the task ultimately received no less than 50% of the GPU computing power to avoid video analytics delays due to insufficient resource allocation.

[0055] This invention, through locating and monitoring the resource feature monitoring area of ​​edge nodes and extracting feature change values, can perceive the dynamic changes in resource status in real time. Combined with real-time task data, it can accurately analyze the relationship between feature change values ​​and task requirements, thereby determining the optimal feature resource matching structure and achieving real-time and accurate resource allocation.

[0056] In one embodiment, after step S7, which involves resource allocation based on the feature resource matching structure, the method further includes: S8. Determine the resource allocation threshold and apply the dynamic task algorithm to perform conflict detection on resource allocation based on the resource allocation threshold.

[0057] In this embodiment of the invention, taking the production line equipment monitoring scenario in the Industrial Internet of Things as an example, dynamic task algorithms can be applied based on resource allocation thresholds to capture occupancy point feature data. These occupancy points include the CPU occupancy points of the PLC controller, the memory occupancy points of the sensor data buffer, and the data transmission occupancy points of the Industrial Ethernet, etc.

[0058] When applying dynamic task algorithms, characteristic values ​​are calculated. For example, for the CPU utilization rate of a PLC controller, the characteristic value is calculated by comprehensively considering the data point weight (set to 0.6), the original characteristic value of the utilization point (e.g., the current CPU utilization rate is 75%), the task value (e.g., the priority of the equipment monitoring task is high), and the category coefficient (the category coefficient for industrial control is 1.2). That is, a comprehensive index is obtained by weighted summation of various factors. After generating the characteristic dataset, outliers such as a sudden jump in CPU utilization rate to 120% due to sensor failure are removed. Sampling errors are corrected using the moving average method. The dataset is stored hierarchically in minute-level intervals and subjected to characteristic processing to generate a characteristic dataset containing the comprehensive characteristic value of each utilization point, such as "the characteristic value of the PLC controller CPU is 82 and the characteristic value of the memory cache is 65 during the period from 14:00 to 14:01".

[0059] This invention can characterize datasets, dividing them into intervals according to resource categories, such as PLC controllers, memory caches, and industrial Ethernet. It extracts trends and fluctuation characteristics, for example, analyzing the change curve of the PLC controller CPU characteristic value over the past hour, calculating its upward slope and fluctuation amplitude, and generating a resource category and characteristic change feature set, such as "the PLC controller CPU characteristic value shows a linear upward trend from 14:00 to 15:00, with a slope of 0.5 / minute and a fluctuation amplitude of ±3%".

[0060] This invention can adjust feature parameters and calibrate trend data based on resource categories and feature change feature sets. For example, the feature parameter of industrial Ethernet transmission delay can be adjusted from the original network delay value to a comprehensive delay index considering packet loss rate. Classification intervals can be extracted, such as dividing resource conflict risk into three intervals: low (feature value < 70), medium (70 - 85), and high (> 85), and performing numerical predictions. For instance, based on the current upward trend of the PLC controller CPU feature value, it is predicted that the value may reach 88 in 30 minutes, entering the high-risk interval. This yields conflict distribution analysis values, clarifying the conflict risk level and changing trend of each resource category at different time periods.

[0061] Taking an edge node of an automotive production line as an example, this node is responsible for monitoring the operating status of 10 welding robots. The resource allocation threshold is set at a safe upper limit of 85% for the PLC controller CPU utilization. During a production changeover, a dynamic task algorithm calculates that the characteristic value of the PLC controller CPU rises from 60 to 80 within 15 minutes. The data point weight is set to 0.5 to account for the temporary nature of the changeover task, the task value is set to high priority (corresponding to a value of 10) due to its involvement in equipment safety, and the category coefficient is set to 1.2 according to the industrial control scenario. The characteristic value is calculated comprehensively. The rate of increase of this characteristic value is 1.3 / minute, with small fluctuations, indicating a stable upward trend. Based on the upward trend prediction, if the current production changeover operation continues, the characteristic value will reach 86 after 20 minutes, exceeding the safe upper limit of 85 and entering a high-risk conflict zone. Therefore, a conflict distribution analysis value is generated, indicating that the PLC controller CPU resource has a high risk of conflict within the next 20 minutes.

[0062] This invention can also extract real-time features and allocation data based on standardized output data, such as the characteristic value change curve of the PLC controller CPU and the corresponding resource allocation record, and analyze the real-time scheduling value and allocation value. The scheduling difference (such as the difference between the actual allocated CPU computing power and the computing power required by the task) is matched with the current feature information to generate a feature scheduling error value. For example, it is found that when the characteristic value reaches 80, the actual allocated CPU computing power is still configured according to the initial threshold, resulting in a scheduling error of 10%.

[0063] This invention can also feature scheduling error values, set training and adjustment parameters for the strategy, and set larger adjustment ranges for resource categories with large errors (such as PLC controller CPUs), for example, increasing the scheduling priority weight of such resources by 15%. For low-error categories (such as memory caches), fine-tuning is performed, such as adjusting the upper and lower limits of their allocation thresholds. By comparing the adjustment effects, matching parameter sets are selected and integrated to generate a strategy adjustment parameter set, for example, forming an adjustment rule that "automatically triggers the secondary computing power allocation strategy when the characteristic value of the PLC controller CPU exceeds 80".

[0064] This invention can adjust the parameter set based on the strategy and apply the adjustment parameters in each training stage to optimize the strategy item by item. Taking the changeover task of an automobile production line as an example, in subsequent training, when the CPU characteristic value of the PLC controller is detected to reach 75, the computing power allocation adjustment is initiated in advance. The characteristics and scheduling are monitored synchronously, and it is recorded whether the rate of increase of the characteristic value slows down after the adjustment and whether the scheduling error is reduced. The optimization order of the strategy for each resource category is gradually adjusted to generate a system optimization parameter set. For example, the optimization order of "PLC controller CPU resources are optimized first, followed by memory cache resources" is determined.

[0065] By determining resource allocation thresholds and applying dynamic task algorithms for conflict detection, this invention can effectively identify and avoid conflicts in the resource allocation process, ensuring that multiple tasks do not fail or experience performance degradation due to resource competition, thereby effectively improving the reliability and stability of resource scheduling.

[0066] In one embodiment, strategy adjustments can also be made based on conflict detection results. For example, conflict detection results can be conflict distribution analysis values. By analyzing the error values ​​of features and resources using real-time features and allocation data, and adjusting allocation rules based on these error values, a resource management correction scheme is obtained. After obtaining the resource management correction scheme, the corrected resource parameters are extracted, the parameter fluctuation range is analyzed, and a management threshold range is set and applied to the conflict distribution analysis to obtain a management threshold parameter set. Based on the management threshold parameter set, the resource allocation results are integrated, and data is formatted according to task type to generate standardized output data. The embodiments of the present invention can also collect system operation feedback information based on standardized output data, analyze the sources of resource scheduling errors, adjust strategy training parameters and demand modeling rules, and obtain a set of optimized system parameters.

[0067] In one embodiment, the load parameter set in the resource-aware status value covers specific data on edge nodes such as CPU utilization, memory usage, and storage consumption. This data intuitively reflects the current workload of the nodes. The latency parameter set includes metrics such as data processing latency and response latency, reflecting the temporal characteristics of resources when processing tasks. The perception difference parameter set calculates the degree of difference in various parameters between different resource nodes, such as differences in CPU processing power and memory bandwidth, forming a quantitative description of the differences between resources. These three parameter sets together constitute a comprehensive understanding of the resource status, providing basic data for subsequent processing. The demand feature parameters in the demand modeling parameter set can extract the specific resource requirements of different task types, such as real-time data processing tasks and batch data storage tasks, such as computing power requirements and storage capacity requirements. The load matching parameter is used to measure the fit between resource load and task requirements. For example, when a task requires high computing power, this parameter will focus on the matching between the CPU performance of the resource node and the task requirements. The combination screening parameters, through a series of rules and algorithms, filter combinations of resources and demands. For example, among multiple resources meeting basic conditions, the optimal combination is selected based on factors such as load matching and resource availability, thus forming an accurate set of demand modeling parameters to provide a basis for resource scheduling. The sample change parameters in the resource scheduling parameter set originate from the collection and analysis of historical scheduling data, including changes in resource redundancy, fluctuations in adaptation bias, and the frequency of task conflicts over different time periods. Feature matching parameters determine the degree of correlation between sample features and scheduling strategies; for example, when a certain resource redundancy pattern is detected, this parameter can indicate the corresponding optimal scheduling strategy. Training and adjustment parameters play a role in the training process of the scheduling strategy. By continuously adjusting parameters such as the learning rate and the number of iterations, the performance of the scheduling model is optimized, making the generated resource scheduling parameter set more consistent with actual scheduling needs. The feature change parameters in the resource allocation threshold monitor changes in resource characteristics in real time, such as a sudden increase in CPU utilization or a sharp change in memory usage. Resource matching parameters determine whether the current resource state matches the task requirements; when resource characteristics change to a certain extent, a re-evaluation of resource matching is triggered. The rule setting parameters define specific allocation rules based on feature changes and resource matching. For example, when CPU utilization exceeds 80%, a resource migration mechanism is activated to determine the resource allocation threshold and ensure the rationality of resource allocation. The feature trend analysis parameters in the conflict distribution analysis values ​​predict and analyze the changing trends of resource characteristics, such as predicting the future trend of CPU utilization based on historical data. The resource feature relationship parameters analyze the correlation between different resource characteristics, such as the correlation between CPU utilization and memory usage. Through the analysis of these two parameters, an accurate judgment of the conflict distribution is formed, providing a basis for strategy adjustment.The error analysis parameters in the resource management correction scheme provide a detailed analysis of errors occurring during resource allocation and scheduling, including the type, cause, and impact of the errors. The rule adjustment parameters, based on the error analysis results, adjust and optimize existing allocation rules, such as modifying resource migration trigger conditions and adjusting scheduling strategy weight factors, thereby forming an effective resource management correction scheme and improving the system's resource management efficiency. The parameter fluctuation parameters in the management threshold parameter set monitor the fluctuation of resource parameters after correction in real time, such as the fluctuation range of CPU utilization and the fluctuation amplitude of memory usage. The interval setting parameters set reasonable management threshold intervals based on parameter fluctuations. When resource parameters exceed these intervals, corresponding management measures are triggered, such as alarms and resource reallocation, to ensure that resource parameters fluctuate within a reasonable range. The type integration parameters in the standardized output data integrate different types of resource allocation results, such as integrating computational resource allocation results and storage resource allocation results into a unified dataset. The format processing parameters format the integrated dataset according to different task types and output requirements, such as converting it to CSV or JSON formats, generating standardized output data for easy data storage, transmission, and analysis. The feedback analysis parameters in the system optimization parameter set analyze feedback information collected during system operation, including user feedback and system performance index feedback, to identify the sources of resource scheduling errors. Based on the feedback analysis results, the strategy adjustment parameters adjust the strategy training parameters and requirement modeling rules, such as modifying the structure of the training model and optimizing the feature extraction algorithm for requirement modeling, thereby forming the system optimization parameter set and improving system performance and efficiency.

[0068] Implementing the embodiments of the present invention has the following beneficial effects: This invention generates a feature parameter dataset based on the resource requirement feature identification results of each task, and determines the parameters matching the current resource perception state value based on the matching relationship between the feature parameter dataset and the task requirements, so as to generate a modeling parameter set and determine the final parameter combination, and transform it into a resource scheduling strategy. It can deeply consider the correlation between task requirements and resource adaptation features, accurately determine the parameter combination, thereby effectively improving the accuracy of resource scheduling strategy generation, and thus effectively improving the resource scheduling effect.

[0069] Furthermore, through matching degree evaluation, the embodiments of the present invention can accurately screen out parameter combinations that match the task requirements. When there are multiple parameter combinations with similar matching degrees, further screening is performed based on the overall resource utilization rate to ensure that the selected parameter combination not only meets the task requirements but also maximizes resource utilization efficiency. This can effectively avoid over-allocation or under-allocation of resources due to inaccurate parameter selection, and significantly improve the rationality of resource allocation and the success rate of task execution.

[0070] like Figure 2 As shown, based on the above method embodiments, corresponding apparatus embodiments are provided; An embodiment of the present invention provides a resource scheduling device, comprising: The resource data conversion module 10 is used to collect resource data of edge nodes in a regionalized manner according to the preset collection range of dynamic parameters of edge node resources, and convert the collected resource data into resource perception status values. The requirement modeling parameter generation module 20 is used to identify resource requirement features for each task, generate a feature parameter dataset based on the identification results, analyze the matching relationship between the feature parameter dataset and task requirements, select parameter combinations that match the current resource awareness state value from the feature dataset based on the matching relationship, and generate a requirement modeling parameter set based on the parameter combinations. The parameter docking result determination module 30 is used to collect sample data related to the requirement modeling parameter set from the training library, perform time series analysis on the sample data to obtain sample analysis data of sample feature distribution in each interval, compare the differences in task completion rate under different sample analysis data, adjust the parameters of the modeling parameter set according to the differences, and obtain the parameter docking result. The resource scheduling strategy generation module 40 is used to filter out parameter combinations that match the current sample conditions from the parameter docking results and convert the parameter combinations into resource scheduling strategies.

[0071] In one embodiment, resource data of edge nodes is collected regionally according to a preset range for collecting dynamic parameters of edge node resources, and the collected resource data is converted into resource awareness status values, including: The edge nodes are divided into different collection areas according to their physical deployment location, logical functional partitions, or resource types; An independent collection frequency is set for each collection area. Based on the preset collection range and collection frequency of edge node resource dynamic parameters, the resource data of the edge nodes is collected in a regionalized manner to obtain resource data. Perform differential analysis on resource data, and integrate the resource data into resource perception status values ​​based on the results of the differential analysis.

[0072] In one embodiment, a set of requirements modeling parameters is generated based on a combination of parameters, including: The matching degree of the parameter combination is accurately evaluated according to the preset selection. When there are multiple parameter combinations with similar matching degree, the optimal parameter combination is selected from multiple parameter combinations based on the overall resource utilization. The pre-built demand modeling parameters are adjusted according to the optimal parameter combination to obtain the adjusted demand modeling parameters, and a demand modeling parameter set is generated based on the adjusted demand modeling parameters.

[0073] In one embodiment, the differences in task completion rates under different sample analysis data are compared, and the parameters of the modeling parameter set are adjusted according to the differences to obtain the parameter alignment results, including: Based on the differences, determine the parameter adjustment weights for the current task; Adjust the weights for each parameter and adjust the corresponding parameter combinations in the modeling parameter set. The adjusted parameter combinations are integrated to obtain the parameter docking results.

[0074] In one embodiment, parameter combinations matching the current sample conditions are selected from the parameter matching results, and these parameter combinations are transformed into a resource scheduling strategy, including: Filter the parameter combinations that match the current sample conditions from the parameter matching results; In the simulated task, different parameter combinations were applied for task scheduling, and the resource utilization and task completion rate under each task scheduling were recorded. The parameter combination for task scheduling corresponding to the resource utilization rate and task completion rate that meet the preset conditions is determined as the final parameter combination. Convert parameter combinations into resource scheduling strategies.

[0075] In one embodiment, after converting the parameter combination into a resource scheduling policy, the method further includes: Based on the resource scheduling strategy, the resource feature monitoring area of ​​the monitoring edge node is located, and the feature change value of the resource feature monitoring area is extracted. Analyze the influence relationship between feature change values ​​and real-time task data, and determine the feature resource matching structure based on the influence relationship; Resource allocation is based on feature-based resource matching structures.

[0076] In one embodiment, after resource allocation based on the feature resource matching structure, the method further includes: Determine the resource allocation threshold and apply a dynamic task algorithm to perform conflict detection on resource allocation based on the resource allocation threshold.

[0077] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the resource scheduling method provided by any of the above-described method embodiments of the present invention.

[0078] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0079] Based on the above-described resource scheduling method embodiments, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the resource scheduling method of any embodiment of the present invention.

[0080] For example, in this embodiment, the computer program can be divided into one or more modules, one or more modules are stored in memory and executed by a processor to complete the present invention. One or more module elements can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a terminal device.

[0081] Terminal devices can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. Terminal devices may include, but are not limited to, processors and memory.

[0082] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device through various interfaces and lines.

[0083] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the resource scheduling method of any of the above-described method embodiments of the present invention.

[0084] The modules / units integrated into the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0085] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A resource scheduling method, characterized in that, include: Based on the preset range of dynamic parameters for edge node resources, the resource data of the edge nodes is collected in a regionalized manner, and the collected resource data is converted into resource perception status values. For each task, resource requirement features are identified, a feature parameter dataset is generated based on the identification results, the matching relationship between the feature parameter dataset and task requirements is analyzed, parameter combinations that match the current resource awareness state value are selected from the feature dataset based on the matching relationship, and a requirement modeling parameter set is generated based on the parameter combinations. Collect sample data related to the required modeling parameter set from the training library, perform time series analysis on the sample data to obtain sample analysis data of sample feature distribution in each interval, compare the differences in task completion rate under different sample analysis data, adjust the parameters of the modeling parameter set according to the differences, and obtain parameter docking results. Parameter combinations that match the current sample conditions are selected from the parameter matching results, and the parameter combinations are transformed into resource scheduling strategies.

2. The resource scheduling method as described in claim 1, characterized in that, The step of collecting resource data of edge nodes in a regionalized manner according to a preset range of dynamic parameters for edge node resources, and converting the collected resource data into resource awareness status values, includes: The edge nodes are divided into different collection areas according to their physical deployment location, logical functional partitions, or resource types; An independent acquisition frequency is set for each acquisition area. Based on the preset acquisition range of dynamic parameters of edge node resources and the acquisition frequency, the resource data of the edge node is acquired in a regionalized manner to obtain resource data. Perform a difference analysis on the resource data, and integrate the resource data into a resource perception status value based on the difference analysis results.

3. The resource scheduling method as described in claim 1, characterized in that, The step of generating a set of requirements modeling parameters based on the parameter combination includes: The matching degree of the parameter combination is evaluated according to the preset selection. When there are multiple parameter combinations with similar matching degrees, the optimal parameter combination is selected from the multiple parameter combinations based on the overall resource utilization rate. The pre-constructed demand modeling parameters are adjusted according to the optimal parameter combination to obtain the adjusted demand modeling parameters, and a demand modeling parameter set is generated based on the adjusted demand modeling parameters.

4. The resource scheduling method as described in claim 1, characterized in that, The comparison of task completion rates under different sample analysis data, and the adjustment of the parameters of the modeling parameter set based on the differences to obtain parameter alignment results, include: Based on the aforementioned differences, determine the parameter adjustment weights for the current task; The weights are adjusted according to each parameter, thereby adjusting the corresponding parameter combinations in the modeling parameter set; The adjusted parameter combinations are integrated to obtain the parameter docking results.

5. The resource scheduling method as described in claim 1, characterized in that, The step of selecting parameter combinations that match the current sample conditions from the parameter matching results and converting the parameter combinations into a resource scheduling strategy includes: Filter the parameter combinations that match the current sample conditions from the parameter matching results; In the simulated task, different parameter combinations were applied for task scheduling, and the resource utilization and task completion rate under each task scheduling were recorded. The parameter combination for task scheduling corresponding to the resource utilization rate and task completion rate that meet the preset conditions is determined as the final parameter combination. The parameter combination is converted into a resource scheduling strategy.

6. The resource scheduling method as described in claim 1, characterized in that, After converting the parameter combination into a resource scheduling policy, the following is also included: Based on the resource scheduling strategy, the resource feature monitoring area of ​​the monitoring edge node is located, and the feature change value of the resource feature monitoring area is extracted. Analyze the influence relationship between the feature change values ​​and real-time task data, and determine the feature resource matching structure based on the influence relationship; Resource allocation is performed based on the aforementioned feature-based resource matching structure.

7. The resource scheduling method as described in claim 6, characterized in that, After resource allocation based on the feature resource matching structure, the process further includes: A resource allocation threshold is determined, and a dynamic task algorithm is applied to perform conflict detection on resource allocation based on the resource allocation threshold.

8. A resource scheduling device, characterized in that, include: The resource data conversion module is used to collect resource data of edge nodes in a regionalized manner according to the preset collection range of dynamic parameters of edge node resources, and convert the collected resource data into resource perception status values. The requirement modeling parameter generation module is used to identify resource requirement features for each task, generate a feature parameter dataset based on the identification results, analyze the matching relationship between the feature parameter dataset and task requirements, select parameter combinations that match the current resource awareness state value from the feature dataset based on the matching relationship, and generate a requirement modeling parameter set based on the parameter combinations. The parameter docking result determination module is used to collect sample data related to the requirement modeling parameter set from the training library, perform time series analysis on the sample data to obtain sample analysis data of sample feature distribution in each interval, compare the differences in task completion rate under different sample analysis data, adjust the parameters of the modeling parameter set according to the differences, and obtain the parameter docking result. The resource scheduling strategy generation module is used to filter out parameter combinations that match the current sample conditions from the parameter docking results and convert the parameter combinations into resource scheduling strategies.

9. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the resource scheduling method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the resource scheduling method as described in any one of claims 1-7.

Images