Self-adaptive task scheduling system and method

Abstract

The invention provides a task scheduling method and system based on an unmanned aerial vehicle, electronic equipment and a storage medium, and relates to the field of task scheduling. The method comprises the following steps: acquiring tasks, resources and user behavior data from a user side and a cloud server through an information collection module, and storing the tasks, the resources and the user behavior data into an information base; the task evaluation module analyzes the data and generates a task value score; the scheduling strategy generation module formulates a preliminary execution scheme in combination with a preset rule, and optimizes through a reinforcement learning algorithm to obtain a dynamic optimal scheduling decision; the task redundancy analysis module is used for identifying mergeable redundant tasks and generating a report for further optimizing a scheduling decision; the dynamic priority adjustment module adjusts task priorities and resource allocation according to the real-time monitoring data; and the scheduling execution module is responsible for performing specific execution and generating a state report. The multi-module collaborative design realizes the intellectualization and adaptivity of task scheduling, and improves the utilization rate of system resources.

Description

Technical Field

[0001] This application relates to the field of task scheduling, and particularly to an adaptive task scheduling system and method. Background Art

[0002] With the rapid development of cloud computing technology, the task scheduling problem in large-scale distributed computing environments has become increasingly prominent. In a cloud computing environment, the system needs to simultaneously process a large number of computing tasks from different users, and these tasks have different resource requirements, execution time limits, and priority requirements. In order to improve system resource utilization and ensure the timely completion of tasks, an efficient task scheduling system is required to reasonably allocate computing resources.

[0003] Currently, common task scheduling systems usually adopt rule-based static scheduling strategies, pre-set task priorities and resource allocation rules, and perform task scheduling accordingly. Although this method is simple to implement, due to the inability to identify and process redundant relationships between tasks, there are often situations where similar tasks are repeatedly executed in the system. For example, multiple users may submit tasks with the same or similar resource requirements and computing objectives, and traditional scheduling systems will treat these tasks as independent tasks, allocate resources and execute them separately, resulting in waste of computing resources and reduced overall execution efficiency of the system. Summary of the Invention

[0004] This application provides an adaptive task scheduling system to improve the overall execution efficiency of the system.

[0005] In a first aspect, this application provides an adaptive task scheduling system, which includes: An information collection module, a task evaluation module, a scheduling strategy generation module, a task redundancy analysis module, a dynamic priority adjustment module, and a scheduling execution module; The information collection module is used to regularly obtain task information, system resource information, and user behavior data from the user side and the cloud server, and store the task information, the system resource information, and the user behavior data in an information library; The task evaluation module is used to extract the task information, the system resource information, and the user behavior data from the information library, and evaluate the importance level and resource requirements of each task according to the task information, the system resource information, and the user behavior data, and generate a task value score; The scheduling strategy generation module is used to generate a preliminary task execution plan according to the task value score in combination with a preset scheduling rule, and optimize the preliminary task execution plan through a reinforcement learning algorithm to obtain a dynamic optimal scheduling decision; The task redundancy analysis module is used to mine mergeable redundant tasks from the task information and generate a redundancy report based on the redundant tasks, and feedback the redundancy report to the scheduling policy generation module; The scheduling policy generation module is used to adjust the dynamic optimal scheduling decision according to the redundancy report to obtain a target scheduling decision; The dynamic priority adjustment module is used to obtain the execution progress information of each task in the information library and the load information of each part of the system in real time, and dynamically adjust the task priority and resource allocation scheme according to the execution progress information and the load information to obtain an adjustment result; The scheduling execution module is used to coordinate the cloud server to execute tasks and allocate resources according to the target scheduling decision and the adjustment result, and collect the execution status of the tasks and the resource usage situation to generate an execution status report.

[0006] Through the above technical solutions, the adaptive task scheduling system provided by this application obtains task information, system resource information, and user behavior data in real time through the information collection module, providing comprehensive data support for the system operation. The task evaluation module conducts multi-dimensional evaluations on tasks based on these data and generates task value scores, enabling the importance and resource requirements of tasks to be accurately quantified. On this basis, the scheduling policy generation module combines preset scheduling rules to generate a preliminary task execution plan and uses reinforcement learning algorithms for optimization, realizing the dynamic optimization of the scheduling policy. At the same time, the system innovatively introduces a task redundancy analysis module. By mining mergeable redundant tasks and generating a redundancy report, it provides an optimization basis for the scheduling policy generation module, enabling it to make targeted adjustments to the dynamic optimal scheduling decision, thereby obtaining a better target scheduling decision. In addition, the dynamic priority adjustment module dynamically adjusts the task priority and resource allocation scheme by monitoring the task execution progress information and system load information in real time, further improving the adaptability of the system. Finally, the scheduling execution module efficiently coordinates the cloud server to execute tasks and allocate resources according to the target scheduling decision and the adjustment result, and realizes the closed-loop monitoring of the task execution process through the execution status report. This mechanism of multi-module collaborative work significantly improves the system resource utilization rate, reduces the redundant overhead of task processing, realizes the intelligence and self-adaptability of task scheduling, and effectively solves the problems of resource waste and low scheduling efficiency in traditional scheduling systems.

[0007] In the second aspect of this application, an adaptive task scheduling method is provided, and the system includes: Regularly obtain task information, system resource information, and user behavior data from the user side and the cloud server, and store the task information, the system resource information, and the user behavior data in the information library; Extract the task information, the system resource information, and the user behavior data from the information repository, and evaluate the importance level and resource requirements of each task according to the task information, the system resource information, and the user behavior data to generate a task value score; Generate a preliminary task execution plan according to the task value score in combination with a preset scheduling rule, and optimize the preliminary task execution plan through a reinforcement learning algorithm to obtain a dynamic optimal scheduling decision; The task redundancy analysis module is configured to mine mergeable redundant tasks from the task information, and generate a redundancy report according to the redundant tasks and feedback the redundancy report to the scheduling policy generation module; Adjust the dynamic optimal scheduling decision according to the redundancy report to obtain a target scheduling decision; Obtain in real time the execution progress information of each task in the information repository and the load information of each part of the system, and dynamically adjust the task priority and the resource allocation plan according to the execution progress information and the load information to obtain an adjustment result; Coordinate the cloud server to execute tasks and allocate resources according to the target scheduling decision and the adjustment result, and collect the execution status of the tasks and the resource usage conditions to generate an execution status report.

[0008] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The adaptive task scheduling system provided by this application obtains task information, system resource information, and user behavior data in real time through the information collection module, providing comprehensive data support for the system operation. The task evaluation module conducts multi-dimensional evaluations on tasks based on this data, generating task value scores to accurately quantify the importance and resource requirements of tasks. On this basis, the scheduling policy generation module combines preset scheduling rules to generate a preliminary task execution plan and optimizes it using reinforcement learning algorithms, achieving dynamic optimization of the scheduling policy. At the same time, the system innovatively introduces a task redundancy analysis module. By mining mergeable redundant tasks and generating a redundancy report, it provides an optimization basis for the scheduling policy generation module, enabling it to make targeted adjustments to the dynamic optimal scheduling decision, thereby obtaining a better target scheduling decision. In addition, the dynamic priority adjustment module dynamically adjusts task priorities and resource allocation plans by monitoring task execution progress information and system load information in real time, further improving the adaptability of the system. Finally, the scheduling execution module efficiently coordinates cloud servers to execute tasks and allocate resources according to the target scheduling decision and adjustment results, and realizes closed-loop monitoring of the task execution process through execution status reports. This mechanism of multi-module collaborative work significantly improves system resource utilization, reduces redundant overhead in task processing, realizes the intelligence and self-adaptability of task scheduling, and effectively solves the problems of resource waste and low scheduling efficiency in traditional scheduling systems.

[0009] 2. By setting up a collaborative work mechanism for the status monitoring sub-module, load balancing sub-module, priority calculation sub-module, and resource adjustment sub-module in the dynamic priority adjustment module, the system realizes real-time monitoring and dynamic adjustment of the task execution status. The status monitoring sub-module continuously collects execution progress information and load information, providing real-time operation status data for the system; the load balancing sub-module calculates the load distribution of each part of the system based on this load information, generating load balancing parameters to effectively avoid over-concentrated use of system resources; the priority calculation sub-module combines the execution progress information and load balancing parameters, dynamically calculates the updated task priorities to ensure the timeliness and rationality of task scheduling; the resource adjustment sub-module reallocates system resources according to the updated task priorities, generating an optimized resource adjustment plan, thereby improving the utilization efficiency of system resources. This multi-level dynamic adjustment mechanism not only enhances the system's ability to quickly respond to changes in the task execution status but also realizes the balanced use of system resources, effectively improving the overall task processing efficiency.

[0010] 3. The task evaluation module of this application achieves precise evaluation of task value through hierarchical processing of multiple sub-modules. The data preprocessing sub-module first normalizes the original data to ensure the data quality for subsequent analysis. The task feature analysis sub-module deeply analyzes the preprocessed data to extract feature parameters, providing basic features for task evaluation. The resource requirement evaluation sub-module accurately estimates the CPU resources, memory usage, storage space, and network bandwidth requirements, generating a resource requirement evaluation result to achieve accurate prediction of task resource requirements. The user behavior analysis sub-module analyzes the user's historical task patterns, evaluates priority levels, identifies usage habits, and calculates credit scores to generate a comprehensive user behavior analysis result. The importance calculation sub-module comprehensively scores the feature parameters, resource requirement evaluation result, and user behavior analysis result to obtain an objective comprehensive importance of the task. Finally, the value score generation sub-module generates a task value score through standardization processing. This multi-dimensional evaluation mechanism not only scientifically quantifies the task value but also provides a reliable basis for scheduling decisions, significantly improving the accuracy of system scheduling and resource utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is an architecture diagram of an adaptive task scheduling system provided by an embodiment of this application; Figure 2 It is an architecture diagram of an adaptive task scheduling system provided by an embodiment of this application; Figure 3 It is a flowchart of an adaptive task scheduling method provided by this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0012] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments.

[0013] In the description of the embodiments of this application, words such as "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design solution described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of words such as "for example" or "for instance" aims to present relevant concepts in a specific manner.

[0014] In the description of the embodiments of the present application, the term "plurality" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "comprise", "include", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0015] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0016] Based on the above background technology, further, please refer to Figure 1 , Figure 1 which is an architecture diagram of an adaptive task scheduling system provided by an embodiment of the present application. This system can be implemented depending on a computer program or run as an independent tool-class application. Specifically, in the embodiments of the present application, this method can be applied to a server, but can also be applied to electronic devices such as a server. An adaptive task scheduling system includes: an information collection module 1, a task evaluation module 2, a scheduling policy generation module 3, a task redundancy analysis module 4, a dynamic priority adjustment module 5, and a scheduling execution module 6; The information collection module 1 is used to regularly obtain task information, system resource information, and user behavior data from the user side and the cloud server, and store the task information, the system resource information, and the user behavior data in an information library; Specifically, the information collection module first establishes data connections with the user side and the cloud server. Among them, the user side refers to the access point where the user submits tasks through terminal devices, and the cloud server refers to the physical server cluster that executes computing tasks. The information collection module adopts a timing trigger mechanism and sends data acquisition requests to the user side and the cloud server respectively at preset time intervals. For the user side, the task information obtained by the information collection module includes task attribute data such as task type, task scale, and expected completion time; the system resource information includes resource status data such as processor occupancy rate, memory usage, storage space capacity, and network bandwidth; the user behavior data includes user operation data such as task submission frequency, resource usage habits, and task completion records. For the cloud server, the information collection module obtains operation data such as the task queue status, resource allocation situation, and load level of the current server cluster. The information collection module uniformly encodes and formats the obtained task information, system resource information, and user behavior data to form a standard data structure, and stores it in the information library through a data writing interface. The information library adopts a distributed database architecture, classifies and stores data and establishes indexes, and supports high-concurrency data reading and writing operations. Through this timing acquisition and centralized storage mechanism, the system realizes the comprehensive monitoring of the task execution environment, provides complete data support for subsequent task evaluation and scheduling optimization, and at the same time ensures the real-time and reliability of the data.

[0017] The task evaluation module 2 is used to extract the task information, the system resource information, and the user behavior data from the information library, and evaluate the importance level and resource requirements of each task according to the task information, the system resource information, and the user behavior data, and generate task value scores; Specifically, the task evaluation module reads the task information, the system resource information, and the user behavior data from the information library through the database query interface. First, the task evaluation module preprocesses the read data, including filling in missing values, correcting outliers, and unifying the data format. Then, the task evaluation module extracts characteristic parameters such as the complexity, time sensitivity, and dependency relationship of the task from the task information; estimates the resource requirements such as the processor resources, memory capacity, storage space, and network bandwidth required for task execution from the system resource information; and analyzes the behavior characteristics such as the user's historical task completion rate, resource utilization efficiency, and task urgency from the user behavior data. Next, the task evaluation module quantitatively evaluates the importance of the task using a weighted calculation method, where the weight of task complexity is 0.3, the weight of time sensitivity is 0.25, the weight of resource requirements is 0.25, and the weight of user behavior characteristics is 0.2. Finally, the task evaluation module normalizes the importance evaluation result and the resource requirement evaluation result to generate a task value score between 0 and 100. The higher the score, the more important the task and the more reasonable the resource requirements. Through this multi-dimensional task evaluation mechanism, the system realizes the precise quantification of the value of task execution, provides a scientific decision-making basis for the subsequent optimization of the scheduling strategy, and ensures that the system resources can be preferentially allocated to tasks with higher value.

[0018] Based on the above embodiments, as an optional embodiment, the task evaluation module includes: a data preprocessing sub-module, a task feature analysis sub-module, a resource requirement evaluation sub-module, a user behavior analysis sub-module, an importance calculation sub-module, and a value score generation sub-module; The data preprocessing sub-module is used to extract the task information, the system resource information, and the user behavior data from the information library, and preprocess the task information, the system resource information, and the user behavior data to obtain preprocessed data; Specifically, the data preprocessing sub-module extracts the task information, the system resource information, and the user behavior data from the information library through the database query interface. First, it preprocesses the task information, including removing duplicate task records, supplementing missing task attribute fields, unifying task type identifiers, and standardizing task parameter formats, and converts the task data into a standard data structure. Then, it preprocesses the system resource information, including correcting abnormal resource usage values, filling in missing monitoring data points, eliminating noise data in resource statistics, and unifying resource measurement units, to ensure the accuracy and consistency of resource data. Next, it preprocesses the user behavior data, including cleaning invalid operation records, correcting timestamp errors, unifying user operation types, and normalizing user behavior characteristics, to form a standard user behavior data format. Finally, the data preprocessing sub-module aligns the processed various types of data in time series and unifies the formats to generate preprocessed data. Through this standardized data preprocessing mechanism, the system ensures the quality of the data used in subsequent analysis, provides a clean and complete data basis for task evaluation, and improves the accuracy and reliability of the evaluation results.

[0019] The task feature analysis sub-module is used to analyze the preprocessed data to obtain feature parameters; Specifically, the task feature analysis sub-module first receives the preprocessed data output by the data preprocessing sub-module and extracts the basic feature information of the tasks from it. For each task, it evaluates the computational complexity of the task through computational complexity analysis methods, including the number of computational instructions, the number of data accesses, and the algorithm complexity; it determines the time characteristics of the task through time sensitivity analysis, including the task submission time, the expected completion time, the task execution cycle, and the deadline constraint; it identifies the precedence relationships between tasks through dependency analysis, including data dependencies, resource dependencies, and execution order dependencies. Subsequently, the task feature analysis sub-module combines and analyzes the extracted basic features to calculate the comprehensive feature indicators of the tasks, including the task urgency index, the execution difficulty coefficient, and the criticality score. Finally, the task feature analysis sub-module integrates all the analysis results to generate feature parameters, including task complexity parameters, time constraint parameters, and dependency relationship parameters. Through this multi-dimensional feature analysis mechanism, the system realizes a comprehensive characterization of task features, providing a specific quantitative basis for subsequent resource requirement assessment and importance calculation.

[0020] The resource requirement assessment sub-module is used to estimate the CPU resources, memory usage requirements, storage space requirements, and network bandwidth requirements required by the tasks based on the processed data, and generate a resource requirement assessment result; Specifically, the resource requirement assessment sub-module first extracts the resource usage history and task type information of the task from the pre-processed data. For the estimation of CPU resource requirements, by analyzing the number of computing instructions, parallelism requirements, and historical CPU usage data of the task, the number of processor cores required for task execution and the CPU time slice allocation ratio are calculated; for the estimation of memory usage requirements, by analyzing the dataset size, intermediate result storage volume, and historical memory occupancy data of the task, the memory capacity requirements and memory access patterns during task operation are calculated; for the estimation of storage space requirements, by analyzing the input data volume, output data volume, and temporary file size of the task, the disk space capacity and storage access bandwidth required during task execution are calculated; for the estimation of network bandwidth requirements, by analyzing the data transmission volume, communication frequency, and network latency requirements of the task, the network bandwidth requirements and data transmission patterns during task operation are calculated. Finally, the resource requirement assessment sub-module integrates the resource requirement data for each item to generate a resource requirement assessment result containing specific resource configuration parameters. Through this precise resource requirement assessment mechanism, the system realizes the accurate prediction of task resource requirements, providing a reliable decision-making basis for subsequent resource allocation optimization.

[0021] The user behavior analysis sub-module is used to analyze the user's historical task patterns, evaluate the user's priority level, identify the user's usage habits, calculate the user's credit score, and generate a user behavior analysis result based on the processed data; Specifically, the user behavior analysis sub-module first extracts the user's historical operation records from the pre-processed data. For the analysis of the user's historical task patterns, by statistically analyzing the time distribution, task type distribution, and task scale distribution of the tasks submitted by the user, the task submission rules of the user are identified, including the peak period task submission frequency, common task type preferences, and typical task scale characteristics; for the evaluation of the user's priority level, by analyzing the user's identity type, historical task completion quality, and resource usage rationality, the basic priority level of the user is determined and dynamically adjusted according to the recent task execution situation; for the identification of the user's usage habits, by analyzing the user's resource application pattern, task execution duration preference, and system interaction method, the operation characteristics of the user are summarized, including resource reservation habits, task execution cycles, and operation behavior patterns; for the calculation of the user's credit score, by evaluating the user's resource usage efficiency, task completion rate, and behavior normality, a quantitative index reflecting the user's credibility is generated. Finally, the user behavior analysis sub-module integrates the analysis results for each item to generate a user behavior analysis result in a standard format. Through this systematic user behavior analysis mechanism, the system realizes the precise characterization of user behavior characteristics, providing a personalized reference basis for task priority allocation and resource scheduling.

[0022] The importance calculation sub-module is used to score according to the feature parameters, the resource requirement evaluation result, the user behavior analysis result and a preset scoring rule, so as to obtain the comprehensive importance of the task; The value score generation sub-module is used to generate a standardized task value score according to the comprehensive importance of the task, in combination with a preset scoring rule and weight.

[0023] Specifically, the importance calculation sub-module first receives the feature parameters, the resource requirement evaluation result and the user behavior analysis result, and reads the preset scoring rule from the configuration file. The scoring rule includes a task complexity scoring criterion, a time constraint scoring criterion, a resource efficiency scoring criterion and a user credit scoring criterion. Subsequently, the importance calculation sub-module scores the task complexity parameter, the time constraint parameter and the dependency parameter in the feature parameters, scores the CPU utilization efficiency, the memory usage rationality, the storage space occupancy rate and the network bandwidth usage rate in the resource requirement evaluation result, and scores the user priority level, the historical task completion rate and the resource usage record in the user behavior analysis result. Then, the importance calculation sub-module obtains the comprehensive importance of the task through weighted calculation of the scoring results of each item. Then, the value score generation sub-module maps the importance value to the standard score interval of 0 to 100 according to the comprehensive importance of the task, in combination with the scoring rule and weight parameter preset by the system, so as to generate a standardized task value score. Among them, the scoring rule determines the benchmark score of different types of tasks, and the weight parameter controls the influence degree of each evaluation dimension on the final score. Through this multi-dimensional scoring mechanism, the system realizes the scientific quantification of the task importance degree and provides an accurate value evaluation basis for task scheduling optimization.

[0024] The scheduling strategy generation module 3 is used to generate a preliminary task execution plan according to the task value score in combination with a preset scheduling rule, and optimize the preliminary task execution plan through a reinforcement learning algorithm to obtain a dynamic optimal scheduling decision; Specifically, the scheduling policy generation module first receives the task value scores generated by the task evaluation module and reads the preset scheduling rules from the configuration file. The preset scheduling rules include task priority rules, resource allocation rules, load balancing rules, and time window rules, which define the basic scheduling constraints. The scheduling policy generation module uses a heuristic algorithm to generate a preliminary task execution plan. This algorithm sorts the tasks according to the task value scores and allocates execution time and computing resources according to the preset scheduling rules. Subsequently, the scheduling policy generation module constructs an optimization environment based on deep reinforcement learning and inputs the preliminary task execution plan as the initial state. During the optimization process, the deep reinforcement learning algorithm sets the task queue state, resource usage state, and system load state as the state space, takes task selection, resource allocation, and execution timing arrangement as the action space, and constructs a reward function based on task completion time, resource utilization rate, and load balancing degree. The scheduling policy generation module uses a deep Q-network to estimate the action value function, continuously optimizes the network parameters through an experience replay mechanism, and finally obtains a dynamic optimal scheduling decision through policy iteration. This optimization method based on reinforcement learning enables the system to adaptively adjust the scheduling policy according to the actual operating state, significantly improving the task execution efficiency and resource utilization rate.

[0025] Based on the above embodiments, as an alternative embodiment, the scheduling policy generation module includes: a rule parsing sub-module, a preliminary scheme generation sub-module, and a reinforcement learning optimization sub-module; The rule parsing sub-module is used to receive and parse the preset scheduling rules. The preset scheduling rules include task priority rules, resource allocation rules, load balancing rules, and time window rules, and generate rule parameters; Specifically, the rule parsing sub-module first reads the preset scheduling rules from the configuration file. For the parsing of the task priority rules, by extracting the priority definition criteria, priority adjustment conditions, and priority update period, the basic constraints for the task scheduling order are formed; for the parsing of the resource allocation rules, by extracting the resource quota limits, resource reservation strategies, and resource recovery thresholds, the allocation boundaries of the system resources are determined; for the parsing of the load balancing rules, by extracting the load balancing target value, load transfer trigger conditions, and balance adjustment step sizes, the balance criteria for the system load are set; for the parsing of the time window rules, by extracting the task execution period division, time slice allocation ratio, and scheduling cycle settings, the time frame for task execution is established. Finally, the rule parsing sub-module integrates and formats the parsed rule parameters to generate a standard rule parameter data structure. Through this rule parsing mechanism, the system realizes the structured expression of the scheduling constraints, provides a clear rule basis for the subsequent generation of the scheduling scheme, and ensures the standardization and consistency of the scheduling decisions.

[0026] The preliminary solution generation sub-module is used to generate a preliminary task execution solution that meets the constraint conditions by using a heuristic algorithm according to the task value score and the rule parameters; Specifically, the preliminary solution generation sub-module first receives the task value score output by the task evaluation module and the rule parameters generated by the rule parsing sub-module. Then, it constructs a preliminary task scheduling solution by using a heuristic algorithm. This algorithm first sorts the tasks based on the task value score, and the tasks with higher value scores obtain higher execution priorities. Then, according to the resource allocation rules in the rule parameters, it allocates the required computing resources for each task, including CPU time slices, memory space, and network bandwidth. Subsequently, according to the load balancing rule, it ensures that the load levels of all nodes in the system are balanced by means of task decentralized deployment. Finally, according to the time window rule, it arranges specific execution time periods for each task and ensures that the task execution timing meets the dependency constraint. The preliminary solution generation sub-module integrates the sorting result, the resource allocation plan, and the execution timing arrangement into a preliminary task execution solution. Through this heuristic-based solution generation mechanism, the system quickly constructs a task execution solution that meets the basic constraint conditions, provides an initial solution for subsequent optimization, and improves the generation efficiency of scheduling decisions.

[0027] The reinforcement learning optimization sub-module is used to optimize the preliminary task execution solution based on the reinforcement learning algorithm to obtain the dynamic optimal scheduling decision.

[0028] Specifically, the reinforcement learning optimization sub-module first receives the preliminary task execution solution output by the preliminary solution generation sub-module and takes it as the initial state input for optimization. Subsequently, it constructs a task scheduling environment model, defines the state space as a combination of the current task queue state, the system resource usage state, and the load distribution state; defines the action space as a set of task selection decisions, resource allocation adjustments, and execution timing changes; designs a reward function to evaluate the quality of the scheduling decision, and takes the task completion time, resource utilization rate, and load balancing degree as the basic indicators for reward calculation. Then, the reinforcement learning optimization sub-module uses a deep Q-network to construct a value function estimation model, which maps the state to the action value through a multi-layer neural network structure. During the optimization process, it stores and samples historical decision data through an experience replay mechanism, continuously updates the network parameters, and improves the action selection strategy. Through repeated iterative training, the reinforcement learning algorithm gradually finds a better scheduling strategy and finally outputs the dynamic optimal scheduling decision. Through this reinforcement learning-based optimization mechanism, the system realizes the adaptive optimization of the scheduling solution, improves the resource utilization efficiency, and reduces the task completion time.

[0029] The task redundancy analysis module 4 is used to mine mergeable redundant tasks from the task information and generate a redundancy report based on the redundant tasks and feedback it to the scheduling policy generation module; Specifically, the task redundancy analysis module first reads the task information from the information library, extracts the resource requirement characteristics, computing characteristics, and timing characteristics of each task to generate a task feature vector. The resource requirement characteristics include processor utilization rate, memory occupancy, storage space size, and network bandwidth requirement; the computing characteristics include algorithm type, data processing scale, computing complexity, and parallelism requirement; the timing characteristics include submission time, expected completion time, execution priority, and deadline. Then, the task redundancy analysis module uses a multi-dimensional similarity calculation method to quantitatively evaluate the similarity between tasks, generates a task similarity matrix by calculating the Euclidean distance, cosine similarity, and time overlap degree between feature vectors. At the same time, the task redundancy analysis module analyzes the data dependency relationship, resource competition relationship, and timing constraint relationship between tasks to construct a task dependency graph. Next, the task redundancy analysis module filters high-similarity task pairs according to a preset similarity threshold and determines redundant tasks that meet the merging conditions in combination with the task dependency graph. Finally, the task redundancy analysis module integrates task similarity data, dependency relationship descriptions, and task merging suggestions to generate a standardized redundancy report and feedback the report to the scheduling policy generation module. Through this task redundancy analysis mechanism, the system realizes the intelligent identification and optimized merging of duplicate computing tasks, effectively reduces the waste of computing resources, and improves the overall execution efficiency.

[0030] Based on the above embodiments, as an optional embodiment, the task redundancy analysis module includes: a feature extraction sub-module, a similarity calculation sub-module, a dependency analysis sub-module, a redundancy merging sub-module, and a report generation sub-module; The feature extraction sub-module is used to read the task information from the information library, extract the resource requirement characteristics, computing characteristics, and timing characteristics of each task to generate a task feature vector; Specifically, the feature extraction sub-module first reads task information from the information library through the database interface. For the extraction of resource requirement features, by analyzing the resource configuration requests of tasks, resource parameters such as CPU core number requirements, memory capacity requirements, storage space size, and network bandwidth requirements are extracted; for the extraction of computing features, by analyzing the execution code and data scale of tasks, computing parameters such as algorithm type identifiers, computing complexity levels, data processing volumes, and parallelism requirements are extracted; for the extraction of timing features, by analyzing the submission records and execution requirements of tasks, time parameters such as task arrival time, expected completion time, execution priority, and deadline are extracted. Subsequently, the feature extraction sub-module performs standardization processing on the extracted feature data of each item, unifies the feature values of different dimensions into the same numerical interval, and organizes them in vector form to generate task feature vectors. Through this feature extraction mechanism, the system realizes the comprehensive collection and structured expression of task features, providing a standardized data basis for subsequent similarity calculation and redundancy analysis.

[0031] The similarity calculation sub-module is used to quantitatively evaluate the similarity degree between tasks according to the task feature vectors by using a multi-dimensional similarity calculation method, and generate a task similarity matrix; Specifically, the similarity calculation sub-module first receives the task feature vectors generated by the feature extraction sub-module. For the feature vectors of each pair of tasks, the similarity degree of resource requirement features is evaluated by calculating the Euclidean distance, including CPU usage difference, memory requirement difference, storage space difference, and network bandwidth difference; the similarity degree of computing features is evaluated by calculating the cosine similarity, including algorithm type matching degree, computing complexity proximity, data scale similarity, and parallel requirement consistency; the similarity degree of timing features is evaluated by calculating the time overlap degree, including submission time proximity, completion time overlap degree, and priority proximity. Then, the similarity calculation sub-module uses a weighted fusion method to comprehensively calculate the similarity evaluation results of the three dimensions to obtain the comprehensive similarity value between tasks. Finally, the similarity values between all task pairs are organized in matrix form to generate a task similarity matrix, where the matrix element values are between 0 and 1, and the larger the value, the higher the similarity between tasks. Through this multi-dimensional similarity calculation mechanism, the system realizes the accurate quantification of task similarity relationships, providing reliable data support for subsequent redundant task identification.

[0032] The dependency analysis sub-module is used to construct a task dependency graph according to the task feature vectors; Specifically, the dependency analysis sub-module first receives task feature vectors, and identifies the dependency relationships between tasks by analyzing the data flow relationships, timing constraints, and resource sharing situations in the feature vectors. For the analysis of data flow relationships, by identifying the corresponding relationships between input and output data among tasks, the dependency chain between data production and consumption tasks is determined; for the analysis of timing constraints, by comparing the execution time windows and priority requirements of tasks, the dependency relationships of task execution sequences are established; for the analysis of resource sharing, by detecting the competition relationships of tasks for the same resources, the dependency constraints of resource access sequences are identified. Subsequently, the dependency analysis sub-module converts the identified various dependency relationships into directed edges, organizes task nodes and dependency edges into the form of a directed acyclic graph, and constructs a complete task dependency graph, where the nodes in the graph represent tasks, the edges represent dependency relationships, and the weights of the edges represent dependency strengths. Through this dependency analysis mechanism, the system realizes the graphical expression of task execution constraints, providing a clear reference for task scheduling and resource allocation for dependency constraints.

[0033] The redundancy merging sub-module is used to identify mergeable tasks according to the task similarity matrix and the task dependency graph; Specifically, the redundancy merging sub-module first receives the task similarity matrix generated by the similarity calculation sub-module and the task dependency graph constructed by the dependency analysis sub-module. The task similarity matrix is filtered according to a preset similarity threshold, and task pairs with similarity higher than the threshold are selected as the preliminary merge candidate set; for each candidate task pair, by analyzing the dependency relationships in the task dependency graph, it is checked whether there are circular dependencies between tasks, whether the execution timing constraints are violated, and whether resource competition conflicts occur. If the candidate task pair meets the similarity requirements and does not violate the dependency constraints, it is marked as a mergeable task. For mergeable tasks, the redundancy merging sub-module calculates the resource requirements after merging, evaluates the resource savings effect brought by the merged execution, and determines the final merge plan. Through this redundancy merging mechanism, the system realizes the intelligent identification and optimal combination of duplicate calculation tasks, improving the resource utilization efficiency on the premise of ensuring the correct execution of tasks.

[0034] The report generation sub-module is used to integrate the task similarity matrix, the task dependency graph, and the mergeable tasks to generate a standardized redundancy analysis report.

[0035] Specifically, the report generation sub-module first collects the intermediate results of task redundancy analysis, including the task similarity matrix generated by the similarity calculation sub-module, the task dependency graph constructed by the dependency analysis sub-module, and the mergeable tasks identified by the redundancy merge sub-module. For the processing of the task similarity matrix, by extracting the information of task pairs with high similarity, the task identifiers, similarity values, and similar feature descriptions are recorded; for the processing of the task dependency graph, by analyzing the graph structure, the key dependency paths, dependency strengths, and execution constraint information are extracted; for the processing of mergeable tasks, by counting the specific content of the merge plan, including the combination method of tasks to be merged, the expected resource savings, and the execution timing arrangement after merging. Subsequently, the report generation sub-module organizes and formats this information according to the standard format to form a structured redundancy analysis report, which includes task similarity data, dependency relationship descriptions, merge plan details, and optimization suggestions. Through this report generation mechanism, the system realizes the systematic induction and clear presentation of the redundancy analysis results, providing a complete decision-making basis for the optimization and adjustment of the scheduling strategy.

[0036] The scheduling strategy generation module 5 is used to adjust the dynamic optimal scheduling decision according to the redundancy report to obtain the target scheduling decision; Specifically, the scheduling strategy generation module first receives the redundancy report generated by the task redundancy analysis module, and extracts the redundancy optimization parameters through the redundancy report parsing sub-module. The redundancy optimization parameters include task similarity data, task dependency relationships, and merge suggestion plans. Subsequently, the task reorganization sub-module in the scheduling strategy generation module reorganizes the tasks in the dynamic optimal scheduling decision according to the redundancy optimization parameters, combines tasks with high similarity and meeting the merge conditions into task groups, and generates the first dynamic optimal scheduling decision. Then, the resource reallocation sub-module optimizes and adjusts the resource allocation strategy in the first dynamic optimal scheduling decision according to the reorganized task structure, including recalculating the resource requirements of the merged tasks, adjusting the resource allocation ratio, and optimizing the resource usage timing, so as to obtain the second dynamic optimal scheduling decision. Finally, the timing optimization sub-module optimizes the execution timing of the second dynamic optimal scheduling decision according to the dependency relationship after task merging, adjusts the execution order and time window of the tasks, and finally forms the target scheduling decision. Through this multi-level scheduling strategy optimization mechanism, the system realizes the intelligent merging of redundant tasks and resource reuse, significantly improving the resource utilization efficiency of the system and reducing the overall execution time of tasks while ensuring the correct execution of tasks.

[0037] Based on the above embodiments, as an alternative embodiment, the scheduling strategy generation module further includes: a redundancy report parsing sub-module, a task reorganization sub-module, a resource reallocation sub-module, and a timing optimization sub-module; The redundant report parsing sub-module is used to extract redundant optimization parameters from the redundant analysis report; The task reorganization sub-module is used to reorganize the tasks in the dynamic optimal scheduling decision according to the redundant optimization parameters to obtain a first dynamic optimal scheduling decision; The resource reallocation sub-module is used to optimize and adjust the resource allocation strategy in the first dynamic optimal scheduling decision according to the redundant optimization parameters to obtain a second dynamic optimal scheduling decision; The timing optimization sub-module is used to optimize the execution timing of the second dynamic optimal scheduling decision according to the redundant optimization parameters to obtain the target scheduling decision.

[0038] Specifically, the scheduling strategy generation module performs multi-stage processing on the task redundancy analysis report. First, the redundant report parsing sub-module reads the redundant analysis report and extracts redundant optimization parameters, including task similarity data, combinable task combinations, resource savings expectations, and execution timing suggestions. Subsequently, the task reorganization sub-module adjusts the task structure in the dynamic optimal scheduling decision according to the extracted redundant optimization parameters, combines highly similar tasks into task groups, and re-plans the organization method of tasks to form a first dynamic optimal scheduling decision. Then, the resource reallocation sub-module optimizes the resource allocation strategy in the first dynamic optimal scheduling decision based on the combined task structure, including recalculating the resource requirements of task groups, adjusting the resource allocation ratio, and optimizing the resource usage timing, to generate a second dynamic optimal scheduling decision. Finally, the timing optimization sub-module adjusts the execution timing of the second dynamic optimal scheduling decision according to the dependency relationship and execution constraints of task groups, including arranging the execution order of task groups, coordinating the time windows between tasks, and balancing the load distribution of the system, and finally forms the target scheduling decision. Through this multi-stage optimization processing mechanism, the system realizes the refined adjustment of the scheduling strategy, improves the resource utilization efficiency while ensuring the correct execution of tasks, and reduces the execution time of tasks.

[0039] The dynamic priority adjustment module 6 is used to obtain the execution progress information of each task in the information library and the load information of each part of the system in real time, and dynamically adjust the task priorities and resource allocation plans according to the execution progress information and the load information to obtain an adjustment result; Specifically, the dynamic priority adjustment module establishes a real-time data connection with the information database through the status monitoring sub-module, continuously obtains and monitors the execution progress information and system load information of tasks. The execution progress information includes progress data such as the completion percentage of tasks, the current execution stage, and the remaining execution time; the load information includes system status data such as the processor utilization rate, memory occupancy rate, storage space usage, and network bandwidth occupancy. Subsequently, the load balancing sub-module calculates the load distribution based on the load information of each part of the system, generates load balancing parameters, and is used to evaluate the balance of system resource utilization. Then, the priority calculation sub-module combines the execution progress information of tasks and the load balancing parameters, and uses a dynamic weight calculation method to update the priority value of tasks. For tasks with a lagging execution progress and a relatively light system load, their priorities are increased, and for tasks with excessive resource occupancy, their priorities are appropriately decreased. Then, the resource adjustment sub-module reallocates system resources according to the updated task priorities, generates a resource adjustment plan, including adjusting the processor time slice allocation, re-planning the memory allocation strategy, and optimizing the network bandwidth allocation. Finally, the feedback processing sub-module integrates the task priority value and the resource adjustment plan to generate an adjustment result in a standard format. Through this dynamic priority adjustment mechanism, the system realizes the real-time monitoring and optimization adjustment of the task execution process, ensuring the efficient utilization of system resources and the balanced execution of tasks.

[0040] Based on the above embodiments, as an alternative embodiment, the dynamic priority adjustment module includes: a status monitoring sub-module, a load balancing sub-module, a priority calculation sub-module, a resource adjustment sub-module, and a feedback processing sub-module; The status monitoring sub-module is used to obtain and monitor in real time the execution progress information and the load information of each part of the system in the information database; The load balancing sub-module is used to calculate the load conditions of each part of the system based on the load information and generate load balancing parameters; The priority calculation sub-module is used to dynamically calculate task priorities based on the execution progress information and the load balancing parameters; The resource adjustment sub-module is used to dynamically adjust the system resource allocation according to the task priorities and generate a resource adjustment plan; The feedback processing sub-module is used to integrate the task priority value and the resource adjustment plan to generate an adjustment result.

[0041] Specifically, the scheduling policy generation module processes the task redundancy analysis report in multiple stages. First, the redundancy report parsing sub-module reads the redundancy analysis report and extracts redundancy optimization parameters, including task similarity data, combinable task combinations, expected resource savings, and execution timing suggestions. Subsequently, the task reorganization sub-module adjusts the task structure in the dynamic optimal scheduling decision according to the extracted redundancy optimization parameters, combines highly similar tasks into task groups, re-plans the task organization method, and forms the first dynamic optimal scheduling decision. Then, the resource reallocation sub-module optimizes the resource allocation strategy in the first dynamic optimal scheduling decision based on the combined task structure, including recalculating the resource requirements of the task groups, adjusting the resource allocation ratio, and optimizing the resource usage timing, to generate the second dynamic optimal scheduling decision. Finally, the timing optimization sub-module adjusts the execution timing of the second dynamic optimal scheduling decision according to the dependency relationship and execution constraints of the task groups, including arranging the execution order of the task groups, coordinating the time windows between tasks, and balancing the system load distribution, and finally forms the target scheduling decision. Through this multi-stage optimization processing mechanism, the system realizes the refined adjustment of the scheduling policy, improves the resource utilization efficiency while ensuring the correct execution of tasks, and reduces the task execution time.

[0042] The scheduling execution module 7 is used to coordinate the cloud server to execute tasks and allocate resources according to the target scheduling decision and the adjustment result, and collect the execution status of tasks and the resource usage situation to generate an execution status report.

[0043] Specifically, the scheduling execution module first parses the received target scheduling decision and adjustment result through the instruction parsing sub-module, extracts the task execution order, resource allocation plan, time window setting, and load balancing parameters, and generates a standardized execution instruction. Subsequently, the task distribution sub-module distributes the tasks to the corresponding cloud server execution nodes according to the task type, resource requirements, and priority information in the execution instruction, where the cloud server execution nodes include processor nodes, storage nodes, and network nodes. During the task execution process, the execution monitoring sub-module real-time collects the execution status data of each task, including task completion progress, resource occupancy rate, execution time, and performance metrics; at the same time, it collects the resource usage situation of each node of the cloud server, including processor load, memory usage, storage space occupancy, and network bandwidth consumption. Finally, the execution monitoring sub-module organizes and formats the collected execution status and resource usage data to generate an execution status report containing task execution conditions, resource utilization conditions, and performance evaluation data. Through this refined execution management mechanism, the system realizes the comprehensive monitoring of the task execution process and the accurate recording of resource usage, providing reliable data support for the continuous optimization and performance improvement of the system.

[0044] Based on the above embodiments, as an alternative embodiment, the scheduling execution module includes: an instruction parsing sub-module, a task distribution sub-module, an execution monitoring sub-module, an exception handling sub-module, and a result collection sub-module; The instruction parsing sub-module is used to parse the target scheduling decision and the adjustment result, and generate a standardized execution instruction; The task distribution sub-module is used to allocate tasks to corresponding cloud server execution nodes according to the execution instruction; The execution monitoring sub-module is used to monitor the execution status of tasks and resource usage in real time, and generate the execution status report.

[0045] Specifically, the scheduling execution module realizes the execution management of tasks through multiple sub-modules. First, the instruction parsing sub-module receives the target scheduling decision output by the scheduling policy generation module and the adjustment result generated by the dynamic priority adjustment module, extracts the task execution order, resource allocation plan, time window setting, and load balancing parameters, and converts this information into a standardized execution instruction, including task identification, execution node allocation, resource quota setting, and scheduling time sequence arrangement. Subsequently, the task distribution sub-module allocates tasks to corresponding cloud server execution nodes according to the content of the execution instruction, where the cloud server execution nodes include processor nodes, storage nodes, and network nodes, and each node configures the corresponding execution environment and resource parameters according to the execution instruction. Finally, the execution monitoring sub-module collects the execution status data of each task in real time, including task running progress, resource occupancy, and performance index data, and at the same time collects the resource usage data of each node of the cloud server, including processor load, memory usage, storage space occupancy, and network bandwidth consumption, and collates and summarizes these monitoring data to generate an execution status report in a standard format. Through this refined execution management mechanism, the system realizes the comprehensive control of the task execution process, ensuring the correct execution of tasks and the efficient utilization of resources.

[0046] Based on the above embodiments, as an alternative embodiment, the system further includes: a user interaction module The user interaction module is used to receive task requests and query requests submitted by users, and provide task information to users according to the query requests.

[0047] Specifically, the user interaction module provides an interactive interface for task management. The user submits a task request through the interface. The task request includes information such as task type, execution parameters, resource requirements, and priority requirements. The user interaction module standardizes these task information and stores them in the information repository. When the user submits a query request, the user interaction module retrieves relevant task information from the information repository according to the query conditions, including task execution status, completion progress, resource usage, and estimated completion time, and displays this information to the user in the form of a chart or a list. At the same time, the user interaction module provides a task monitoring function, enabling the user to view the execution progress of the task in real time, including the current execution stage, resource occupancy, and performance metric data. Through this interaction mechanism, the system realizes the effective communication between the user and the task scheduling system, enabling the user to conveniently submit tasks and timely understand the task execution situation, and improving the usability of the system and the user experience.

[0048] Based on the above embodiments, as an optional embodiment, the user interaction module includes: a request parsing unit and a request parsing unit The request parsing unit is used to parse the task request, extract the task type, task parameters, and priority information, and store them in the information repository; The request parsing unit is used to extract relevant task information from the information repository according to the query request.

[0049] Specifically, the user interaction module processes the task request and the query request through the request parsing unit and the query parsing unit respectively. First, the request parsing unit receives the task request submitted by the user, parses and processes the request content, extracts the task type information, including computing tasks, storage tasks, or network tasks; extracts the task parameter information, including the input data scale, processing algorithm type, and output requirements; extracts the priority information, including the task urgency and importance level. The request parsing unit organizes the parsed and extracted information in a standard format and stores it in the information repository through the data writing interface. Subsequently, the query parsing unit receives the query request sent by the user, parses the query conditions and filtering parameters, retrieves the task information that meets the conditions from the information repository, including the task execution status, completion progress, resource occupancy, and estimated completion time, and returns the retrieval result to the user interface for display. Through this request processing mechanism, the system realizes the standardized processing of the user task request and the accurate query of the task information, improving the efficiency and accuracy of the user interaction.

[0050] Please refer to Figure 3 This application also discloses a flowchart of an adaptive task scheduling method provided by an embodiment of this application. The method includes: S101. Regularly obtain task information, system resource information, and user behavior data from the user side and the cloud server, and store the task information, the system resource information, and the user behavior data in the information repository. S102. Extract the task information, the system resource information, and the user behavior data from the information repository, and evaluate the importance level and resource requirements of each task based on the task information, the system resource information, and the user behavior data to generate a task value score. S103. Generate a preliminary task execution plan according to the task value score in combination with a preset scheduling rule, and optimize the preliminary task execution plan through a reinforcement learning algorithm to obtain a dynamic optimal scheduling decision. S104. The task redundancy analysis module is configured to mine mergeable redundant tasks from the task information, and generate a redundancy report based on the redundant tasks and feedback the redundancy report to the scheduling policy generation module. S105. Adjust the dynamic optimal scheduling decision according to the redundancy report to obtain a target scheduling decision. S106. Real-time obtain the execution progress information of each task in the information repository and the load information of each part of the system, and dynamically adjust the task priority and resource allocation plan according to the execution progress information and the load information to obtain an adjustment result. S107. Coordinate the cloud server to execute tasks and allocate resources according to the target scheduling decision and the adjustment result, and collect the execution status of the tasks and the resource usage situation to generate an execution status report.

[0051] The above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. Those skilled in the art will readily conceive of other implementation manners of the present disclosure after considering the specification and the practice of the present disclosure.

[0052] This application aims to cover any variations, uses, or adaptive changes of the present disclosure. These variations, uses, or adaptive changes follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An adaptive task scheduling system, characterized in that: The system includes: an information collection module, a task evaluation module, a scheduling strategy generation module, a task redundancy analysis module, a dynamic priority adjustment module and a scheduling execution module; The information collection module is used to periodically obtain task information, system resource information and user behavior data from the user side and the cloud server, and store the task information, system resource information and user behavior data in the information library; The task evaluation module is used to extract the task information, the system resource information and the user behavior data from the information base, and evaluate the importance and resource requirements of each task according to the task information, the system resource information and the user behavior data to generate a task value score; The scheduling strategy generation module is used to generate a preliminary task execution plan according to the task value score combined with the preset scheduling rules, and optimize the preliminary task execution plan through the reinforcement learning algorithm to obtain a dynamic optimal scheduling decision; The task redundancy analysis module is used to mine the redundant tasks that can be merged from the task information, and generate a redundancy report based on the redundant tasks and feed it back to the scheduling strategy generation module; The scheduling strategy generation module is used to adjust the dynamic optimal scheduling decision according to the redundant report to obtain a target scheduling decision; The dynamic priority adjustment module is used to obtain the execution progress information of each task in the information base and the load information of each part of the system in real time, and dynamically adjust the task priority and resource allocation plan according to the execution progress information and the load information to obtain the adjustment result; The scheduling execution module is used to coordinate the cloud server to execute tasks and allocate resources according to the target scheduling decision and the adjustment result, collect the execution status and resource usage of the task, and generate an execution status report.

2. The system according to claim 1, characterized in that The task evaluation module includes: a data preprocessing submodule, a task feature analysis submodule, a resource demand evaluation submodule, a user behavior analysis submodule, an importance calculation submodule and a value score generation submodule; The data preprocessing submodule is used to extract the task information, the system resource information and the user behavior data from the information base, and preprocess the task information, the system resource information and the user behavior data to obtain preprocessed data; The task feature analysis submodule is used to analyze the preprocessed data to obtain feature parameters; The resource requirement evaluation submodule is used to estimate the CPU resources, memory usage requirements, storage space requirements and network bandwidth requirements required for the task based on the processed data, and generate a resource requirement evaluation result; The user behavior analysis submodule is used to analyze the user's historical task mode, evaluate the user's priority level, identify the user's usage habits, calculate the user's credit score, and generate the user behavior analysis result based on the processed data; The importance calculation submodule is used to score according to the characteristic parameters, the resource demand assessment results, the user behavior analysis results and the preset scoring rules to obtain the comprehensive importance of the task; The value score generating submodule is used to generate a standardized task value score according to the comprehensive importance of the task in combination with preset scoring rules and weights.

3. The system according to claim 1, characterized in that The scheduling strategy generation module includes: a rule parsing submodule, a preliminary plan generation submodule and a reinforcement learning optimization submodule; The rule parsing submodule is used to receive and parse preset scheduling rules, wherein the preset scheduling rules include task priority rules, resource allocation rules, load balancing rules and time window rules, and generate rule parameters; The preliminary plan generation submodule is used to generate a preliminary task execution plan that meets the constraint conditions using a heuristic algorithm according to the task value score and the rule parameters; The reinforcement learning optimization submodule is used to optimize the preliminary task execution plan based on the reinforcement learning algorithm to obtain the dynamic optimal scheduling decision.

4. The system according to claim 1, characterized in that The task redundancy analysis module includes: a feature extraction submodule, a similarity calculation submodule, a dependency analysis submodule, a redundancy merging submodule and a report generation submodule; The feature extraction submodule is used to read the task information from the information base, extract the resource requirement feature, calculation feature and timing feature of each task, and generate a task feature vector; The similarity calculation submodule is used to quantitatively evaluate the similarity between tasks using a multi-dimensional similarity calculation method according to the task feature vectors to generate a task similarity matrix; The dependency analysis submodule is used to construct a task dependency graph according to the task feature vector; The redundant merging submodule is used to identify mergeable tasks according to the task similarity matrix and the task dependency graph; The report generation submodule is used to integrate the task similarity matrix, the task dependency graph and the mergeable tasks to generate a standardized redundancy analysis report.

5. The system according to claim 1, characterized in that The scheduling strategy generation module also includes: a redundant report analysis submodule, a task reorganization submodule, a resource reallocation submodule and a timing optimization submodule; The redundancy report parsing submodule is used to extract the redundancy optimization parameters in the redundancy analysis report; The task reorganization submodule is used to reorganize the tasks in the dynamic optimal scheduling decision according to the redundant optimization parameters to obtain a first dynamic optimal scheduling decision; The resource reallocation submodule is used to optimize and adjust the resource allocation strategy in the first dynamic optimal scheduling decision according to the redundant optimization parameter to obtain a second dynamic optimal scheduling decision; The timing optimization submodule is used to optimize the execution timing of the second dynamic optimal scheduling decision according to the redundant optimization parameters to obtain the target scheduling decision.

6. The system according to claim 1, characterized in that The dynamic priority adjustment module includes: a state monitoring submodule, a load balancing submodule, a priority calculation submodule, a resource adjustment submodule and a feedback processing submodule; The state monitoring submodule is used to obtain and monitor the execution progress information in the information base and the load information of each part of the system in real time; The load balancing submodule is used to calculate the load conditions of various parts of the system according to the load information and generate load balancing parameters; The priority calculation submodule is used to dynamically calculate the task priority according to the execution progress information and the load balancing parameter; The resource adjustment submodule is used to dynamically adjust the system resource allocation according to the task priority and generate a resource adjustment plan; The feedback processing submodule is used to integrate the task priority value and the resource adjustment plan to generate an adjustment result.

7. The system according to claim 1, characterized in that The scheduling execution module includes: an instruction parsing submodule, a task distribution submodule, an execution monitoring submodule, an exception handling submodule and a result collection submodule; The instruction parsing submodule is used to parse the target scheduling decision and the adjustment result to generate standardized execution instructions; The task distribution submodule is used to distribute the task to the corresponding cloud server execution node according to the execution instruction; The execution monitoring submodule is used to monitor the execution status and resource usage of the task in real time and generate the execution status report.

8. The system according to claim 1, characterized in that The system also includes: a user interaction module The user interaction module is used to receive task requests and query requests submitted by users, and provide task information to users according to the query requests.

9. The system according to claim 8, characterized in that The user interaction module includes: a request parsing unit and a request parsing unit The request parsing unit is used to parse the task request, extract the task type, task parameters, and priority information and store them in the information library; The request parsing unit is used to extract relevant task information from the information base according to the query request.

10. An adaptive task scheduling method, characterized in that: The method comprises: Obtaining task information, system resource information and user behavior data from the user side and the cloud server at regular intervals, and storing the task information, system resource information and user behavior data in an information library; Extracting the task information, the system resource information and the user behavior data from the information base, and evaluating the importance and resource requirements of each task based on the task information, the system resource information and the user behavior data to generate a task value score; Generate a preliminary task execution plan based on the task value score combined with preset scheduling rules, and optimize the preliminary task execution plan through a reinforcement learning algorithm to obtain a dynamic optimal scheduling decision; The task redundancy analysis module is used to mine the redundant tasks that can be merged from the task information, and generate a redundancy report based on the redundant tasks and feed it back to the scheduling strategy generation module; Adjusting the dynamic optimal scheduling decision according to the redundant report to obtain a target scheduling decision; Acquire the execution progress information of each task in the information base and the load information of each part of the system in real time, and dynamically adjust the task priority and resource allocation scheme according to the execution progress information and the load information to obtain the adjustment result; According to the target scheduling decision and the adjustment result, the cloud server is coordinated to execute tasks and allocate resources, and the execution status and resource usage of the tasks are collected to generate an execution status report.

Images