Intelligent scheduling system and method based on calculation power demand prediction

Abstract

The invention relates to the technical field of resource scheduling, in particular to an intelligent scheduling system and method based on computing power demand prediction, and the system comprises a task structure analysis module, a dependency evaluation module, a computing power demand quantification module, a resource distribution module and a scheduling dynamic adjustment module. According to the method, the task structure, the dependency relationship, the computing power demand and the resource allocation are accurately modeled and analyzed, the refinement and dynamic optimization of resource scheduling are realized, the task execution duration, the priority and the resource demand are integrated into a multi-dimensional association network, the dependency intensity is quantified, the interaction influence is evaluated, the resource allocation precision is improved, and the resource allocation efficiency is improved. Computing power resources and demands are dynamically adjusted in combination with weighted analysis and a matching algorithm, so that resource allocation is flexible and efficient, real-time dynamic scheduling is achieved, allocation is adjusted according to task progress and resource vacancy, resource conflicts and waste are reduced, the utilization rate is increased, task scheduling efficiency is optimized, and concurrent task allocation rationality and system stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of resource scheduling, and particularly to an intelligent scheduling system and method based on computing power demand prediction. Background Art

[0002] The technical field of resource scheduling includes related technologies for dynamically allocating and optimally configuring computing resources, storage resources, network resources, etc. The core content of this technical field is to reasonably schedule resources through predictive analysis, algorithm optimization, and rule formulation to ensure the maximization of resource utilization in a complex computing environment, while reducing resource conflicts and waste. This field covers a variety of technical solutions, including load balancing scheduling for cloud computing, resource allocation strategies based on computing power demand, priority optimization of large-scale concurrent tasks, etc. Its overall technical field involves the full-process technology from resource demand prediction to the generation of scheduling plans.

[0003] Among them, an intelligent scheduling system refers to the intelligent resource allocation based on computing power demand prediction and task priority. This system mainly addresses the problems in the dynamic demand allocation of computing power resources, covering technical matters such as demand data collection, computing power demand prediction, and the generation and execution of scheduling strategies. Specifically, through the computing power demand prediction technology based on time series analysis, combined with task classification and priority setting algorithms, task scheduling planning is carried out, and finally resource allocation instructions are generated to ensure the dynamic scheduling and reasonable allocation of computing resources.

[0004] The prior art mainly relies on static rules and preset scheduling algorithms in resource scheduling, and it is difficult to cope with high-dynamic and high-concurrent computing demands. Due to the inability to update the status of tasks and resources in real time, the prior art is prone to situations such as uneven resource allocation, idle computing resources, or task delays. For example, in the face of large-scale concurrent tasks, the existing scheduling system cannot flexibly adjust the priority of tasks or the allocation path of computing resources, resulting in some important tasks being unable to be executed in time due to resource competition, and even affecting the stability of the entire system. The prior art is difficult to make timely responses and adjustments to the actual load situation of resources, thereby affecting the optimization of resource utilization. Such problems are particularly prominent in scenarios such as cloud computing, large-scale data processing, and high-performance computing, resulting in waste of computing resources and a decline in the overall system efficiency. Summary of the Invention

[0005] To address the problem in the prior art that in resource scheduling, it mainly relies on static rules and preset scheduling algorithms, making it difficult to handle highly dynamic and highly concurrent computing requirements. Since the status of tasks and resources cannot be updated in real time, the prior art is prone to uneven resource allocation, idle computing resources, or task delays. For example, in the face of a large number of concurrent tasks, the existing scheduling system cannot flexibly adjust the priorities of tasks or the allocation paths of computing resources, resulting in some important tasks being unable to be executed in a timely manner due to resource competition, and even affecting the stability of the entire system. The prior art is difficult to make timely responses and adjustments to the actual load conditions of resources, thereby affecting the optimization of resource utilization. Such problems are particularly prominent in scenarios such as cloud computing, large-scale data processing, and high-performance computing, leading to technical problems of wasted computing resources and reduced overall system efficiency. Embodiments of the present invention provide an intelligent scheduling system and method based on computing power demand prediction. The technical solutions are as follows: On the one hand, an intelligent scheduling system based on computing power demand prediction is provided. The system includes: The task structure parsing module extracts execution duration, priority, dependency strength, CPU occupancy, memory requirements, and bandwidth distribution data according to the task description information, arranges the task dependency path and resource distribution situation, and obtains a task-resource association network model; The dependency relationship evaluation module extracts the connection weights between task nodes and identifies the dependency path lengths based on the task-resource association network model, quantifies the direct and indirect dependency strengths, and summarizes the interaction weights to obtain a task-resource interaction strength data set; The computing power demand quantification module extracts the computing power demand parameters of task nodes based on the task-resource interaction strength data set, combines the availability of resource nodes and task priorities for hierarchical weighted analysis, and obtains a task computing power demand allocation table; The resource allocation module sorts according to the task priorities and demand allocation matching values based on the task computing power demand allocation table, identifies the allocation relationships between task and resource nodes, adjusts the allocation order according to the idle situation of resource nodes, and constructs an initial scheduling resource allocation plan; The scheduling dynamic adjustment module monitors the execution status of task nodes and the idle weights of resource nodes based on the initial scheduling resource allocation plan, compares in real time the computing power demands of unallocated tasks and the remaining capabilities of resources, and fills the idle time slots by adjusting the matching order of task and resource nodes to generate a task dynamic scheduling optimization result.

[0006] As a further solution of the present invention, the task and resource association network model includes task nodes, resource nodes, task and resource attribute matching weights, task dependency paths, and resource distribution conditions. The task and resource interaction intensity data set includes connection weights between task nodes, dependency path lengths, direct dependency intensities, indirect dependency intensities, and task interaction weights. The task computing power demand allocation table includes task node computing power demand parameters, resource node availability, task priority hierarchical weighting values, task path matching values, and resource load conditions. The initial scheduling resource allocation plan includes task and resource node allocation relationships, task priority sorting, demand allocation matching values, and resource node idle conditions. The task dynamic scheduling optimization result includes task node execution status, resource node idle weights, unallocated task computing power demands, remaining resource capabilities, and task execution feedback.

[0007] As a further solution of the present invention, the task structure parsing module includes: The task information extraction sub-module parses the execution duration, priority, dependency intensity, CPU occupancy, memory requirements, and bandwidth distribution according to the task description information, classifies and sorts the parameters by task priority, labels the attribute relationships between tasks, and generates a set of task basic attributes. The resource distribution matching sub-module matches the CPU, memory, and bandwidth parameters in the resource nodes based on the set of task basic attributes, extracts the idle state nodes in the resource distribution, assigns resource node parameters according to task priority, adjusts the allocation order, integrates the task priority and resource allocation conditions, and generates a resource matching result. The task-resource association modeling sub-module sorts out the dependency paths between tasks and resources based on the resource matching result, extracts the distribution of resource nodes associated in the task dependency relationship, assigns weights according to the dependency intensity and integrates them into model data to obtain a task and resource association network model.

[0008] As a further solution of the present invention, the dependency relationship evaluation module includes: The weight acquisition sub-module extracts the connection weights between task nodes based on the task and resource association network model, calls the weight data between nodes in the network model, analyzes the weight distribution in the connection path, classifies and statistically processes the weight data, and generates a node connection weight data table. The path dependency quantification sub-module extracts the cumulative weights of each task node in the path based on the node connection weight data table, analyzes the direct connection intensity between nodes, identifies the dependency intensity, and statistically processes the weight differences between nodes to obtain a dependency intensity quantification result. The interaction impact analysis sub-module analyzes the priority relationship data between task nodes based on the dependency intensity quantification result, calculates the optimization weights of the interaction paths, analyzes the changes in the interaction weights between nodes, and adjusts the interaction impact path relationship to obtain a task and resource interaction intensity data set.

[0009] As a further solution of the present invention, calculate the optimization weight of the interaction path according to the formula: ; Wherein, represents the optimization weight of the interaction path, represents the weight value of each path in the original interaction path, represents the dynamic adjustment coefficient between paths, represents the correlation complexity between nodes in the path, represents the priority factor of the task node, represents the regularization parameter of the adjustment coefficient in the path.

[0010] As a further solution of the present invention, the computing power requirement quantification module includes: The computing power parameter capture sub-module extracts the computing power requirement parameters of the task nodes based on the task and resource interaction intensity data set, analyzes the resource load records, locates the task nodes, classifies and calculates the requirement data, and generates a task node computing power requirement data table; The resource matching weighting sub-module analyzes the availability parameters of the resource nodes based on the task node computing power requirement data table, matches the computing requirements of the tasks and resource nodes, adjusts the resource allocation amount, and obtains a task-resource matching weight data table; The task computing power allocation sub-module extracts the resource allocation weight values based on the task-resource matching weight data table, adjusts the task computing power allocation and resource distribution, and generates a task computing power requirement allocation table.

[0011] As a further solution of the present invention, the resource allocation module includes: The allocation relationship identification sub-module extracts the computing power requirement data and allocation parameters of the task nodes and resource nodes based on the task computing power requirement allocation table, analyzes the association information, matches the computing power allocation relationship, and generates a task-resource allocation relationship table; The allocation order adjustment sub-module extracts the task priority data based on the task-resource allocation relationship table, analyzes the load conflict between the tasks and resource nodes, sorts the task allocation order, adjusts the resource node mapping relationship, and generates a task allocation priority order table; The initial resource allocation sub-module analyzes the allocation order in the task path and the resource node load parameters based on the task allocation priority order table, calculates the initial allocation amount of the adjusted resource nodes, and establishes an initial scheduling resource allocation scheme.

[0012] As a further solution of the present invention, calculate the initial allocation amount of the adjusted resource nodes according to the formula: ; Wherein, represents the initial allocation quantity of the adjusted resource node, represents the task 's weight coefficient, represents the task 's priority, represents the resource node 's current load, represents the load adjustment coefficient, represents the conventional adjustment coefficient, represents the total number of resource nodes.

[0013] As a further solution of the present invention, the scheduling dynamic adjustment module includes: The execution status monitoring sub-module extracts the execution status of the task node and the idle weight value of the resource node based on the initial scheduling resource allocation scheme, monitors the execution time and completion percentage, records the dynamic association data of the task and the resource node, and generates a task and resource status monitoring table; The intelligent task-resource matching sub-module extracts the computing power requirements of the unallocated tasks and the remaining capabilities of the resource nodes based on the task and resource status monitoring table, compares the computing power requirements of the task nodes with the computing capabilities of the resource nodes, updates the task allocation order, and generates a task and resource dynamic matching sequence table; The intelligent dynamic optimization sub-module analyzes the resource dynamic allocation data based on the task and resource dynamic matching sequence table, adjusts the allocation weight value and computing power distribution of the resource node, and generates a task dynamic scheduling optimization result.

[0014] On the other hand, an intelligent scheduling method based on computing power demand prediction, the intelligent scheduling method based on computing power demand prediction is executed based on the above intelligent scheduling system based on computing power demand prediction, and includes the following steps: S1: According to the task description information, extract the execution duration, priority, dependence intensity, CPU occupancy, memory requirement, bandwidth distribution of the task node, analyze the ratio of execution duration to dependence intensity, normalize the CPU occupancy rate and memory requirement, match the relationship between bandwidth distribution and priority, sort out the priority resource matching sequence, and generate an initial matching weight value table of the task node and the resource node; S2: Based on the initial matching weight value table of the task node and the resource node, extract the direct dependence intensity value and path length data of the task node, screen and sort the connection intensity values of the path task nodes, summarize the path dependence distribution and resource occupancy relationship, analyze the difference between the resource node load parameter and the path demand ratio, and generate a task and resource distribution balance data set; S3: Based on the task and resource distribution balance dataset, extract the computing power requirement parameters of task nodes, the CPU occupancy rate and memory carrying value of resource nodes, conduct multi-dimensional comparison of computing power requirements and resource occupancy, classify the idle state values of resource nodes and the cumulative matching parameters of task node priorities, screen the adaptation relationship between the computing power requirement distribution of path task nodes, and generate an adaptation table for task computing power requirements and resource node allocation; S4: Based on the adaptation table for task computing power requirements and resource node allocation, analyze the priority parameters of task nodes and the idle distribution parameters of resource nodes, adjust the priority order of resource allocation for task nodes, reconstruct the load distribution weight value of resource nodes and the resource matching path, screen the resource load conditions of task nodes in the path, and construct an initial task path and resource allocation plan; S5: Based on the initial task path and resource allocation plan, screen the computing power requirement parameters of unallocated task nodes and the idle state values of resource nodes, adjust the priority order of resources within the task node path, re-analyze the idle time slot distribution of resource nodes and the load matching status of task nodes, and generate the task dynamic scheduling optimization result.

[0015] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include: Through precise modeling and analysis of task structure, dependency relationship, computing power requirement, and resource allocation, the refinement and dynamic optimization of resource scheduling are realized. Factors such as task execution duration, priority, and resource requirements are integrated into a multi-dimensional association network between task nodes and resource nodes, thus forming a comprehensive and detailed task-resource interaction model. The interaction model can not only quantify the dependency strength between tasks but also reasonably evaluate the interaction impact between tasks according to the priority relationship, enhancing the intelligence and accuracy of resource allocation. Through weighted analysis and matching algorithms, the task requirements and resource allocation capabilities are dynamically adjusted, enabling the computing power resources to achieve the best ratio with the task requirements. The real-time dynamic scheduling process between tasks and resources can be flexibly adjusted according to the task execution progress and the idle situation of resource nodes, effectively reducing resource conflicts and waste, improving resource utilization rate, being able to quickly respond to changes in task requirements and resource status, making resource allocation more flexible and intelligent, minimizing the problems of resource idleness and scheduling lag to the greatest extent. This method effectively improves the utilization rate of computing resources, optimizes the efficiency of task scheduling, and ensures the reasonable allocation and use of computing resources in the environment of multi-task concurrent execution, thereby enhancing the overall operation efficiency and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0017] Figure 1 is a schematic diagram of an intelligent scheduling system based on computing power demand prediction provided by an embodiment of the present invention; Figure 2 is a schematic diagram of the system framework of the present invention; Figure 3 is a flowchart of the task structure analysis module in the present invention; Figure 4 is a flowchart of the dependency relationship evaluation module in the present invention; Figure 5 is a flowchart of the computing power demand quantification module in the present invention; Figure 6 is a flowchart of the resource allocation module in the present invention; Figure 7 is a flowchart of the scheduling dynamic adjustment module in the present invention; Figure 8 is a flowchart of an intelligent scheduling method based on computing power demand prediction provided by an embodiment of the present invention. Detailed implementation manners

[0018] The following will describe the technical solutions in the present invention with reference to the accompanying drawings.

[0019] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0020] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0021] In the embodiments of the present invention, sometimes subscripts such as W 1It may be written in non-subscript form such as W1. When the difference is not emphasized, the meaning it conveys is the same.

[0022] To make the technical problems to be solved, technical solutions and advantages of the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0023] The embodiment of the present invention provides an intelligent scheduling system based on computing power demand prediction, such as Figure 1-2 the schematic diagram of the intelligent scheduling system based on computing power demand prediction shown in the figure. The system includes: The task structure parsing module extracts the execution duration, priority, dependency strength, CPU occupancy, memory requirement, and bandwidth distribution data according to the task description information, obtains the task nodes and resource nodes, calculates the weights for the matching of task and resource attributes, sorts out the task dependency paths and resource distribution conditions, and obtains the task-resource association network model; The dependency relationship evaluation module extracts the connection weights between task nodes and identifies the dependency path lengths based on the task-resource association network model, quantifies the direct and indirect dependency strengths, evaluates the interaction effects between nodes through the priority relationship between tasks, and summarizes the interaction weights to obtain the task-resource interaction strength data set; The computing power demand quantification module extracts the computing power demand parameters of task nodes based on the task-resource interaction strength data set, performs hierarchical weighted analysis by combining the availability of resource nodes and task priorities, identifies the matching values of the requirements and resource allocation capabilities of task nodes, and combines the task paths and resource load conditions to obtain the task computing power demand allocation table; The resource allocation module sorts according to the task priorities and demand allocation matching values based on the task computing power demand allocation table, identifies the allocation relationships between task and resource nodes, adjusts the allocation order according to the idle conditions of resource nodes, and constructs an initial scheduling resource allocation plan; The scheduling dynamic adjustment module monitors the execution status of task nodes and the idle weights of resource nodes based on the initial scheduling resource allocation plan, compares the computing power demands of unallocated tasks and the remaining capabilities of resources in real time, fills the idle time slots by adjusting the matching order of task and resource nodes, and generates the task dynamic scheduling optimization result in combination with the task execution feedback.

[0024] The task and resource association network model includes task nodes, resource nodes, task-resource attribute matching weights, task dependency paths, and resource distribution. The task and resource interaction intensity dataset includes connection weights between task nodes, dependency path lengths, direct dependency intensities, indirect dependency intensities, and task interaction weights. The task computing power demand allocation table includes task node computing power demand parameters, resource node availability, task priority hierarchical weighted values, task path matching values, and resource load conditions. The initial scheduling resource allocation plan includes task-resource node allocation relationships, task priority sorting, demand allocation matching values, and resource node idle conditions. The task dynamic scheduling optimization result includes task node execution status, resource node idle weights, unallocated task computing power demands, remaining resource capabilities, and task execution feedback.

[0025] Specifically, as Figure 2 , 3 shown, the task structure analysis module includes: The task information extraction sub-module analyzes the execution duration, priority, dependency intensity, CPU occupancy, memory requirements, and bandwidth distribution according to the task description information, classifies and sorts the parameters by task priority, marks the attribute relationships between tasks, and generates a set of task basic attributes; First, calculating the execution duration, priority, dependency intensity, CPU occupancy, memory requirements, and bandwidth distribution of each task is the basic data for task scheduling. The specific operations include using data analysis techniques to extract key attributes from the task description. For example, determining the priority level through text analysis, analyzing the execution duration based on the complexity and urgency of the task content. The attributes are obtained through quantitative analysis and then sorted according to the priority. Using a customized sorting algorithm to ensure that tasks can be executed in the order of urgency and importance. Not only classifying the basic attributes of each task, but also marking the relationships between tasks. For example, the dependency intensity is marked as a relative numerical weight, forming a set of task basic attributes. This set reflects the sequence relationship and resource demand degree between tasks, providing a basis for resource allocation.

[0026] The resource distribution matching sub-module, based on the set of task basic attributes, matches the CPU, memory, and bandwidth parameters in the resource nodes, extracts the idle state nodes in the resource distribution, allocates resource node parameters according to the task priority, adjusts the allocation order, integrates the task priority and resource allocation situation, and generates a resource matching result; Combined with the real-time status of the current resource nodes, perform node-by-node matching. During the execution, match the idle nodes in sequence according to the priority of each task. By calculating the real-time idle rate of the nodes, screen the CPU resources that meet the requirements and record the allocation status. At the same time, based on the memory requirements in the task description and the remaining distribution of the memory capacity in the actual nodes, delimit the available memory range. On the premise of ensuring that the task resource requirements are met, optimize the occupancy ratio of the nodes. During the bandwidth parameter matching process, according to the data traffic involved in the task input and output paths, combined with the current throughput rate in the node network distribution, gradually adjust the bandwidth allocation order, and give priority to reserving data transmission bandwidth for high-priority tasks, forming a resource allocation priority sequence. After integration, adjust the corresponding relationship between the task priority and the node resource allocation, eliminate the resource nodes that cannot be matched, and finally output the complete resource matching result.

[0027] Based on the resource matching result, the task-resource association modeling sub-module sorts out the dependency paths between tasks and resources, extracts the distribution of resource nodes associated in the task dependency relationship, assigns weights according to the dependency strength and integrates them into model data to obtain the task-resource association network model. By sorting out the corresponding relationship between the task distribution and the resource occupancy, extract the strongly dependent resource nodes in the association path. For example, if a task needs to rely on multiple resource nodes to cooperate to complete, then divide the weights of its associated nodes according to the dependency strength. By extracting the dependency relationships item by item in the task description, mark the resource importance weights of each task path. During the integration process of the resource node distribution, assign weights in combination with the dependency strength, aggregate the strongly dependent nodes into the core path, and mark the weakly dependent nodes as redundant paths. According to the balance of the resource node distribution, perform a diversion operation on the nodes with too high dependency weights to avoid the decrease in task execution efficiency caused by the concentration of resources in a single path. After the model integration is completed, generate the task-resource association network model. Analyze the resource conflicts between tasks through the graph structure of the model, and further provide a basis for the optimization of intelligent scheduling to achieve the efficient matching and real-time allocation of tasks and resources.

[0028] Specifically, as Figure 2 、 4 shown, the dependency relationship evaluation module includes: Based on the task-resource association network model, the weight acquisition sub-module extracts the connection weights between task nodes, calls the weight data between nodes in the network model, analyzes the weight distribution in the connection path, classifies the weight data and conducts statistics to generate a node connection weight data table. Extract the connection weights between task nodes from the model one by one, analyze all node connection paths in the associated network model, record the weight distribution on each path, and classify the weights according to the direct association strength and resource allocation ratio between task nodes. During the classification process, archive the path connection weights according to the task priority to form high, medium, and low priority weight sets. When counting the weight data, calculate the weight density of each connection path according to the cumulative sum and frequency distribution of the weight distribution. To ensure the accuracy of the statistical results, check the resource occupancy and path weights item by item, and eliminate invalid path data according to the actual resource usage of the path. Finally, generate a node connection weight data table to provide accurate numerical support for subsequent path dependence and interaction analysis.

[0029] Based on the node connection weight data table, the path dependence quantification sub-module extracts the cumulative weights of each task node in the path, analyzes the direct connection strength between nodes, identifies the dependence strength, and counts the weight differences between nodes to obtain the dependence strength quantification result. Analyze the cumulative weight situation between task nodes in sequence. During the execution, extract the weight distribution of each path, calculate the cumulative weight according to the resource allocation ratio of the nodes, and form a cumulative weight sequence of the nodes within the path. During the analysis of the direct connection strength between nodes, record the weight change rate on the path one by one, and mark the parts with significant weight fluctuations in the path through the weight change trend of the nodes. In terms of dependence strength quantification, calculate the weight difference of each path, and generate a path dependence strength matrix according to the weight difference, which includes the dependence strength between nodes and the weight change data of the connection path. When counting the dependence weights between nodes, match the task priority with the path cumulative weight, delimit the critical path and the secondary path, and generate the dependence strength quantification result in combination with the resource consumption ratio of the path distribution, providing an accurate quantification basis for the interaction impact analysis.

[0030] Based on the dependence strength quantification result, the interaction impact analysis sub-module analyzes the priority relationship data between task nodes, calculates the optimized weight of the interaction path, analyzes the change of the interaction weight between nodes, adjusts the interaction impact path relationship, and obtains the task and resource interaction strength data set. Calculate the optimized weight of the interaction path according to the formula: ; Where, represents the optimized weight of the interaction path, represents the weight value of each path in the original interaction path, represents the dynamic adjustment coefficient between paths, represents the association complexity between nodes in the path, represents the priority factor of the task node, represents the regularization parameter of the adjustment coefficient in the path; Detailed Explanation of the Formula and Derivation Process of Formula Calculation: This formula is used to calculate the optimization weight of the interaction path between task nodes , and the obtained result is used to evaluate and adjust the interaction intensity between tasks and resources; represents the weight value of each path in the original interaction path, which is used to quantify the intensity of the dependency relationship between task nodes. It is obtained by monitoring the communication frequency and data transmission volume between task nodes. The set value is 0.75, which reflects a medium-intensity dependency relationship between nodes; represents the dynamic adjustment coefficient between paths, which is adjusted according to the interaction change weight between paths. It is calculated by real-time monitoring of the interaction change situation between task nodes. The set value is 1.2, indicating that the interaction intensity of the current path has increased; represents the correlation complexity between nodes in the path, which is used to quantify the coupling degree of tasks between nodes. By analyzing the dependency relationship graph between tasks and calculating the direct and indirect dependency quantities between nodes, the set value is 3, reflecting a high degree of coupling with multiple dependency relationships between nodes; represents the priority factor of task nodes, which reflects the difference in priorities between task nodes. It is obtained through the priority parameters set in the task scheduling strategy. The set value is 2, indicating a higher priority for this task node; represents the regularization parameter of the adjustment coefficient in the path, which is used to balance the adjustment amplitude of weights between paths. It is set through empirical parameters to avoid the impact of the overall optimization of the result caused by the excessive bias of path weights towards a single node adjustment. The set value is 0.5; Substitute the parameters into the formula for calculation: ; Result indicates that in the interaction path between current task nodes, the optimized weight after dynamic adjustment is approximately 0.245. This result is used to evaluate the interaction intensity between tasks and resources. The higher the value, the greater the interaction intensity. Through this optimized weight, the task scheduling strategy can be further adjusted, resource allocation optimized, and the overall performance of the system improved.

[0031] Specifically, as Figure 2 、 5 shown, the computing power requirement quantification module includes: The computing power parameter capture sub-module extracts the computing power requirement parameters of task nodes based on the task and resource interaction intensity data set, analyzes the resource load records, locates the task nodes, classifies and calculates the demand data, and generates a task node computing power requirement data table; Extract the computing power requirement parameters for each task node one by one, analyze the interaction intensity of each task node, locate the computing power requirement situation of the current node according to the historical resource call records of the task, extract the resource consumption distribution of the task through the running data in the resource load record, such as CPU occupancy duration, memory demand, and network bandwidth throughput, etc., classify the computing demand data of different tasks. When classifying, separately count the compute-intensive tasks and data-intensive tasks to label the different priority requirements of the tasks for computing power. According to the classification results of the computing power requirements, divide the tasks into high-computing-power requirement nodes and low-computing-power requirement nodes, and generate a task node computing power requirement data table by combining the priority of the tasks and the resource usage efficiency. This data table records the specific values of the computing power requirements of each task node and marks the corresponding relationship with the resource distribution, providing a clear basis for computing power requirements for subsequent resource matching.

[0032] Based on the task node computing power requirement data table, the resource matching weighting sub-module analyzes the availability parameters of the resource nodes, matches the computing requirements of the tasks and the resource nodes, adjusts the resource allocation amount, and obtains a task-resource matching weight data table; Check the idle state of each resource node one by one, combine real-time parameters such as the number of CPU cores, memory capacity, and bandwidth utilization rate to confirm the availability conditions of the nodes. According to the task computing power requirements, perform hierarchical matching on the resource nodes, give priority to allocating high-computing-power requirement tasks to resource nodes with higher performance, and allocate low-computing-power tasks to nodes with more available resources. During the allocation process, gradually adjust the allocation ratio according to the utilization rate of the resource nodes to ensure the load balance of the node resources, obtain a task-resource matching weight data table, record the weight values between each task node and the resource node, use the weight values to represent the priority of the allocation and the degree of adaptation of the computing requirements, provide a basis for subsequent computing power allocation, and at the same time eliminate the redundant paths of resource allocation to ensure the efficient and accurate resource allocation for all tasks.

[0033] Based on the task-resource matching weight data table, the task computing power allocation sub-module extracts the resource allocation weight values, adjusts the task computing power allocation and resource distribution, and generates a task computing power requirement allocation table; Adjust the computing power allocation of tasks and the distribution of resource nodes. During execution, analyze the suitability between the computing power requirements of tasks and resource nodes one by one. Prioritize the allocation of tasks with higher weight values to the nodes with the best availability. At the same time, avoid the concentrated allocation of high-computing-power tasks to prevent the overall system scheduling efficiency from decreasing due to computing power overload of resource nodes. Through the redistribution of weight values, optimize the computing power resource configuration of tasks, record the computing power resource data of task allocation, such as the CPU time, memory capacity, and bandwidth resource ratio allocated to each node, and mark the node location where the task is executed. At the same time, dynamically adjust the allocation table to adapt to the changes of real-time tasks and the improvement of resource utilization efficiency, so as to ensure the dynamic balance between tasks and resources and improve the overall computing performance of intelligent scheduling.

[0034] Specifically, as Figure 2 、 6 shown, the resource allocation module includes: The allocation relationship recognition sub-module extracts the computing power requirement data and allocation parameters of task nodes and resource nodes based on the task computing power requirement allocation table, analyzes the associated information, matches the computing power allocation relationship, and generates a task resource allocation relationship table; Analyze each task node and the corresponding resource node, record the relationship between its computing power requirement and the actual allocation amount, analyze the associated information between tasks and resources, focus on marking the nodes where the computing power requirement exceeds the resource allocation amount, and classify and file this unbalanced allocation situation. When matching the computing power allocation relationship, map the allocation priority of tasks according to the actual computing power utilization rate of resource nodes, and preferentially match the task nodes with high computing power requirements to the nodes with low resource utilization rate to avoid redundant waste of node resources. For the mapped relationship of task resources after allocation, generate a task resource allocation relationship table, record the actual resource occupancy data of each task node, including the CPU allocation time period, memory usage, and bandwidth consumption ratio, etc., and at the same time mark the direct connection strength between the task node and the resource node to provide an accurate allocation basis for subsequent task order adjustment.

[0035] The allocation order adjustment sub-module extracts the task priority data based on the task resource allocation relationship table, analyzes the load conflicts between tasks and resource nodes, sorts the task allocation order, adjusts the resource node mapping relationship, and generates a task allocation priority order table; By sorting the priorities of tasks, the load conflicts between tasks and resource nodes are analyzed item by item. For example, when the allocated resources of a high-priority task are occupied by low-priority tasks, the resource nodes will be released for the high-priority task first. By adjusting the allocation order of tasks, efficient resource utilization is achieved. When sorting the task allocation order, a task resource conflict matrix is ​​generated according to the task resource demand and the resource node load rate. Task paths with larger conflicts are marked as critical paths that need to be adjusted. When adjusting the resource node mapping relationship, low-priority tasks are moved to nodes with lower loads or their allocation order is moved back to ensure the stability of resource allocation for high-priority tasks. Finally, a task allocation priority table is generated, and the adjusted task execution sequence and resource node occupancy are recorded to provide a sequentially optimized scheduling strategy for intelligent scheduling.

[0036] The initial resource allocation submodule analyzes the allocation order and resource node load parameters in the task path based on the task allocation priority table, calculates the adjusted initial allocation amount of resource nodes, and establishes the initial scheduling resource allocation plan; Calculate the adjusted initial allocation of resource nodes according to the formula: ; in, Represents the adjusted initial allocation of resource nodes, Representative tasks The weight coefficient of Representative tasks The priority of Represents a resource node The current load, represents the load adjustment factor, represents the conventional adjustment factor, Represents the total number of resource nodes; Detailed explanation of the formula and the process of formula calculation and derivation: This formula is used to calculate the initial allocation of resource nodes , the purpose is to prioritize tasks based on Load on resource nodes The weighted synthesis of the task priority, resource load and adjustment coefficient is used to calculate the resource allocation amount for each resource node. The parameters in the formula are weighed by the hierarchical analysis method to accurately calculate the resource allocation amount. :Task The weight coefficient of a task indicates its importance in the overall scheduling. The weight coefficient of a task is obtained by comprehensively evaluating the complexity, required resources, and time requirements of the task. It is usually quantified using the analytic hierarchy process (AHP). The complexity assessment is 0.7, the required resource assessment is 0.6, and the task time requirement assessment is 0.5. Then the weight coefficient ; : Task The priority, which represents the execution urgency of the task. The priority is determined by the task's deadline, dependencies, and resource demand period. Set the priority calculation result of task to be 8 (out of 10 full marks). Then , : Resource node The load, which represents the current load level of the node. Usually, it is the ratio of the number of tasks running on the node to the maximum capacity of the node. Set the current load of resource node to be 0.75 (75% load). Then ; : Load adjustment coefficient, which is used to adjust the impact of load on the allocation quantity. It is set based on the assessment of actual demand and load impact. Set this coefficient to 0.9, indicating that the load has a relatively high impact on the final allocation quantity. The adjustment coefficient is 0.9; : Conventional adjustment coefficient, which is used to normalize the load differences of each resource node and avoid excessive deviations. When setting this coefficient, it is generally determined based on the degree of difference in the load of resource nodes. Set , indicating a moderate balance of the load differences of resource nodes; : The total number of resource nodes, which represents the number of resource nodes involved in the scheduling plan. Set the total number of resource nodes to be 3, that is ; Substitute the parameters into the formula for calculation: Given: , , , , , , , , , , , ; First, calculate the numerator part: ; ; ; Then, calculate the denominator part: ; ; Finally, substitute into the complete formula for calculation: ; The result shows that the initial allocation amount of the resource node is 8.57 units, which means that according to the comprehensive evaluation of the task priority and the resource node load, this node needs to be allocated 8.57 units of resources. This value integrates the urgency of the task, the load situation of the resource node, and other adjustment factors, and can ensure the reasonable allocation of resources among multiple nodes.

[0037] Specifically, as Figure 2 , 7 shown, the scheduling dynamic adjustment module includes: The execution status monitoring sub-module extracts the execution status of the task node and the idle weight of the resource node based on the initial scheduling resource allocation plan, monitors the execution time and completion percentage, records the dynamic association data of the task and the resource node, and generates a task and resource status monitoring table; Record the execution status of each task node one by one, including the execution time of the task, the real-time completion percentage, and the current resource consumption. By regularly collecting the computing load of the node during the task execution process, mark the execution progress of each task. When monitoring the idle weight of the resource node, according to the current computing power and load level of the resource node, gradually update the change trend of the idle weight, and analyze the dynamic idle ratio of the resource node at different times. During the recording process of the dynamic association data of the task and the resource node, associate the real-time execution status of the task with the load change of the resource node, generate a comprehensive data table including the task execution time, completion percentage, resource consumption, and node idle status, integrate the above monitoring results to generate a task and resource status monitoring table, clarify the dynamic progress of the task node and the real-time distribution of the resource node, and provide basic data support for subsequent scheduling optimization.

[0038] The intelligent task-resource matching sub-module extracts the computing power requirements of the unallocated tasks and the remaining capabilities of the resource nodes based on the task and resource status monitoring table, compares the computing power requirements of the task nodes with the computing capabilities of the resource nodes, updates the task allocation order, and generates a task and resource dynamic matching sequence table; By comparing the computing power requirements of unassigned task nodes with the remaining computing power of resource nodes, select the nodes with the lowest load and highest availability of resource nodes to preferentially allocate task computing power. During the adjustment process, preferentially allocate high-computing-power-requirement tasks to nodes with stronger computing power and higher idle weights, and appropriately delay the allocation of low-computing-power-requirement tasks or allocate them to the redundant resource part of the nodes. Record the priority changes of task nodes and the actual allocation status of corresponding resource nodes one by one, and mark the adjusted preferential allocation path. Dynamically adjust the allocation order according to the task execution progress and the load of resource nodes, and match the task computing power requirements distribution with the remaining capacity of resources one by one, so as to optimize the real-time order of task allocation and improve the accuracy and execution efficiency of computing power scheduling.

[0039] The intelligent dynamic optimization sub-module analyzes the dynamically allocated resource data based on the dynamic matching sequence table of tasks and resources, adjusts the allocation weights and computing power distribution of resource nodes, and generates the optimized result of task dynamic scheduling; Combined with the current computing power distribution of resource nodes and the computing power requirements of task nodes, re-allocate the computing power of resource nodes, dynamically adjust the allocation weights of each node. For example, for task nodes with a sudden increase in computing power requirements, allocate high-performance resource nodes to them in a timely manner to ensure the smooth completion of the task calculation process. At the same time, appropriately reduce the resource allocation ratio of low-priority task nodes to release more computing power. During the adjustment process of resource allocation weights, conduct a correlation analysis of the real-time load changes of nodes and the trend of task requirements to optimize the balance of computing power distribution among nodes and prevent resource overload or idle situations. When generating the optimized result of task dynamic scheduling, record the allocation optimization of task paths, the dynamic optimization strategy of resource nodes, and the final computing power occupancy ratio of node distribution one by one to ensure that the optimized result is intelligent and efficient.

[0040] Please refer to Figure 8 , the intelligent scheduling method based on computing power demand prediction is executed based on the above intelligent scheduling system based on computing power demand prediction, and includes the following steps: S1: According to the task description information, extract the execution duration, priority, dependency intensity, CPU occupancy, memory requirement, and bandwidth distribution of task nodes, analyze the ratio of execution duration to dependency intensity, normalize the CPU occupancy rate and memory requirement, match the relationship between bandwidth distribution and priority, sort out the priority resource matching sequence, and generate the initial matching weight table of task nodes and resource nodes; S2: Based on the initial matching weight table of task nodes and resource nodes, extract the direct dependency intensity values and path length data of task nodes, screen and sort the connection intensity values of path task nodes, summarize the relationship between path dependency distribution and resource occupancy, analyze the difference between the load parameters of resource nodes and the path requirement ratio, and generate the balanced data set of task and resource distribution; S3: Based on the task and resource distribution equilibrium dataset, extract the computing power requirement parameters of task nodes, the CPU occupancy rate and memory carrying value of resource nodes, conduct multi-dimensional comparison of computing power requirements and resource occupancy, classify the idle state values of resource nodes and the cumulative matching parameters of task node priorities, screen the adaptation relationship between the computing power requirements distribution of path task nodes, and generate an adaptation table for task computing power requirements and resource node allocation; S4: Based on the adaptation table for task computing power requirements and resource node allocation, analyze the priority parameters of task nodes and the idle distribution parameters of resource nodes, adjust the resource priority allocation order of task nodes, reconstruct the load distribution weights of resource nodes and the resource matching path, screen the resource load conditions of task nodes in the path, and construct an initial task path and resource allocation plan; S5: Based on the initial task path and resource allocation plan, screen the computing power requirement parameters of unallocated task nodes and the idle state values of resource nodes, adjust the resource priority order within the task node path, re-analyze the idle time slot distribution of resource nodes and the load matching state of task nodes, and generate the task dynamic scheduling optimization result.

[0041] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An intelligent scheduling system based on computing power demand prediction, characterized by: The system comprises: The task structure parsing module extracts execution time, priority, dependency intensity, CPU usage, memory requirements, and bandwidth distribution data based on the task description information, organizes the task dependency path and resource distribution, and obtains a task and resource association network model; The dependency evaluation module extracts the connection weights between task nodes and identifies the dependency path length based on the task and resource association network model, quantifies the direct and indirect dependency strengths, summarizes the interaction weights, and obtains the task and resource interaction strength dataset; The computing power demand quantification module extracts the computing power demand parameters of the task nodes based on the task and resource interaction intensity data set, performs hierarchical weighted analysis based on the resource node availability and task priority, and obtains the task computing power demand allocation table; The resource allocation module is based on the task computing power demand allocation table, sorts the tasks according to the task priority and demand allocation matching value, identifies the allocation relationship between the tasks and the resource nodes, adjusts the allocation order according to the idle status of the resource nodes, and constructs an initial scheduling resource allocation plan; The scheduling dynamic adjustment module monitors the execution status of task nodes and the idle weight of resource nodes based on the initial scheduling resource allocation plan, compares the computing power requirements of unallocated tasks and the remaining resource capacity in real time, fills the idle time slots by adjusting the matching order of tasks and resource nodes, and generates task dynamic scheduling optimization results.

2. The intelligent scheduling system based on computing power demand prediction according to claim 1 is characterized in that: The task and resource association network model includes task nodes, resource nodes, task and resource attribute matching weights, task dependency paths and resource distribution. The task and resource interaction intensity data set includes connection weights between task nodes, dependency path length, direct dependency intensity, indirect dependency intensity and interaction weights between tasks. The task computing power demand allocation table includes task node computing power demand parameters, resource node availability, task priority hierarchical weighted values, task path matching values ​​and resource load conditions. The initial scheduling resource allocation plan includes the task and resource node allocation relationship, task priority sorting, demand allocation matching values ​​and resource node idle conditions. The task dynamic scheduling optimization results include task node execution status, resource node idle weights, unallocated task computing power requirements, resource remaining capacity and task execution feedback.

3. The intelligent scheduling system based on computing power demand prediction according to claim 1 is characterized in that: The task structure parsing module includes: The task information extraction submodule analyzes the execution time, priority, dependency intensity, CPU usage, memory requirements, and bandwidth distribution according to the task description information, classifies and sorts the parameters according to the task priority, annotates the attribute relationship between tasks, and generates a set of basic task attributes; The resource distribution matching submodule matches the CPU, memory and bandwidth parameters in the resource nodes based on the task basic attribute set, extracts the idle nodes in the resource distribution, allocates resource node parameters according to the task priority, adjusts the allocation order, integrates the task priority and resource allocation, and generates a resource matching result; The task-resource association modeling submodule organizes the dependency paths between tasks and resources based on the resource matching results, extracts the distribution of resource nodes associated with the task dependency relationship, allocates weights according to the dependency strength and integrates them into model data to obtain a task-resource association network model.

4. The intelligent scheduling system based on computing power demand prediction according to claim 1 is characterized in that: The dependency evaluation module includes: The weight acquisition submodule extracts the connection weights between task nodes based on the task and resource association network model, calls the weight data between nodes in the network model, analyzes the weight distribution in the connection path, classifies the weight data and performs statistics, and generates a node connection weight data table; The path dependency quantification submodule extracts the cumulative weight of each task node in the path based on the node connection weight data table, analyzes the direct connection strength between nodes, identifies the dependency strength, and counts the weight differences between nodes to obtain the dependency strength quantification result; Based on the dependency strength quantification results, the interaction impact analysis submodule parses the priority relationship data between task nodes, calculates the optimization weight of the interaction path, analyzes the change of interaction weight between nodes, adjusts the interaction impact path relationship, and obtains the task and resource interaction strength data set.

5. The intelligent scheduling system based on computing power demand prediction according to claim 4 is characterized in that: Calculate the optimization weight of the interaction path according to the formula: ； in, represents the optimization weight of the interaction path, Represents the weight value of each path in the original interaction path, Represents the dynamic adjustment coefficient between paths, represents the association complexity between nodes in the path, Represents the priority factor of the task node, A regularization parameter representing the adjustment coefficients in the path.

6. The intelligent scheduling system based on computing power demand prediction according to claim 1 is characterized in that: The computing power demand quantification module includes: The computing power parameter capture submodule extracts the computing power requirement parameters of the task nodes based on the task and resource interaction intensity data set, parses the resource load records, locates the task nodes, classifies the computing demand data, and generates a task node computing power requirement data table; The resource matching weighted submodule analyzes the availability parameters of the resource nodes based on the task node computing power demand data table, matches the task with the computing demand of the resource node, adjusts the resource allocation amount, and obtains the task resource matching weight data table; The task computing power allocation submodule extracts the resource allocation weight value based on the task resource matching weight data table, adjusts the task computing power allocation and resource distribution, and generates a task computing power demand allocation table.

7. The intelligent scheduling system based on computing power demand prediction according to claim 1 is characterized in that: The resource allocation module comprises: The allocation relationship identification submodule extracts the computing power demand data and allocation parameters of the task nodes and resource nodes based on the task computing power demand allocation table, analyzes the associated information, matches the computing power allocation relationship, and generates a task resource allocation relationship table; The allocation order adjustment submodule extracts task priority data based on the task resource allocation relationship table, analyzes the load conflict between tasks and resource nodes, sorts the task allocation order, adjusts the resource node mapping relationship, and generates a task allocation priority table; The initial resource allocation submodule analyzes the allocation order and resource node load parameters in the task path based on the task allocation priority table, calculates the adjusted initial allocation amount of the resource node, and establishes an initial scheduling resource allocation plan.

8. The intelligent scheduling system based on computing power demand prediction according to claim 7 is characterized in that: Calculate the adjusted initial allocation of resource nodes according to the formula: ； in, Represents the adjusted initial allocation of resource nodes, Representative tasks The weight coefficient of Representative tasks The priority of Represents a resource node The current load, represents the load adjustment factor, represents the conventional adjustment factor, Represents the total number of resource nodes.

9. The intelligent scheduling system based on computing power demand prediction according to claim 1 is characterized in that: The scheduling dynamic adjustment module includes: The execution status monitoring submodule extracts the execution status of the task node and the idle weight of the resource node based on the initial scheduling resource allocation scheme, monitors the execution time and completion percentage, records the dynamic association data of the task and resource node, and generates a task and resource status monitoring table; The intelligent task resource matching submodule extracts the computing power requirements of unassigned tasks and the remaining capacity of resource nodes based on the task and resource status monitoring table, compares the computing power requirements of task nodes with the computing capacity of resource nodes, updates the task allocation order, and generates a task and resource dynamic matching order table; The intelligent dynamic tuning submodule analyzes the dynamic resource allocation data based on the task and resource dynamic matching sequence table, adjusts the allocation weights and computing power distribution of resource nodes, and generates task dynamic scheduling tuning results.

10. An intelligent scheduling method based on computing power demand prediction is characterized in that: The intelligent scheduling system based on computing power demand prediction according to any one of claims 1 to 9 comprises the following steps: S1: According to the task description information, extract the task node execution time, priority, dependency strength, CPU usage, memory requirement, and bandwidth distribution, analyze the execution time and dependency strength ratio, normalize the CPU usage and memory requirement, match the bandwidth distribution and priority relationship, sort out the priority resource matching sequence, and generate the initial matching weight table of task nodes and resource nodes; S2: Based on the initial matching weight table of task nodes and resource nodes, extract the direct dependency strength value and path length data of task nodes, filter and sort the connection strength values ​​of path task nodes, summarize the relationship between path dependency distribution and resource occupancy, analyze the difference between resource node load parameters and path demand ratio, and generate a task and resource distribution balanced data set; S3: Based on the task and resource distribution balance data set, extract the task node computing power requirement parameters and the resource node CPU occupancy rate and memory carrying value, perform a multi-dimensional comparison of computing power requirement and resource occupancy, classify the resource node idle state value and the task node priority cumulative matching parameter, screen the path task node computing power requirement distribution adaptation relationship, and generate the task computing power requirement and resource node allocation adaptation table; S4: Based on the task computing power demand and resource node allocation adaptation table, analyze the task node priority parameters and resource node idle distribution parameters, adjust the task node resource priority allocation order, reconstruct the resource node load distribution weight and resource matching path, screen the task node resource load in the path, and build an initial task path and resource allocation plan; S5: Based on the initial task path and resource allocation plan, filter the computing power requirement parameters of unassigned task nodes and the idle state values ​​of resource nodes, adjust the resource priority order within the task node path, re-analyze the idle time slot distribution of resource nodes and the load matching status of task nodes, and generate task dynamic scheduling optimization results.

Images