Task scheduling optimization method and system based on decision tree algorithm

Abstract

The application provides a task scheduling optimization method and system based on a decision tree algorithm, and relates to the technical field of computer task scheduling. First, the growth basis of the task scheduling decision tree is initialized, the multidimensional growth direction is determined, and the initial growth node is set. Then, the node expansion is performed based on the multidimensional direction, the data is extracted from different systems to generate the subnode, and the exclusive scheduling judgment rule is constructed. Then, the association link between the real-time scheduling feedback and the decision tree is established. The decision tree structure is iteratively optimized based on the feedback data. The attribute data of the task to be scheduled is input into the optimized decision tree to generate a multidimensional adaptive scheduling scheme. Finally, the task scheduling execution instruction is generated based on the scheme and is sent to the task execution control system. The application can comprehensively consider multidimensional factors, and improve the task scheduling efficiency and reliability.

Description

Technical Field

[0001] This invention relates to the field of computer task scheduling technology, and more specifically, to a task scheduling optimization method and system based on decision tree algorithm. Background Technology

[0002] In the field of computer task scheduling, with the continuous expansion of system scale and the increasing complexity of tasks, how to perform task scheduling efficiently and rationally has become a critical issue that urgently needs to be addressed. Traditional task scheduling methods often make decisions based on simple rules or experience, lacking comprehensive consideration of factors such as tasks, resources, and environment.

[0003] For example, some common scheduling methods focus only on task priority or resource availability, ignoring the dependencies between tasks. When multiple interdependent tasks exist, the above simplistic scheduling methods may lead to an unreasonable task execution order, thus affecting the overall system efficiency. Furthermore, resource constraints are also a problem that traditional scheduling methods struggle to handle effectively. In a multi-task concurrent execution environment, unreasonable resource allocation can easily lead to resource waste or contention, reducing the overall system performance.

[0004] Meanwhile, environmental interference factors, such as network latency and hardware failures, also play a significant role in task scheduling. Traditional scheduling methods often fail to adequately consider these environmental interferences, making tasks susceptible to external factors during execution, leading to task failure or inefficiency. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a task scheduling optimization method based on a decision tree algorithm, the method comprising: The growth basis of the initial task scheduling decision tree is initialized, the multi-dimensional growth direction of the task scheduling decision tree is determined, and an initial growth node is set for each growth dimension. The initial growth node carries the basic scheduling attribute information of the growth dimension. The multi-dimensional growth direction includes task dependency growth dimension, resource constraint growth dimension, and environmental interference growth dimension. Based on the multi-dimensional growth direction, the task scheduling decision tree node expansion is performed. Task-related data is extracted from the task management system, resource status data is extracted from the resource monitoring system, and environmental parameter data is extracted from the environmental monitoring system. Sub-nodes are generated for each growth dimension, and dimension-specific scheduling judgment rules are constructed for each sub-node. Establish a link between real-time scheduling feedback and the task scheduling decision tree, collect scheduling execution feedback data from the task execution control system, and construct association matching rules between the feedback data and the sub-nodes of each growth dimension. The scheduling execution feedback data includes task execution progress data, resource utilization efficiency data, and environmental interference impact data. Based on the scheduling execution feedback data transmitted through the associated link, the task scheduling decision tree structure is iteratively optimized, the scheduling judgment rule parameters of each growth dimension sub-node are adjusted, the path association relationship between sub-nodes is optimized, and an iteratively optimized task scheduling decision tree is generated. The task attribute data of the task to be scheduled is input into the optimized task scheduling decision tree. The scheduling judgment rules are executed layer by layer through the child nodes of each growth dimension to generate a multi-dimensional adaptive scheduling scheme for the task to be scheduled. The multi-dimensional adaptive scheduling scheme includes task execution order planning, resource allocation scheme, and environmental interference response scheme. Based on the multi-dimensional adaptation scheduling scheme, a task scheduling execution instruction is generated and sent to the task execution management system. The task scheduling execution instruction carries a task execution order identifier, a resource allocation identifier, and interference response trigger conditions.

[0006] Furthermore, embodiments of the present invention also provide a task scheduling optimization system based on a decision tree algorithm, characterized in that it includes: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described task scheduling optimization method based on the decision tree algorithm by executing the machine-executable instructions.

[0007] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, the processor of a task scheduling optimization system based on a decision tree algorithm reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the task scheduling optimization system based on a decision tree algorithm to execute the aforementioned task scheduling optimization method based on a decision tree algorithm.

[0008] Based on the above, by initializing the growth foundation of the task scheduling decision tree and determining multi-dimensional growth directions, task scheduling can be analyzed from multiple key dimensions such as task dependency, resource constraints, and environmental interference. Then, node expansion is performed based on the multi-dimensional growth directions, and relevant data is extracted from different systems to generate child nodes. A dedicated scheduling judgment rule is constructed for each child node, enabling precise judgment based on specific circumstances of different dimensions, thus improving the flexibility and adaptability of scheduling. Establishing a link between real-time scheduling feedback and the task scheduling decision tree allows for timely acquisition of various feedback data during task execution. Based on this feedback data, the decision tree structure is iteratively optimized, adjusting scheduling judgment rule parameters and path relationships between child nodes, enabling the decision tree to continuously adapt to changes in the actual operating environment and maintain optimal scheduling performance. The final multi-dimensional adaptive scheduling scheme for the tasks to be scheduled comprehensively considers multiple aspects such as task execution order, resource allocation, and environmental interference response. The task scheduling execution instructions generated based on this multi-dimensional adaptive scheduling scheme accurately convey scheduling intent, ensuring that tasks are executed smoothly according to the optimal solution, effectively improving the efficiency and reliability of task scheduling. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the execution flow of the task scheduling optimization method based on the decision tree algorithm provided in the embodiment of the present invention.

[0010] Figure 2 This is a schematic diagram of exemplary hardware and software components of a task scheduling optimization system based on a decision tree algorithm provided in an embodiment of the present invention. Detailed Implementation

[0011] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a task scheduling optimization method based on a decision tree algorithm according to an embodiment of the present invention. The following is a detailed description of the task scheduling optimization method based on the decision tree algorithm.

[0012] Step S110: Initialize the growth basis of the task scheduling decision tree, determine the multi-dimensional growth direction of the task scheduling decision tree, set an initial growth node for each growth dimension, the initial growth node carries the basic scheduling attribute information of the growth dimension, and the multi-dimensional growth direction includes task dependency growth dimension, resource constraint growth dimension, and environmental interference growth dimension.

[0013] This embodiment uses a distributed data processing task scheduling scenario of an internet company as an example. First, the growth foundation of the task scheduling decision tree needs to be initialized. This initialization process includes determining multi-dimensional growth directions. In this scenario, the multi-dimensional growth directions are specifically set as task dependency growth dimensions, resource constraint growth dimensions, and environmental interference growth dimensions. For each growth dimension, a corresponding initial growth node must be set. These initial growth nodes carry the basic scheduling attribute information for each growth dimension. For example, the initial growth node for the task dependency growth dimension carries basic scheduling attribute information including the type of dependency relationship between tasks and the range of dependency strength; the initial growth node for the resource constraint growth dimension carries basic scheduling attribute information including resource type and the range of total resource quantity; and the initial growth node for the environmental interference growth dimension carries basic scheduling attribute information including environmental parameter type and the range of environmental parameter influence weights.

[0014] Step S120: Based on the multi-dimensional growth direction, perform the task scheduling decision tree node expansion, extract task-related data from the task management system, extract resource status data from the resource monitoring system, and extract environmental parameter data from the environmental monitoring system, generate sub-nodes corresponding to each growth dimension, and construct dimension-specific scheduling judgment rules for each sub-node.

[0015] In the distributed data processing task scheduling scenario of the aforementioned internet company, after initializing the task scheduling decision tree, node expansion is required based on the determined multi-dimensional growth direction. Specifically, relevant data needs to be extracted from different systems to support the generation of child nodes for each growth dimension. Task-related correlation data is extracted from the task management system, reflecting various relationships between tasks; resource status data is extracted from the resource monitoring system to understand the current resource usage and availability; and environmental parameter data is extracted from the environmental monitoring system to understand environmental factors that may affect task execution. Then, based on this extracted data, corresponding child nodes are generated for the task dependency growth dimension, resource constraint growth dimension, and environmental interference growth dimension, and a dedicated scheduling judgment rule is constructed for each child node. These rules will be used for subsequent task scheduling judgments.

[0016] Step S121: For the task dependency growth dimension, extract the task association data from the task management system. The task association data includes task pre-dependency relationship data, task post-dependency relationship data, and task parallel constraint relationship data.

[0017] In this distributed data processing task scheduling scenario, when expanding nodes based on task dependency growth dimensions, the first step is to extract task association data from the task management system. This task association data specifically includes pre-task dependency data, post-task dependency data, and parallel constraint data. This data forms the basis for generating sub-nodes in the task dependency growth dimension. By analyzing and processing this data, the dependencies, associations, and parallel constraints between tasks can be clearly identified.

[0018] Step S1211: Determine the extraction range of the task prerequisite dependency relationship data, and retrieve all associated task information of the task group to which the task to be scheduled belongs from the task association database of the task management system.

[0019] In the above scenario, when extracting task dependency data, the first step is to determine the extraction scope. The task to be scheduled belongs to a specific task group; therefore, all related task information for that task group should be retrieved from the task association database of the task management system. For example, if the task to be scheduled is a data cleaning task, its task group may also include data acquisition tasks, data analysis tasks, etc. All information for these related tasks needs to be retrieved in order to filter out the dependency data related to the task to be scheduled.

[0020] Step S1212: Extract the task prerequisite dependency data from the associated task information. The task prerequisite dependency data includes the prerequisite task identifier, the prerequisite task completion status requirement, and the dependency strength data between the prerequisite task and the current task. The dependency strength data is generated based on the degree of influence of the prerequisite task on the execution of the current task.

[0021] From all associated task information retrieved from the task group, further extract task prerequisite dependency data. This data specifically includes prerequisite task identifiers, used to uniquely identify prerequisite tasks; prerequisite task completion status requirements, specifying the required completion status of the prerequisite task before the current task can be initiated; and dependency strength data between the prerequisite and current tasks, determined based on the impact of the prerequisite task on the execution of the current task. For example, for a data cleaning task, its prerequisite task might be a data acquisition task. The completion status requirement for a data acquisition task is that data acquisition is completed and data integrity meets a certain standard. The dependency strength data is generated based on the impact of the quality and integrity of data acquisition on the data cleaning task.

[0022] Step S1213: Determine the extraction range of the task post-association data, and retrieve the subsequent task information affected by the current task from the task association database based on the current task identifier.

[0023] Next, determine the scope for extracting task-related relationship data. Based on the identifier of the currently scheduled task, retrieve information on all subsequent tasks affected by this current task from the task association database. For example, if the current task is a data cleaning task, then the subsequent tasks affected by it might be data analysis tasks, requiring the retrieval of relevant information about data analysis tasks from the database as the basis for extracting task-related relationship data.

[0024] Step S1214: Extract the task post-relationship data from the subsequent task information. The task post-relationship data includes the post-task identifier, the post-task start condition requirements, and the correlation strength data between the current task and the post-task. The correlation strength data is generated based on the degree of influence of the current task on the execution of the post-task.

[0025] Extract task-related relationship data from the retrieved subsequent task information. This data includes subsequent task identifiers for identifying subsequent tasks; subsequent task activation conditions, i.e., what conditions the current task must meet to activate the subsequent task; and correlation strength data between the current task and the subsequent task, which is generated based on the degree of influence of the current task on the execution of the subsequent task. For example, if the subsequent task of a data cleaning task is a data analysis task, the activation conditions for the subsequent task might be that the data cleaning task is completed and the data cleaning quality meets the standards, and the correlation strength data is determined based on the degree of influence of the data cleaning results on the accuracy of the data analysis.

[0026] Step S1215: Determine the extraction range of the task parallel constraint relationship data, and retrieve information on parallel tasks that may overlap with the current task in execution time from the task association database.

[0027] Next, determine the scope for extracting task parallel constraint relationship data. Retrieve information on parallel tasks that may overlap in execution time with the currently scheduled task from the task association database. For example, if the current task is a data cleaning task, a data backup task may also be running during its execution, and the execution times of these two tasks may overlap, then it is necessary to retrieve the relevant information of the data backup task.

[0028] Step S1216: Extract the parallel task constraint relationship data from the parallel task information. The parallel task constraint relationship data includes the parallel task identifier, the parallel task resource conflict type, and the parallel task execution time conflict threshold. The resource conflict type is determined based on the overlap type of resources required by the parallel task and the current task.

[0029] The parallel task constraint relationship data is extracted from the retrieved parallel task information. This data includes parallel task identifiers to distinguish different parallel tasks; parallel task resource conflict types, which are determined based on the overlap of resource requirements between the parallel task and the current task, such as both needing to occupy the same server's CPU or memory resources; and parallel task execution time conflict thresholds, i.e., the degree to which the execution times of two tasks overlap before a conflict occurs. Taking data cleaning and data backup tasks as examples, if both require a large amount of storage resources, the resource conflict type is storage resource conflict, and the execution time conflict threshold is set based on the impact of both simultaneously occupying storage resources on system performance.

[0030] Step S1217: Classify and store the extracted task pre-dependency relationship data, task post-association relationship data, and task parallel constraint relationship data, and set a unique data identifier for each type of data.

[0031] The extracted task prerequisite dependency data, task post-relationship data, and task parallel constraint data are categorized and stored for later use and management. A unique data identifier is assigned to each data type, allowing for quick and accurate identification and retrieval of the corresponding data. For example, a "Prerequisite Dependency - Data Identifier" is assigned to task prerequisite dependency data, a "Post-Relationship - Data Identifier" is assigned to task post-relationship data, and a "Parallel Constraint - Data Identifier" is assigned to task parallel constraint data, etc.

[0032] Step S122: Generate first-level child nodes of the task dependency growth dimension based on the task prerequisite dependency relationship data. The first-level child nodes correspond to different prerequisite dependency types. Construct prerequisite dependency matching rules for the first-level child nodes. The prerequisite dependency matching rules are used to determine whether the task to be scheduled meets the execution conditions of the corresponding prerequisite dependency type.

[0033] After extracting the task association data for the task dependency growth dimension, first-level child nodes for this growth dimension are generated based on the task's prerequisite dependency data. These first-level child nodes correspond to different prerequisite dependency types, such as strong dependency, weak dependency, and conditional dependency. For each first-level child node, a prerequisite dependency matching rule needs to be constructed. This rule is used to determine whether the task to be scheduled meets the execution conditions of the corresponding prerequisite dependency type. For example, the prerequisite dependency matching rule for a first-level child node of a strong dependency type might stipulate that the task to be scheduled can only meet the execution conditions when the prerequisite task is fully completed and passes the quality check; while the rule for a weak dependency type might be that the task to be scheduled can be started as soon as most of the prerequisite tasks are completed.

[0034] Step S123: Generate second-level child nodes of the task dependency growth dimension based on the task post-association relationship data. The second-level child nodes are associated with the first-level child nodes. Each first-level child node corresponds to multiple second-level child nodes. The second-level child nodes correspond to different post-association types. Construct post-association matching rules for the second-level child nodes. The post-association matching rules are used to determine the type of post-task that can be triggered after the scheduled task satisfies the preceding dependency.

[0035] After generating the first-level child nodes of the task dependency growth dimension, second-level child nodes are generated based on the task post-association data. Second-level child nodes are associated with first-level child nodes, and each first-level child node can correspond to multiple different second-level child nodes. These second-level child nodes correspond to different post-association types, such as direct triggering, delayed triggering, and conditional triggering. A post-association matching rule is constructed for each second-level child node. This rule determines the type of post-task that can be triggered after the scheduled task meets the preceding dependency conditions. For example, for a strong dependency type in the first-level child nodes, the post-association matching rule for the direct triggering type in the corresponding second-level child nodes might be: if the preceding task meets the strong dependency condition and the scheduled task is completed, the post-task is triggered immediately; while the rule for the delayed triggering type might be: the post-task is triggered some time after the scheduled task is completed.

[0036] Step S124: For the resource constraint growth dimension, extract the resource status data from the resource monitoring system. The resource status data includes the available resource quantity data, resource usage duration data, and resource load limit data.

[0037] In this distributed data processing task scheduling scenario, when expanding nodes based on resource constraint growth dimensions, it is necessary to extract resource status data from the resource monitoring system. Specifically, resource status data includes available resource quantity data, resource usage duration data, and resource load limit data. This data reflects the current status of various resources in the system and is the key basis for generating sub-nodes for the resource constraint growth dimension. By analyzing this data, the rules for resource allocation and usage can be determined.

[0038] Step S1241: Determine the target for extracting the available resource quantity data, and retrieve the type identifier and corresponding quantity information of all currently available resources from the resource ledger database of the resource monitoring system.

[0039] When extracting data on available resource quantities, the first step is to determine the target. Retrieve the type identifiers and corresponding quantity information of all currently available resources from the resource ledger database of the resource monitoring system. Resource type identifiers distinguish different types of resources, such as servers, CPUs, memory, and storage devices, while the corresponding quantity information indicates the current available quantity of each type of resource. For example, retrieving server resource type identifiers would be "Server-Type-A," and the corresponding quantity information would be the number of servers of that type currently in an idle state.

[0040] Step S1242: Classify the available resources according to resource type, count the current available quantity of each resource type, and generate the available quantity data of the resources. The available quantity data of the resources includes resource type identifier, current available quantity value, and resource quantity statistics timestamp.

[0041] All available resources are categorized by resource type, and the current available quantity for each resource type is then calculated to generate resource availability data. This data includes a resource type identifier to identify the specific type of resource; a current available quantity value, indicating the quantity of that type of resource currently available; and a timestamp recording the exact time the available quantity was calculated to ensure data timeliness. For example, server resources, CPU resources, and memory resources are categorized separately, and the current available quantity value for "Server-Type-A" is calculated, along with the corresponding timestamp.

[0042] Step S1243: Determine the target for extracting the resource usage duration data, and retrieve the historical usage records of each resource type from the resource usage log database of the resource monitoring system.

[0043] The data to be extracted for resource usage duration is determined by retrieving historical usage records for each resource type from the resource usage log database of the resource monitoring system. These historical usage records contain information about past resource usage; analysis of these records reveals usage patterns and duration distribution. For example, the usage records for the "Server-Type-A" server resource over a past period can be retrieved, including the start and end times of each usage session.

[0044] Step S1244: Extract the single continuous usage duration, daily cumulative usage duration, and weekly cumulative usage duration data for each resource type from the historical usage records to generate the resource usage duration data. The resource usage duration data includes the resource type identifier, the single usage duration range, the daily cumulative usage duration range, and the weekly cumulative usage duration range.

[0045] From the historical usage records of each resource type, extract the single continuous usage duration, daily cumulative usage duration, and weekly cumulative usage duration data. Single continuous usage duration refers to the length of time a resource is used continuously in one session; daily cumulative usage duration is the total duration the resource is used within a day; and weekly cumulative usage duration is the total duration the resource is used within a week. Based on this data, generate resource usage duration data, which includes the resource type identifier, as well as the ranges for single usage duration, daily cumulative usage duration, and weekly cumulative usage duration. For example, the single usage duration range for a "Server-Type-A" server resource may be within a certain range, and the daily and weekly cumulative usage duration ranges also each have corresponding ranges.

[0046] Step S1245: Determine the target for extracting the resource load limit data, and retrieve the hardware performance parameters and software operating parameters for each resource type from the resource performance parameter database of the resource monitoring system.

[0047] The data to be extracted for determining the upper limit of resource load is retrieved from the resource performance parameter database of the resource monitoring system, including hardware performance parameters and software operating parameters for each resource type. Hardware performance parameters include processor performance metrics, memory capacity metrics, and storage capacity metrics; software operating parameters include resource utilization ratio metrics. These parameters are crucial for determining the upper limit of resource load. For example, for server resources, it is necessary to retrieve their CPU processing power, memory size, storage capacity, and other hardware performance parameters, as well as the resource utilization ratio parameters of the operating system and related software during runtime.

[0048] Step S1246: Based on the processor performance index, memory capacity index, and storage capacity index in the hardware performance parameters, and combined with the resource usage ratio index in the software operation parameters, generate the maximum load capacity for each resource type, and generate the resource load upper limit data. The resource load upper limit data includes the resource type identifier, processor load upper limit value, memory load upper limit value, and storage load upper limit value.

[0049] Based on the retrieved hardware performance parameters (processor performance, memory capacity, and storage capacity) and the resource utilization rate from the software operating parameters, a comprehensive analysis is performed to calculate the maximum load capacity for each resource type, thus generating resource load ceiling data. This resource load ceiling data includes the resource type identifier, as well as the processor load ceiling, memory load ceiling, and storage load ceiling. For example, for server resources, based on their CPU performance, memory capacity, and storage capacity, and the resource utilization rate during software operation, the maximum load values ​​that the server can withstand in terms of processor, memory, and storage are calculated; these are the processor load ceiling, memory load ceiling, and storage load ceiling.

[0050] Step S1247: The extracted data on the number of available resources, the duration of resource usage, and the upper limit of resource load are associated and stored, and a correspondence between the resource type identifier and each data item is established.

[0051] The extracted data on available resource quantity, usage duration, and load limit are stored in a correlated manner, with each data item linked by a resource type identifier. This way, when querying the status of a resource, the available quantity, usage duration, and load limit can be quickly retrieved using the resource type identifier. For example, the available quantity, usage duration, and load limit data for a "Server-Type-A" server resource can be linked using this type identifier and stored in the corresponding location in the database.

[0052] Step S125: Generate a first-level child node of the resource constraint growth dimension based on the resource availability quantity data. The first-level child node corresponds to different resource availability quantity ranges. Construct resource quantity matching rules for the first-level child node. The resource quantity matching rules are used to determine whether the resource requirements of the task to be scheduled meet the corresponding availability quantity range.

[0053] After extracting the resource status data for the resource constraint growth dimension, first-level child nodes for this growth dimension are generated based on the available resource quantity data. These first-level child nodes correspond to different available resource quantity ranges, such as high availability, medium availability, and low availability. A resource quantity matching rule is constructed for each first-level child node. This rule is used to determine whether the resource requirements of the task to be scheduled conform to the corresponding available quantity range. For example, the resource quantity matching rule for the first-level child node in the high availability range might stipulate that the task's requirement conforms to the range if the quantity of a certain resource required by the task is less than or equal to the available quantity in that range. The rule for the low availability range might be that the task's required resource quantity is within the available quantity range of that range, but the resource scarcity level needs to be considered.

[0054] Step S126: Generate second-level child nodes of the resource constraint growth dimension based on the resource usage duration data. The second-level child nodes are associated with the first-level child nodes. Each first-level child node corresponds to multiple second-level child nodes. The second-level child nodes correspond to different resource usage duration intervals. Construct resource duration matching rules for the second-level child nodes. The resource duration matching rules are used to determine whether the execution duration requirement of the task to be scheduled meets the corresponding usage duration interval.

[0055] Second-level child nodes are generated based on resource usage duration data to create resource constraint growth dimensions. These second-level child nodes are associated with first-level child nodes, and each first-level child node corresponds to multiple different second-level child nodes. These second-level child nodes correspond to different resource usage duration intervals, such as short, medium, and long duration intervals. A resource duration matching rule is constructed for each second-level child node. This rule determines whether the execution duration requirement of the task to be scheduled conforms to the corresponding usage duration interval. For example, for the high availability interval in the first-level child node, the resource duration matching rule for the short duration interval in the corresponding second-level child node might be that the execution duration of the task to be scheduled is less than or equal to the upper limit of the short duration interval; while the rule for the medium duration interval might be that the task execution duration falls within the range of the medium duration interval.

[0056] Step S127: For the environmental interference growth dimension, extract the environmental parameter data from the environmental monitoring system. The environmental parameter data includes temperature change data, network stability data, and external interference frequency data.

[0057] In this distributed data processing task scheduling scenario, when expanding nodes based on the environmental interference growth dimension, it is necessary to extract environmental parameter data from the environmental monitoring system. Specifically, the environmental parameter data includes temperature change data, network stability data, and external interference frequency data. This data reflects the current state of the task execution environment and is crucial for assessing the impact of the environment on task execution and formulating corresponding countermeasures. It serves as the basis for generating sub-nodes for the environmental interference growth dimension.

[0058] Step S1271: Determine the extraction range of the temperature change data, and retrieve the real-time temperature data and historical temperature data of the task execution area from the temperature monitoring database of the environmental monitoring system.

[0059] When extracting temperature change data, the extraction range must first be determined. Real-time and historical temperature data for the task execution area are retrieved from the temperature monitoring database of the environmental monitoring system. The task execution area refers to the physical area where the distributed data processing task actually runs, such as a data center server room. Real-time temperature data reflects the current temperature situation in this area, while historical temperature data includes temperature records for this area over a past period. By comparing and analyzing the two, the temperature change trend can be understood.

[0060] Step S1272: Calculate the difference between the real-time temperature data and the historical average temperature data for the same period, statistically analyze the change range of the temperature difference within a unit time, and generate the temperature change data, which includes the temperature monitoring timestamp, real-time temperature value, historical average temperature value, and temperature difference change range.

[0061] The retrieved real-time and historical temperature data are processed to calculate the difference between the real-time temperature data and the historical average temperature data for the same period. The historical average temperature data refers to the average temperature value within the same historical time period. Then, the variation range of the temperature difference per unit time is statistically analyzed, such as the change in the temperature difference per hour. Through these calculations and statistics, temperature change data is generated. This temperature change data includes a temperature monitoring timestamp, recording the specific time of temperature monitoring; the real-time temperature value, the current actual temperature; the historical average temperature value, the average temperature for the same period in history; and the variation range of the temperature difference, the magnitude of the change in the temperature difference per unit time.

[0062] Step S1273: Determine the extraction range of the network stability data, and retrieve the real-time transmission data of the network link required for task execution from the network monitoring database of the environmental monitoring system.

[0063] The scope of network stability data extraction is determined, and real-time transmission data of the network links required for task execution is retrieved from the network monitoring database of the environmental monitoring system. The network links required for task execution refer to the network paths traversed by the distributed data processing task during data transmission, such as communication links between servers and connection links between the data center and external networks. Real-time transmission data includes information such as network transmission rate, network latency, and network packet loss rate.

[0064] Step S1274: Extract network transmission rate, network latency, and network packet loss rate data from the real-time transmission data, statistically analyze the fluctuation range of these data per unit time, and generate the network stability data. The network stability data includes network monitoring timestamps, transmission rate fluctuation range, latency fluctuation range, and packet loss rate fluctuation range.

[0065] Network transmission rate, network latency, and network packet loss rate data are extracted from the retrieved real-time network link transmission data. Then, the fluctuation range of these data within a unit of time is statistically analyzed, such as the difference between the maximum and minimum network transmission rate per minute, the fluctuation range of network latency, and the variation range of network packet loss rate. Based on these statistical results, network stability data is generated. This network stability data includes a network monitoring timestamp, recording the time of network monitoring; the transmission rate fluctuation range, the variation range of transmission rate within a unit of time; the latency fluctuation range, the fluctuation of latency within a unit of time; and the packet loss rate fluctuation range, the variation range of packet loss rate within a unit of time.

[0066] Step S1275: Determine the extraction range of the external interference frequency data, and retrieve the external interference event records within the task execution area from the interference monitoring database of the environmental monitoring system.

[0067] The extraction range of external interference frequency data is determined, and records of external interference events within the task execution area are retrieved from the interference monitoring database of the environmental monitoring system. External interference events include power fluctuations, electromagnetic interference, equipment failures, and other events that may affect task execution. These interference event records within the task execution area reflect the occurrence of external interference.

[0068] Step S1276: Count the number of occurrences of different types of external interference events per unit time, generate the occurrence frequency of each type of interference event, and generate the external interference frequency data. The external interference frequency data includes interference type identifier, number of occurrences per unit time, and interference duration range.

[0069] The system analyzes records of external and external interference events retrieved within the task execution area, statistically analyzing the frequency of different types of external interference events per unit time. Interference type identifiers are used to distinguish different types of external interference, such as power fluctuations, electromagnetic interference, and equipment failures. The occurrence frequency of each type of interference event is generated based on the frequency of occurrence, thus generating external interference frequency data. This external interference frequency data includes an interference type identifier, specifying the type of interference; the number of occurrences per unit time, indicating the number of times this type of interference occurs per unit time; and the interference duration range, indicating the duration range of each instance of this type of interference.

[0070] Step S1277: Perform timestamp alignment processing on the extracted temperature change data, network stability data, and external interference frequency data to ensure that the monitoring time of different types of environmental parameter data remains consistent.

[0071] The extracted temperature change data, network stability data, and external interference frequency data were timestamped. Since different types of environmental parameter data may have been monitored at different times, their monitoring times need to be aligned for easier subsequent analysis and use. For example, using the temperature monitoring timestamp of the temperature change data as a baseline, the network monitoring timestamp of the network stability data and the occurrence timestamps of corresponding interference events in the external interference frequency data were adjusted to the same or similar times, ensuring temporal comparability of different types of environmental parameter data.

[0072] Step S128: Generate a first-level sub-node of the environmental disturbance growth dimension based on the temperature change data. The first-level sub-node corresponds to different temperature change ranges. Construct a temperature matching rule for the first-level sub-node. The temperature matching rule is used to determine whether the temperature tolerance requirement of the task to be scheduled meets the corresponding temperature change range.

[0073] After extracting the environmental parameter data for the growth dimension of environmental disturbances, first-level child nodes for this growth dimension are generated based on temperature change data. These first-level child nodes correspond to different temperature change ranges, such as small change ranges, medium change ranges, and large change ranges. A temperature matching rule is constructed for each first-level child node. This rule is used to determine whether the temperature tolerance requirements of the task to be scheduled meet the corresponding temperature change range. For example, the temperature matching rule for a first-level child node in a small change range might stipulate that the temperature tolerance requirement of the task to be scheduled can withstand temperature changes within that range; while the rule for a large change range might have higher requirements for the task's temperature tolerance.

[0074] Step S129: Generate second-level child nodes for the environmental interference growth dimension based on the network stability data. The second-level child nodes are associated with the first-level child nodes. Each first-level child node corresponds to multiple second-level child nodes. The second-level child nodes correspond to different network stability levels. Construct network matching rules for the second-level child nodes. The network matching rules are used to determine whether the network requirements of the task to be scheduled meet the corresponding stability level.

[0075] Second-level child nodes are generated based on network stability data to represent the environmental interference growth dimension. These second-level child nodes are associated with first-level child nodes, and each first-level child node corresponds to multiple different second-level child nodes. These second-level child nodes correspond to different network stability levels, such as high stability, medium stability, and low stability. A network matching rule is constructed for each second-level child node to determine whether the network requirements of the task to be scheduled meet the corresponding stability level. For example, for a small temperature variation range in a first-level child node, the corresponding high-stability-level network matching rule in the second-level child node might state that the requirements of the task to be scheduled for network transmission rate, latency, and packet loss rate can be met in a high-stability network environment; while the low-stability-level rule might allow network parameters to fluctuate within a certain range.

[0076] Step S130: Establish the association link between real-time scheduling feedback and the task scheduling decision tree, collect scheduling execution feedback data from the task execution control system, and construct the association matching rules between the feedback data and the sub-nodes of each growth dimension. The scheduling execution feedback data includes task execution progress data, resource utilization efficiency data, and environmental interference impact data.

[0077] After expanding the nodes of the task scheduling decision tree, it is necessary to establish a link between real-time scheduling feedback and the task scheduling decision tree. This process includes collecting scheduling execution feedback data from the task execution control system. This data reflects the actual execution status of the task, specifically including task execution progress data, resource utilization efficiency data, and environmental interference impact data. Then, association matching rules are constructed between the feedback data and the child nodes of each growth dimension. These rules map the feedback data to the child nodes in the decision tree, enabling optimization of the decision tree based on the feedback data. In this distributed data processing task scheduling scenario, the task execution control system is responsible for monitoring the task execution process.

[0078] Step S131: Set the collection frequency of the scheduling execution feedback data, and obtain the task execution progress data from the task execution control system according to the collection frequency. The task execution progress data includes the actual start time of the task, the actual completion time of the task, and the time consumption data of each stage of task execution.

[0079] During the establishment of the connection link, the collection frequency of scheduling execution feedback data is first set. The collection frequency is determined based on the importance of the task, the execution cycle, and the requirement for timely feedback; for example, it can be set to collect data once per hour or once per half day. Task execution progress data is then obtained from the task execution control system according to the set collection frequency. Specifically, the task execution progress data includes the actual start time of the task (i.e., the time when the task truly starts); the actual completion time of the task (i.e., the time when the task is completed); and the time consumed in each stage of task execution, such as the time consumed in the data loading stage, the time consumed in the data processing stage, and the time consumed in the result output stage. This data can reflect the progress of task execution in detail.

[0080] Step S132: Obtain the resource utilization efficiency data from the task execution control system. The resource utilization efficiency data includes the actual number of resources allocated, the actual usage time of resources, and the idle time of resources.

[0081] Simultaneously, resource utilization efficiency data is obtained from the task execution control system. Actual resource allocation quantity refers to the quantity of various resources actually allocated to a task; actual resource usage time refers to the time these resources are actually occupied by the task; and resource idle time data refers to the time resources are allocated to a task but not used. This data allows for the evaluation of resource utilization efficiency and the determination of whether resource allocation is reasonable.

[0082] Step S133: Obtain the environmental interference impact data from the task execution control system. The environmental interference impact data includes the task execution delay caused by temperature changes, the task data transmission delay caused by network fluctuations, and the number of task interruptions caused by external interference.

[0083] In addition, it is necessary to obtain environmental interference impact data from the task execution control system. This environmental interference impact data specifically includes the task execution delay caused by temperature changes, that is, the time when the task execution time is extended due to temperature changes; the task data transmission delay caused by network fluctuations, that is, the time when data transmission time is increased due to network instability; and the number of task interruptions caused by external interference, that is, the number of times the task was forced to be interrupted by external interference events. The above data can reflect the degree of impact of environmental factors on task execution.

[0084] Step S134: For the child nodes of the task dependency growth dimension, construct the association matching rules between the task execution progress data and the child nodes. The association matching rules specify the deviation range of the actual start time of the task corresponding to different pre-dependency type child nodes and the deviation range of the actual completion time of the task corresponding to different post-association type child nodes.

[0085] For child nodes in the task dependency growth dimension, association and matching rules are constructed between task execution progress data and child nodes. For child nodes with different pre-dependency types, the rules specify the corresponding deviation range of the actual task start time, i.e., the allowable deviation range between the actual task start time and the planned start time. For child nodes with different post-dependency types, the rules specify the corresponding deviation range of the actual task completion time, i.e., the allowable deviation range between the actual task completion time and the planned completion time. For example, for child nodes with strong pre-dependencies, the deviation range of the actual task start time may be small, requiring the task to start as close to the planned time as possible; while for child nodes with weak pre-dependencies, the deviation range may be large.

[0086] Step S135: For the child nodes of the resource constraint growth dimension, construct the association matching rules between the resource utilization efficiency data and the child nodes. The association matching rules specify the deviation range of the actual resource allocation quantity corresponding to the child nodes in different resource availability quantity ranges, and the deviation range of the actual resource usage time corresponding to the child nodes in different resource usage time ranges.

[0087] For child nodes in the resource constraint growth dimension, association matching rules are constructed between resource utilization efficiency data and child nodes. Child nodes with different resource availability intervals correspond to different deviation ranges in the actual resource allocation quantity, i.e., the deviation range between the actual resource allocation quantity to the task and the quantity that should be allocated according to the rule for that child node. Child nodes with different resource usage duration intervals correspond to different deviation ranges in the actual resource usage duration, i.e., the deviation range between the actual usage duration of the resource and the expected usage duration according to the rule for that child node. For example, child nodes in the high resource availability interval may have a larger deviation range in the actual resource allocation quantity because resources are abundant, allowing for some allocation fluctuations; while child nodes in the low resource availability interval may have a smaller deviation range to ensure the rational utilization of resources.

[0088] Step S136: For the child nodes of the environmental interference growth dimension, construct the association matching rules between the environmental interference impact data and the child nodes. The association matching rules specify the temperature-induced delay time deviation range corresponding to the child nodes in different temperature change ranges, and the network-induced transmission delay time deviation range corresponding to the child nodes in different network stability levels.

[0089] For child nodes in the environmental interference growth dimension, association matching rules are constructed between environmental interference impact data and child nodes. Child nodes with different temperature change ranges correspond to different temperature-induced delay duration deviation ranges, i.e., the deviation range between the task execution delay duration caused by temperature changes and the delay duration predicted according to the rules for that child node. Child nodes with different network stability levels correspond to different network-induced transmission delay duration deviation ranges, i.e., the deviation range between the task data transmission delay duration caused by network fluctuations and the delay duration predicted according to the rules for that child node. For example, for child nodes in ranges of large temperature changes, the temperature-induced delay duration deviation range may be large because the impact of large temperature changes on the task is uncertain; while for child nodes in ranges of small temperature changes, the deviation range may be small.

[0090] Step S137: Establish a mapping relationship table between the association matching rules and the child nodes of each growth dimension, and store the mapping relationship table in the association link configuration module of the task scheduling decision tree. The mapping relationship table records the association matching rule identifier and feedback data type corresponding to each child node.

[0091] After constructing the association matching rules between each growth dimension's child nodes and feedback data, a mapping table is established between these rules and each growth dimension's child nodes. This mapping table records the association matching rule identifier for each child node, allowing the identification of the corresponding association matching rule; and the feedback data type, clarifying the type of feedback data associated with each child node. This mapping table is stored in the association link configuration module of the task scheduling decision tree, enabling rapid lookup of the corresponding association matching rule and feedback data type based on the child node during subsequent decision tree optimization.

[0092] Step S140: Based on the scheduling execution feedback data transmitted through the associated link, iteratively optimize the task scheduling decision tree structure, adjust the scheduling judgment rule parameters of each growth dimension sub-node, optimize the path association relationship between sub-nodes, and generate the iteratively optimized task scheduling decision tree.

[0093] After establishing the link between real-time scheduling feedback and the task scheduling decision tree, the task scheduling decision tree structure is iteratively optimized based on the scheduling execution feedback data transmitted through this link. This process includes adjusting the scheduling judgment rule parameters of each growth dimension's child nodes to make the rules more consistent with actual task execution; and optimizing the path relationships between child nodes to ensure more rational and efficient task scheduling paths. Through continuous iterative optimization, an optimized task scheduling decision tree is generated to improve the accuracy and efficiency of task scheduling. In this distributed data processing task scheduling scenario, as tasks are continuously executed and feedback data accumulates, the decision tree needs continuous optimization to adapt to changing circumstances.

[0094] Step S141: Receive the scheduling execution feedback data through the association link, and match the corresponding growth dimension sub-nodes and the association matching rules based on the mapping relationship table.

[0095] During the iterative optimization process, scheduling execution feedback data is first received via the associated link. This data is acquired from the task execution management system at a set collection frequency. Then, based on the previously established mapping table, the corresponding growth dimension sub-nodes and associated matching rules are matched. That is, based on the type and content of the feedback data, the mapping table is searched to determine which growth dimension and which sub-node the feedback data is associated with, and what the corresponding associated matching rule is. For example, if the received task execution progress data shows a significant deviation in the actual start time of the task, the mapping table can be used to match a sub-node of a certain pre-dependency type in the task dependency growth dimension and its corresponding associated matching rule.

[0096] Step S142: For the child nodes of the task dependency growth dimension, compare the deviation range between the task execution progress data and the association matching rules. If the actual deviation exceeds the deviation range, extract the parameters of the preceding dependency matching rule or the following association matching rule of the child node, and adjust the execution condition threshold in the parameters.

[0097] For child nodes in the task dependency growth dimension, the received task execution progress data is compared with the deviation range in the association matching rules. If the actual deviation exceeds the deviation range specified in the rules, it indicates that the scheduling judgment rule parameters of the current child node may be unreasonable and need to be adjusted. In this case, the parameters of the preceding dependency matching rules or the following association matching rules for this child node are extracted. These parameters include execution condition thresholds, etc. Then, the execution condition thresholds in the parameters are adjusted to adapt to the actual task execution situation. For example, in the association matching rules of a certain preceding dependency type child node, the deviation range at the actual start time of task execution is a certain value, but the deviation in the actual feedback data exceeds this range. Then, it is necessary to extract the execution condition thresholds in the preceding dependency matching rules of this child node, such as the strictness threshold of the preceding task completion status requirements, and adjust them, possibly by appropriately relaxing or tightening the threshold.

[0098] Step S1421: Extract the actual deviation data of the corresponding child node from the task execution progress data. If the child node is the first-level child node of the task dependency growth dimension and the first-level child node is a pre-dependency type child node, then the actual deviation data is the difference between the actual start time of the task and the planned start time.

[0099] When adjusting parameters for child nodes in the task dependency growth dimension, the actual deviation data for the corresponding child node is first extracted from the task execution progress data. If the child node is a first-level child node in the task dependency growth dimension and is a prerequisite dependency type child node, then the actual deviation data is the difference between the actual start time and the planned start time of the task. The planned start time is the expected start time of the task set according to the prerequisite dependency matching rules. By calculating the difference between the two, the deviation of the task start time can be understood.

[0100] Step S1422: Retrieve the deviation range in the association matching rule of the preceding dependency type sub-node, and compare the actual deviation data with the deviation range.

[0101] The algorithm retrieves the deviation range specified in the association matching rules for the child nodes of the preceding dependency type, and then compares the extracted actual deviation data with this deviation range. It determines whether the actual deviation data is within the deviation range; if it is outside the range, subsequent parameter adjustments are required.

[0102] Step S1423: If the actual deviation data exceeds the upper limit of the deviation range, then extract the execution condition threshold from the pre-dependency matching rule of the sub-node. The execution condition threshold includes the pre-task completion advance threshold and the pre-task result verification time threshold.

[0103] When the actual deviation data exceeds the upper limit of the deviation range, the execution condition threshold in the preceding dependency matching rule of that child node is extracted. The execution condition threshold specifically includes the preceding task completion advance threshold, that is, the amount of time that the preceding task needs to be completed in advance; and the preceding task result verification time threshold, that is, the time threshold required to verify the result of the preceding task. The above thresholds directly affect the judgment result of the preceding dependency matching rule.

[0104] Step S1424: Increase the value of the threshold for the completion advance of the preceding task to allow the preceding task more time to complete and reduce the delay in the start of the current task; increase the value of the threshold for the verification time of the result of the preceding task to extend the verification time of the result of the preceding task and make the result of the preceding task meet the execution requirements of the current task.

[0105] To address situations where the deviation exceeds the upper limit, increase the threshold value for the lead time of the preceding tasks. This allows the preceding tasks more time to complete, thereby reducing the delay in starting the current task. Simultaneously, increase the threshold value for the verification time of the preceding task results, extending the verification period to ensure that the results of the preceding tasks meet the requirements of the current task, reducing task delays caused by unqualified preceding task results.

[0106] Step S1425: If the child node is a second-level child node of the task-dependent growth dimension, and the second-level child node is a post-association type child node, then the actual deviation data is the difference between the actual completion time of the task and the planned completion time.

[0107] If the child node is a second-level child node in the task dependency growth dimension and is a subsequent association type child node, then the actual deviation data is the difference between the actual completion time and the planned completion time of the task. The planned completion time is the expected completion time of the task set according to the subsequent association matching rules. This difference can be used to understand the deviation of the task completion time.

[0108] Step S1426: Retrieve the deviation range in the association matching rules of the subsequent association type sub-node, and compare the actual deviation data with the deviation range.

[0109] Retrieve the deviation range from the association matching rules of the child node of the corresponding association type, compare the actual deviation data with the deviation range, and determine whether it exceeds the range.

[0110] Step S1427: If the actual deviation data exceeds the upper limit of the deviation range, then extract the execution condition threshold from the post-association matching rule of the child node. The execution condition threshold includes the post-task start delay threshold and the post-task resource reservation time threshold.

[0111] When the actual deviation data exceeds the upper limit of the deviation range of the child node of the subsequent association type, the execution condition threshold in the subsequent association matching rule of the child node is extracted. The above thresholds include the subsequent task start delay threshold, that is, the time to delay the start of the subsequent task after the current task is completed; and the subsequent task resource reservation time threshold, which is the time threshold for reserving resources for the subsequent task.

[0112] Step S1428: Increase the value of the post-task start delay threshold and the value of the post-task resource reservation time threshold.

[0113] To address situations where actual deviations exceed the upper limit, the threshold for delaying the start of subsequent tasks is increased to allow more time for the current task to finish. At the same time, the threshold for reserving resource duration for subsequent tasks is increased to ensure that subsequent tasks have sufficient resources available and reduce delays caused by resource issues.

[0114] Step S143: For the child nodes of the resource constraint growth dimension, compare the deviation range between the resource utilization efficiency data and the association matching rules. If the actual deviation exceeds the deviation range, extract the resource quantity matching rule or resource duration matching rule parameters of the child node, and adjust the quantity interval threshold or duration interval threshold in the parameters.

[0115] For child nodes in the resource constraint growth dimension, compare the resource utilization efficiency data with the deviation range in the association matching rules. If the actual deviation exceeds the deviation range, it indicates that the parameters of the resource quantity matching rule or resource duration matching rule need to be adjusted. Extract the parameters of the resource quantity matching rule or resource duration matching rule for that child node, such as the quantity range threshold or duration range threshold, and adjust them. For example, if the deviation range of the actual resource allocation quantity in the association matching rule for a child node with a high available resource quantity exceeds the specified limit, then it is necessary to extract the quantity range threshold in the resource quantity matching rule and adjust the upper and lower limits of this threshold to make the resource allocation more reasonable.

[0116] Step S144: For the child nodes of the environmental interference growth dimension, compare the deviation range of the environmental interference impact data with the correlation matching rules. If the actual deviation exceeds the deviation range, extract the temperature matching rule or network matching rule parameters of the child node and adjust the temperature range threshold or stability level threshold in the parameters.

[0117] For sub-nodes in the environmental interference growth dimension, the impact data of environmental interference is compared with the deviation range in the association matching rules. If the actual deviation exceeds the deviation range, the temperature matching rule or network matching rule parameters for that sub-node, such as the temperature range threshold or stability level threshold, are extracted and adjusted. For example, in the association matching rules for sub-nodes with large temperature changes, if temperature causes the delay time deviation to exceed the range, then the temperature range threshold in the temperature matching rule is extracted and adjusted to more accurately reflect the impact of temperature changes on the task.

[0118] Step S145: Count the frequency of deviation exceeding the child node in each growth dimension. For child nodes whose deviation exceeds the preset frequency threshold, increase the number of associated child nodes of that child node and expand the judgment dimension of the child node.

[0119] While adjusting the rule parameters of each child node, the deviation exceedance frequency of each growth dimension child node is statistically analyzed. Deviation exceedance frequency refers to the number of times within a certain time period that the actual deviation corresponding to a child node exceeds the deviation range in the associated matching rule. The preset frequency threshold is set based on the importance of the child node and its impact on the scheduling results. For a child node whose deviation exceedance frequency reaches the preset frequency threshold, it indicates that the judgment dimension of that child node may be insufficient, and it is necessary to increase the number of its associated child nodes to expand the judgment dimension of the child node. For example, if the deviation exceedance frequency of a child node in a certain resource usage duration interval is high, it may be because the interval division is not fine enough. In this case, the associated child nodes of that child node can be increased to further subdivide the original duration interval, thereby improving the accuracy of the judgment.

[0120] Step S146: Analyze the path association between child nodes and calculate the scheduling execution success rate corresponding to each path. The scheduling execution success rate is determined based on the comparison between the actual completion status and the planned completion status of the task under that path.

[0121] Besides adjusting the parameters and number of child nodes, it's also necessary to analyze the path relationships between them. Each path starts from the initial growth node and passes through a sequence of child nodes at different levels across various growth dimensions. The scheduling execution success rate for each path is calculated. This success rate is determined by comparing the actual task completion rate with the planned completion rate for that path, such as the ratio of the actual number of completed tasks to the planned number. Calculating the success rate allows us to evaluate the quality of each path.

[0122] Step S1461: Traverse the paths between all child nodes in the task scheduling decision tree. Each path contains a sequence of nodes starting from the initial growth node and passing through child nodes at different levels of each growth dimension.

[0123] When calculating the success rate of scheduling execution, the paths between all child nodes in the task scheduling decision tree are traversed first. Each path has its specific sequence of nodes, starting from the initial growth node and sequentially passing through child nodes at different levels of each growth dimension. For example, any path might start from the initial growth node of the task dependency growth dimension, pass through the prerequisite dependency type child nodes, then to the resource availability range child nodes of the resource constraint growth dimension, and finally to the temperature change range child nodes of the environmental disturbance growth dimension.

[0124] Step S1462: Set a unique path identifier for each path and record the identifiers of the child nodes contained in the path and the connection order between the child nodes.

[0125] Each traversed path is assigned a unique path identifier for differentiation and management. Simultaneously, the identifiers of the child nodes within the path and the connection order between them are recorded, providing insight into the path's structure and direction. For example, the path identifier in the above example is set to "Path-001," and the identifiers of its child nodes, such as "Task-Dep-Node-001," "Res-Avail-Node-002," and "Env-Temp-Node-003," and their connection order are recorded.

[0126] Step S1463: Filter out all task execution records corresponding to the path from the scheduling execution feedback data. The task execution records include task identifier, planned completion time, actual completion time, and task completion status.

[0127] Based on the path identifier and child node identifier, filter out all task execution records corresponding to the specified path from the scheduling execution feedback data. A task execution record is a record of the execution status of tasks that have been scheduled through this path, including information such as task identifier, planned completion time, actual completion time, and task completion status. For example, filter out the execution records of all tasks scheduled through the "Path-001" path.

[0128] Step S1464: Count the total number of tasks in the task execution records under this path, and the number of tasks with a completion status among the total number of tasks.

[0129] The task execution records under the selected path are statistically analyzed to calculate the total number of tasks, i.e., the total number of tasks that have been scheduled under this path; and the number of tasks in the total number of tasks with a completion status. The task completion status can be divided into completed status and incomplete status. Incomplete status may include task failure, task interruption, etc.

[0130] Step S1465: Based on the comparison between the number of completed tasks and the total number of tasks, determine the scheduling execution success rate of the path.

[0131] Based on the statistical comparison between the number of completed tasks and the total number of tasks, the ratio between the two is calculated, and this ratio is determined as the scheduling execution success rate of that path. For example, if the total number of tasks is N and the number of completed tasks is M, then the scheduling execution success rate is the ratio of M to N.

[0132] Step S1466: If the number of task execution records corresponding to any path is lower than the preset record number threshold, which is insufficient to generate an effective scheduling execution success rate, then retrieve the past task execution records of that path from the historical scheduling database and supplement them to the current task execution record.

[0133] If the number of task execution records for any path is low, below a preset record count threshold, the scheduling execution success rate calculated based on these records may be inaccurate and insufficient to reflect the true situation of the path. In this case, past task execution records for that path are retrieved from the historical scheduling database and added to the current task execution records to increase the sample size. The preset record count threshold is set based on the path's usage frequency and task stability.

[0134] Step S1467: Recalculate the total number of tasks and the number of completed tasks after the supplementation. Based on the comparison between the number of completed tasks after the supplementation and the total number of tasks, determine the success rate of the supplemented scheduling execution.

[0135] The total number of tasks and the number of completed tasks are recalculated after supplementing the task execution records. Then, based on the comparison between the supplemented number of completed tasks and the total number of tasks, the scheduling execution success rate of this path is recalculated and determined. By supplementing historical data, the accuracy and reliability of scheduling execution success rate calculation can be improved.

[0136] Step S1468: Organize the path identifier of each path, the scheduling execution success rate, and the number of task execution records used for calculation into a path success rate statistics table, and store the path success rate statistics table in the path optimization module of the task scheduling decision tree.

[0137] The path identifier for each path, the calculated scheduling execution success rate, and the number of task execution records used for the calculation are compiled into a path success rate statistics table. This table displays the performance of each path and is stored in the path optimization module of the task scheduling decision tree.

[0138] Step S147: For paths with a scheduling execution success rate lower than a preset success rate threshold, adjust the order of the child nodes associated with the path, or replace some child nodes in the path to optimize the path's association logic.

[0139] Based on the path success rate statistics table, identify paths with a scheduling execution success rate lower than a preset success rate threshold. The preset success rate threshold is set according to the task scheduling goals and requirements, such as a certain percentage. For these paths with low success rates, their association logic needs to be optimized. Optimization methods include adjusting the order of child nodes associated with the path, changing the order in which child nodes are judged; or replacing some child nodes in the path, replacing the child nodes with low success rates with other more suitable child nodes. For example, if the scheduling execution success rate of any path is lower than the preset threshold, and analysis reveals that a certain resource-constrained child node is causing unreasonable resource allocation, then that child node can be replaced with other resource-constrained child nodes to improve the path's success rate.

[0140] Step S148: Integrate the adjusted child node parameters, the expanded number of child nodes, and the optimized path association to generate the iterative optimized task scheduling decision tree, and store the structural information of the optimized task scheduling decision tree and the scheduling judgment rules of each child node.

[0141] After adjusting child node parameters, expanding the number of child nodes, and optimizing path relationships, the results of these adjustments and optimizations are integrated. The adjusted child node parameters are integrated to ensure the accuracy of the rule parameters for each child node; the expanded number of child nodes is integrated to update the node structure of the decision tree; and the optimized path relationships are integrated to make the paths more rational and efficient. This integration generates an iteratively optimized task scheduling decision tree, and the structure information of this optimized decision tree and the scheduling judgment rules for each child node are stored for use in subsequent task scheduling.

[0142] Step S150: Input the task attribute data of the task to be scheduled into the optimized task scheduling decision tree, and execute the scheduling judgment rules layer by layer through the child nodes of each growth dimension to generate a multi-dimensional adaptive scheduling scheme for the task to be scheduled. The multi-dimensional adaptive scheduling scheme includes task execution order planning, resource allocation scheme, and environmental interference response scheme.

[0143] After generating the optimized task scheduling decision tree, the task attribute data of the tasks to be scheduled is input into this decision tree. The tasks to be scheduled refer to the distributed data processing tasks that are about to be scheduled and executed; their task attribute data contains various characteristic information of the tasks. By executing scheduling judgment rules layer by layer through the child nodes of each growth dimension, the task attribute data is judged and filtered, ultimately generating a multi-dimensional adaptive scheduling scheme for the tasks to be scheduled. This multi-dimensional adaptive scheduling scheme comprehensively considers factors such as task dependencies, resource constraints, and environmental interference, including task execution order planning, resource allocation schemes, and environmental interference response schemes. In this scenario, the task to be scheduled may be a new data analysis task, requiring the formulation of a suitable scheduling scheme based on its attribute data.

[0144] Step S151: Extract the task attribute data of the task to be scheduled. The task attribute data includes the task identifier, the type and quantity of resources required for task execution, the task execution time requirement, the identifiers of the tasks that depend on the task before it, the identifiers of the tasks that are associated with the task after it, the task temperature tolerance range, and the task network requirement level.

[0145] When generating a multi-dimensional adaptive scheduling scheme, the task attribute data of the task to be scheduled is extracted first. The task identifier uniquely identifies the task; the resource types and quantities required for task execution specify which types of resources the task needs and the quantity of each type; the task execution duration requirement is the estimated execution time of the task; the task's pre-dependent task identifiers indicate the pre-tasks that the task depends on; the task's post-related task identifiers indicate the post-tasks affected by this task; the task's temperature tolerance range is the temperature range within which the task can execute normally; and the task's network requirement level reflects the task's requirements for network stability and transmission rate. For example, the task identifier for the task to be scheduled might be "Task-Analysis-001," the required resource types might include servers, CPU, and memory, the task execution duration requirement might be a specific time period, and the pre-dependent task identifier might be "Task-Clean-001," etc.

[0146] Step S152: Input the task pre-dependent task identifier and task post-related task identifier from the task attribute data into the task dependency growth dimension of the optimized task scheduling decision tree.

[0147] The extracted task attribute data, including the identifiers of pre-dependent tasks and post-dependent tasks, is input into the task dependency growth dimension of the optimized task scheduling decision tree. This growth dimension handles the dependencies between tasks; by inputting these identifiers, the decision tree can determine the task dependencies based on the child nodes of the growth dimension and the scheduling rules.

[0148] Step S153: Execute the prerequisite dependency matching rule through the first-level child node of the task dependency growth dimension, determine whether the prerequisite task corresponding to the task prerequisite dependency task identifier meets the completion status requirements, and determine the matching first-level child node.

[0149] The first-level child nodes in the task dependency growth dimension execute the prerequisite dependency matching rules. Based on the input prerequisite task identifier, they determine whether the corresponding prerequisite task meets the completion status requirements. Completion status requirements include whether the prerequisite task has been completed and whether the completion quality meets the standards. Through this determination, a first-level child node matching the prerequisite dependency status of the task to be scheduled is identified. For example, if the prerequisite task "Task-Clean-001" corresponding to the prerequisite task identifier meets the completion status requirements, then it will match the corresponding first-level child node of the prerequisite dependency type in the task dependency growth dimension.

[0150] Step S154: Based on the matched first-level child node, execute the post-association matching rule through the corresponding second-level child node, determine whether the post-association task corresponding to the task post-association task identifier meets the startup condition requirements, and determine the matched second-level child node.

[0151] Based on the matched first-level child node, the subsequent association matching rules are executed through the corresponding second-level child nodes. According to the task's subsequent association task identifier, it is determined whether the corresponding subsequent task meets the startup conditions. Startup conditions include whether the subsequent task's resources are ready and whether there are other conflicting tasks. Through this determination, the matching second-level child node is identified. For example, if the subsequent association matching rule of a certain subsequent association type child node in the second-level child nodes corresponding to the matched first-level child node determines that the subsequent task meets the startup conditions, then that second-level child node is identified as the matching node.

[0152] Step S155: Input the resource type and quantity required for task execution and the task execution duration requirement from the task attribute data into the resource constraint growth dimension of the optimized task scheduling decision tree.

[0153] Simultaneously, the resource types and quantities required for task execution, as well as the task execution duration requirements, from the task attribute data are input into the resource constraint growth dimension of the optimized task scheduling decision tree. This resource constraint growth dimension is used to make scheduling decisions based on resource availability; these data are crucial for resource scheduling.

[0154] Step S156: Execute the resource quantity matching rule through the first-level child node of the resource constraint growth dimension to determine whether the resource quantity required by the task conforms to the corresponding available resource quantity range, and determine the matching first-level child node.

[0155] The first-level child nodes of the resource constraint growth dimension execute resource quantity matching rules. Based on the type and quantity of resources required for the input task, they determine whether it falls within the corresponding available resource quantity range. The available resource quantity range is the basis for dividing the first-level child nodes of the resource constraint growth dimension. By comparing the resource quantity required by the task with the range, the matching first-level child node is determined. For example, if the number of server resources required by the task falls within the range of the higher available resource quantity range child node, then that first-level child node is matched.

[0156] Step S157: Based on the matched first-level child node, execute the resource duration matching rule through the corresponding second-level child node, determine whether the task execution duration requirement conforms to the corresponding resource usage duration range, and determine the matched second-level child node.

[0157] Based on the matched first-level child nodes of the resource constraint growth dimension, resource duration matching rules are executed through the corresponding second-level child nodes. According to the task execution duration requirement, it is determined whether it conforms to the resource usage duration range corresponding to that child node. The resource usage duration range is the basis for dividing the second-level child nodes. By comparing the task execution duration requirement with the range, the matching second-level child nodes are determined. For example, among the second-level child nodes corresponding to the matched first-level child node, if the rule of a certain resource usage duration range child node determines that the task execution duration requirement conforms to that range, then the matching second-level child node is determined.

[0158] Step S158: Input the task temperature tolerance range and task network requirement level from the task attribute data into the environmental disturbance growth dimension of the optimized task scheduling decision tree.

[0159] In addition, the task's temperature tolerance range and network requirement level from the task attribute data are input into the environmental disturbance growth dimension of the optimized task scheduling decision tree. The environmental disturbance growth dimension is used to consider the impact of environmental factors on task execution, and this data forms the basis for developing environmental disturbance response plans.

[0160] Step S159: Execute the temperature matching rule through the first-level child node of the environmental interference growth dimension to determine whether the task temperature tolerance range meets the corresponding temperature change range, and determine the matching first-level child node.

[0161] The first-level child nodes in the environmental disturbance growth dimension execute temperature matching rules, determining whether they conform to the corresponding temperature variation range based on the task's temperature tolerance range. The temperature variation range is the basis for dividing the first-level child nodes; by comparing the task's temperature tolerance range with the range, the matching first-level child node is determined. For example, if the task's temperature tolerance range can accommodate a small temperature variation range, then it matches the corresponding first-level child node in the environmental disturbance growth dimension.

[0162] Step S1510: Based on the matched first-level child node, execute the network matching rules through the corresponding second-level child node to determine whether the task network requirement level meets the corresponding network stability level, and determine the matched second-level child node.

[0163] Based on the matched first-level child nodes in the environmental interference growth dimension, network matching rules are executed through the corresponding second-level child nodes. According to the task's network requirement level, it is determined whether it matches the network stability level corresponding to that child node. The network stability level is the basis for dividing the second-level child nodes; by comparing the task's network requirement level with the level standard, the matching second-level child nodes are determined. For example, if the task's network requirement level is high, a second-level child node with a high network stability level will be matched.

[0164] Step S1511: Integrate the output results of the child nodes matched by each growth dimension to generate task execution sequence planning, resource allocation scheme, and environmental interference response scheme.

[0165] After identifying matching child nodes for each growth dimension, the outputs of these child nodes are integrated. The outputs of child nodes in the task-dependent growth dimension are used to generate task execution sequence planning, determining the order in which tasks are executed; the outputs of child nodes in the resource-constraint growth dimension are used to generate resource allocation schemes, specifying the type, quantity, and duration of resources allocated to tasks; and the outputs of child nodes in the environmental interference growth dimension are used to generate environmental interference response schemes, formulating measures to address environmental interference such as temperature changes and network fluctuations.

[0166] Step S1512: Combine the task execution sequence planning, the resource allocation scheme, and the environmental interference response scheme to form a multi-dimensional adaptive scheduling scheme for the task to be scheduled; set a scheme version identifier for the multi-dimensional adaptive scheduling scheme, and record the scheme generation time and the corresponding optimized task scheduling decision tree version.

[0167] The generated task execution sequence plan, resource allocation scheme, and environmental interference response scheme are combined to form a multi-dimensional adaptive scheduling scheme for the tasks to be scheduled. A scheme version identifier is set for this scheme to facilitate version management and traceability. Simultaneously, the scheme generation time and the corresponding optimized task scheduling decision tree version are recorded to clarify which version of the decision tree generated the scheme, facilitating subsequent analysis and optimization.

[0168] Step S160: Generate a task scheduling execution instruction based on the multi-dimensional adaptation scheduling scheme, and send the task scheduling execution instruction to the task execution management system. The task scheduling execution instruction carries a task execution order identifier, a resource allocation identifier, and interference response triggering conditions.

[0169] After generating a multi-dimensional adaptive scheduling scheme, task scheduling execution instructions are generated based on this scheme. These instructions are specific execution commands, carrying a task execution order identifier to indicate the execution order of tasks; a resource allocation identifier to identify the resources allocated to the tasks; and interference response trigger conditions, indicating the environmental interference conditions that will trigger corresponding countermeasures. The generated task scheduling execution instructions are sent to the task execution management system, which then executes the task scheduling according to the instructions. In this distributed data processing task scheduling scenario, after receiving the instructions, the task execution management system organizes the execution of tasks according to the requirements of the instructions.

[0170] In one exemplary embodiment, a task scheduling optimization system based on a decision tree algorithm is provided. This system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2 As shown, the task scheduling optimization system based on the decision tree algorithm includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements a task scheduling optimization method based on the decision tree algorithm. The display unit of this task scheduling optimization system based on decision tree algorithm is used to form a visually visible image, and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of this task scheduling optimization system based on decision tree algorithm can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad set on the shell of the task scheduling optimization system based on decision tree algorithm algorithm, or an external keyboard, touchpad, or mouse, etc.

[0171] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A task scheduling optimization method based on decision tree algorithm, characterized in that, The method includes: The growth basis of the initial task scheduling decision tree is initialized, the multi-dimensional growth direction of the task scheduling decision tree is determined, and an initial growth node is set for each growth dimension. The initial growth node carries the basic scheduling attribute information of the growth dimension. The multi-dimensional growth direction includes task dependency growth dimension, resource constraint growth dimension, and environmental interference growth dimension. Based on the multi-dimensional growth direction, the task scheduling decision tree node expansion is performed. Task-related data is extracted from the task management system, resource status data is extracted from the resource monitoring system, and environmental parameter data is extracted from the environmental monitoring system. Sub-nodes are generated for each growth dimension, and dimension-specific scheduling judgment rules are constructed for each sub-node. Establish a link between real-time scheduling feedback and the task scheduling decision tree, collect scheduling execution feedback data from the task execution control system, and construct association matching rules between the feedback data and the sub-nodes of each growth dimension. The scheduling execution feedback data includes task execution progress data, resource utilization efficiency data, and environmental interference impact data. Based on the scheduling execution feedback data transmitted through the associated link, the task scheduling decision tree structure is iteratively optimized, the scheduling judgment rule parameters of each growth dimension sub-node are adjusted, the path association relationship between sub-nodes is optimized, and an iteratively optimized task scheduling decision tree is generated. The task attribute data of the task to be scheduled is input into the optimized task scheduling decision tree. The scheduling judgment rules are executed layer by layer through the child nodes of each growth dimension to generate a multi-dimensional adaptive scheduling scheme for the task to be scheduled. The multi-dimensional adaptive scheduling scheme includes task execution order planning, resource allocation scheme, and environmental interference response scheme. Based on the multi-dimensional adaptation scheduling scheme, a task scheduling execution instruction is generated and sent to the task execution management system. The task scheduling execution instruction carries a task execution order identifier, a resource allocation identifier, and interference response trigger conditions.

2. The task scheduling optimization method based on decision tree algorithm according to claim 1, characterized in that, The process of expanding the task scheduling decision tree nodes based on the multi-dimensional growth direction involves extracting task-related data from the task management system, resource status data from the resource monitoring system, and environmental parameter data from the environmental monitoring system. Sub-nodes are generated corresponding to each growth dimension, and dimension-specific scheduling judgment rules are constructed for each sub-node, including: For the task dependency growth dimension, the task association data is extracted from the task management system. The task association data includes task pre-dependency relationship data, task post-dependency relationship data, and task parallel constraint relationship data. Based on the task prerequisite dependency relationship data, a first-level child node of the task dependency growth dimension is generated. The first-level child node corresponds to different prerequisite dependency types. A prerequisite dependency matching rule is constructed for the first-level child node. The prerequisite dependency matching rule is used to determine whether the task to be scheduled meets the execution conditions of the corresponding prerequisite dependency type. Based on the task post-association relationship data, a second-level child node of the task dependency growth dimension is generated. The second-level child node is associated with the first-level child node. Each first-level child node corresponds to multiple second-level child nodes. The second-level child node corresponds to different post-association types. A post-association matching rule is constructed for the second-level child node. The post-association matching rule is used to determine the type of post-task that can be triggered after the task to be scheduled satisfies the preceding dependency. For the aforementioned resource constraint growth dimension, the resource status data is extracted from the resource monitoring system. The resource status data includes data on the number of available resources, resource usage duration, and resource load limit. Based on the available resource quantity data, a first-level child node of the resource constraint growth dimension is generated. The first-level child node corresponds to different available resource quantity ranges. A resource quantity matching rule is constructed for the first-level child node. The resource quantity matching rule is used to determine whether the resource requirements of the task to be scheduled meet the corresponding available quantity range. Based on the resource usage duration data, a second-level child node of the resource constraint growth dimension is generated. The second-level child node is associated with the first-level child node. Each first-level child node corresponds to multiple second-level child nodes. The second-level child node corresponds to different resource usage duration intervals. A resource duration matching rule is constructed for the second-level child node. The resource duration matching rule is used to determine whether the execution duration requirement of the task to be scheduled meets the corresponding usage duration interval. For the aforementioned environmental interference growth dimension, environmental parameter data is extracted from the environmental monitoring system. The environmental parameter data includes temperature change data, network stability data, and external interference frequency data. Based on the temperature change data, a first-level sub-node of the environmental disturbance growth dimension is generated. The first-level sub-node corresponds to different temperature change ranges. A temperature matching rule is constructed for the first-level sub-node. The temperature matching rule is used to determine whether the temperature tolerance requirement of the task to be scheduled meets the corresponding temperature change range. Based on the network stability data, a second-level child node of the environmental interference growth dimension is generated. The second-level child node is associated with the first-level child node. Each first-level child node corresponds to multiple second-level child nodes. The second-level child nodes correspond to different network stability levels. A network matching rule is constructed for the second-level child node. The network matching rule is used to determine whether the network requirements of the task to be scheduled meet the corresponding stability level.

3. The task scheduling optimization method based on decision tree algorithm according to claim 1, characterized in that, The process of establishing the link between real-time scheduling feedback and the task scheduling decision tree involves collecting scheduling execution feedback data from the task execution control system and constructing association matching rules between the feedback data and the sub-nodes of each growth dimension, including: Set the collection frequency of the scheduling execution feedback data, and obtain the task execution progress data from the task execution control system according to the collection frequency. The task execution progress data includes the actual start time of the task, the actual completion time of the task, and the time consumption data of each stage of task execution. The resource utilization efficiency data is obtained from the task execution control system. The resource utilization efficiency data includes the actual number of resources allocated, the actual duration of resource use, and the idle time of resources. The environmental interference impact data is obtained from the task execution control system. The environmental interference impact data includes the task execution delay caused by temperature changes, the task data transmission delay caused by network fluctuations, and the number of task interruptions caused by external interference. For the child nodes of the task-dependent growth dimension, an association matching rule is constructed between the task execution progress data and the child nodes. The association matching rule specifies the deviation range of the actual start time of the task corresponding to different pre-dependent child nodes and the deviation range of the actual completion time of the task corresponding to different post-dependent child nodes. For the child nodes of the resource constraint growth dimension, an association matching rule is constructed between the resource utilization efficiency data and the child nodes. The association matching rule specifies the deviation range of the actual resource allocation quantity corresponding to the child nodes in different resource availability quantity ranges, and the deviation range of the actual resource usage time corresponding to the child nodes in different resource usage time ranges. For the child nodes of the environmental interference growth dimension, an association matching rule is constructed between the environmental interference impact data and the child nodes. The association matching rule specifies the temperature-induced delay deviation range corresponding to the child nodes in different temperature change ranges, and the network-induced transmission delay deviation range corresponding to the child nodes in different network stability levels. Establish a mapping relationship table between the association matching rules and the child nodes of each growth dimension, and store the mapping relationship table in the association link configuration module of the task scheduling decision tree. The mapping relationship table records the association matching rule identifier and feedback data type corresponding to each child node.

4. The task scheduling optimization method based on decision tree algorithm according to claim 3, characterized in that, The iterative optimization of the task scheduling decision tree structure based on the scheduling execution feedback data transmitted through the associated link, adjusting the scheduling judgment rule parameters of each growth dimension child node, optimizing the path association relationship between child nodes, and generating the iteratively optimized task scheduling decision tree includes: The scheduling execution feedback data is received through the associated link, and the corresponding growth dimension sub-nodes and the associated matching rules are matched based on the mapping relationship table. For the child nodes of the task dependency growth dimension, compare the deviation range between the task execution progress data and the association matching rule. If the actual deviation exceeds the deviation range, extract the parameters of the preceding dependency matching rule or the following association matching rule of the child node and adjust the execution condition threshold in the parameters. For the child nodes of the resource constraint growth dimension, compare the deviation range between the resource utilization efficiency data and the association matching rules. If the actual deviation exceeds the deviation range, extract the resource quantity matching rule or resource duration matching rule parameters of the child node, and adjust the quantity range threshold or duration range threshold in the parameters. For the child nodes of the environmental interference growth dimension, compare the deviation range between the environmental interference impact data and the association matching rules. If the actual deviation exceeds the deviation range, extract the temperature matching rule or network matching rule parameters of the child node and adjust the temperature range threshold or stability level threshold in the parameters. The frequency of deviation exceeding the threshold for each growth dimension of child nodes is counted. For child nodes whose deviation exceeds the threshold, the number of associated child nodes of that child node is increased, thus expanding the judgment dimensions of the child node. Analyze the path relationships between child nodes and calculate the scheduling execution success rate for each path. The scheduling execution success rate is determined based on the comparison between the actual completion status and the planned completion status of the task under that path. For paths with a scheduling execution success rate lower than a preset success rate threshold, adjust the order of the child nodes associated with the path, or replace some child nodes in the path to optimize the path's association logic. By integrating and adjusting the child node parameters, expanding the number of child nodes, and optimizing the path relationships, the iterative optimized task scheduling decision tree is generated, and the structural information of the optimized task scheduling decision tree and the scheduling judgment rules of each child node are stored.

5. The task scheduling optimization method based on decision tree algorithm according to claim 2, characterized in that, The step of extracting task-related data from the task management system for the task dependency growth dimension includes: Determine the extraction range of the task prerequisite dependency data, and retrieve all associated task information of the task group to which the task to be scheduled belongs from the task association database of the task management system; Extract the task prerequisite dependency data from the associated task information. The task prerequisite dependency data includes the prerequisite task identifier, the prerequisite task completion status requirement, and the dependency strength data between the prerequisite task and the current task. The dependency strength data is generated based on the degree of influence of the prerequisite task on the execution of the current task. Determine the extraction range of the task-related relationship data, and retrieve the information of subsequent tasks affected by the current task from the task association database based on the current task identifier; The task post-relationship data is extracted from the subsequent task information. The task post-relationship data includes the post-task identifier, the post-task start condition requirements, and the correlation strength data between the current task and the post-task. The correlation strength data is generated based on the degree of influence of the current task on the execution of the post-task. Determine the extraction range of the task parallel constraint relationship data, and retrieve information on parallel tasks that may overlap with the current task in execution time from the task association database; The parallel task constraint relationship data is extracted from the parallel task information. The parallel task constraint relationship data includes the parallel task identifier, the parallel task resource conflict type, and the parallel task execution time conflict threshold. The resource conflict type is determined based on the overlap type of the resources required by the parallel task and the current task. The extracted task pre-dependency relationship data, task post-association relationship data, and task parallel constraint relationship data are classified and stored, and a unique data identifier is set for each type of data; The step of extracting resource status data from the resource monitoring system for the resource constraint growth dimension includes: Determine the target for extracting the available resource quantity data, and retrieve the type identifiers and corresponding quantity information of all currently available resources from the resource ledger database of the resource monitoring system; The available resources are classified according to resource type, and the current available quantity of each resource type is counted to generate the resource available quantity data. The resource available quantity data includes resource type identifier, current available quantity value, and resource quantity statistics timestamp. Determine the objects to be extracted from the resource usage duration data, and retrieve historical usage records for each resource type from the resource usage log database of the resource monitoring system; Extract the single continuous usage duration, daily cumulative usage duration, and weekly cumulative usage duration data for each resource type from the historical usage records to generate the resource usage duration data. The resource usage duration data includes the resource type identifier, the single usage duration range, the daily cumulative usage duration range, and the weekly cumulative usage duration range. Determine the target for extracting the resource load limit data, and retrieve the hardware performance parameters and software operation parameters for each resource type from the resource performance parameter database of the resource monitoring system; Based on the processor performance index, memory capacity index, and storage capacity index in the hardware performance parameters, and combined with the resource utilization ratio index in the software operation parameters, the maximum load capacity of each resource type is generated, and the resource load limit data is generated. The resource load limit data includes resource type identifier, processor load limit value, memory load limit value, and storage load limit value. The extracted data on the number of available resources, the duration of resource usage, and the upper limit of resource load are stored together to establish a correspondence between resource type identifiers and each data item. The step of extracting environmental parameter data from the environmental monitoring system for the aforementioned environmental disturbance growth dimension includes: Determine the extraction range of the temperature change data, and retrieve the real-time temperature data and historical temperature data of the task execution area from the temperature monitoring database of the environmental monitoring system; The difference between the real-time temperature data and the historical average temperature data for the same period is calculated, the change range of the temperature difference within a unit time is statistically analyzed, and the temperature change data is generated. The temperature change data includes the temperature monitoring timestamp, real-time temperature value, historical average temperature value, and the change range of the temperature difference. Determine the extraction range of the network stability data, and retrieve the real-time transmission data of the network links required for task execution from the network monitoring database of the environmental monitoring system; Network transmission rate, network latency, and network packet loss rate data are extracted from the real-time transmission data. The fluctuation range of these data within a unit of time is statistically analyzed to generate the network stability data. The network stability data includes network monitoring timestamps, transmission rate fluctuation range, latency fluctuation range, and packet loss rate fluctuation range. Determine the extraction range of the external interference frequency data, and retrieve the external interference event records within the task execution area from the interference monitoring database of the environmental monitoring system; The occurrence frequency of different types of external interference events is counted within a unit of time, and the occurrence frequency of each type of interference event is generated. The external interference frequency data includes interference type identifier, occurrence frequency per unit of time, and interference duration range. The extracted temperature change data, network stability data, and external interference frequency data are timestamped to ensure that the monitoring time of different types of environmental parameter data remains consistent.

6. The task scheduling optimization method based on decision tree algorithm according to claim 4, characterized in that, For the child nodes of the task dependency growth dimension, the deviation range between the task execution progress data and the association matching rules is compared. If the actual deviation exceeds the deviation range, the parameters of the preceding dependency matching rule or the following association matching rule of the child node are extracted, and the execution condition threshold in the parameters is adjusted, including: Extract the actual deviation data of the corresponding child node from the task execution progress data. If the child node is the first-level child node of the task dependency growth dimension and the first-level child node is a pre-dependency type child node, then the actual deviation data is the difference between the actual start time of the task and the planned start time. Retrieve the deviation range from the association matching rules of the preceding dependency type sub-node, and compare the actual deviation data with the deviation range; If the actual deviation data exceeds the upper limit of the deviation range, the execution condition threshold in the pre-dependency matching rule of the sub-node is extracted. The execution condition threshold includes the pre-task completion advance threshold and the pre-task result verification time threshold. Increase the value of the threshold for the completion lead time of the preceding tasks to allow the preceding tasks more time to complete and reduce the delay in the start of the current task. Increase the value of the threshold for the verification time of the results of the preceding tasks, and extend the verification time of the results of the preceding tasks so that the results of the preceding tasks meet the requirements of the current task execution. If the child node is a second-level child node of the task-dependent growth dimension, and the second-level child node is a post-association type child node, then the actual deviation data is the difference between the actual completion time of the task and the planned completion time. Retrieve the deviation range from the association matching rules of the subsequent association type sub-node, and compare the actual deviation data with the deviation range; If the actual deviation data exceeds the upper limit of the deviation range, the execution condition threshold in the subsequent association matching rule of the child node is extracted. The execution condition threshold includes the subsequent task start delay threshold and the subsequent task resource reservation time threshold. Increase the values ​​of the post-task start delay threshold and the post-task resource reservation time threshold.

7. The task scheduling optimization method based on decision tree algorithm according to claim 4, characterized in that, The analysis of path relationships between child nodes and the calculation of the scheduling execution success rate for each path include: Traverse the paths between all child nodes in the task scheduling decision tree. Each path contains a sequence of nodes starting from the initial growth node and passing through child nodes at different levels of each growth dimension. Assign a unique path identifier to each path and record the identifiers of the child nodes contained in the path and the connection order between the child nodes; Filter out all task execution records corresponding to the path from the scheduling execution feedback data. The task execution records include task identifier, planned completion time, actual completion time, and task completion status. Count the total number of tasks in the task execution records under this path, and the number of tasks with a completion status among the total number of tasks; Based on the comparison between the number of completed tasks and the total number of tasks, the scheduling execution success rate of this path is determined; If the number of task execution records corresponding to any path is lower than the preset record number threshold, which is insufficient to generate an effective scheduling execution success rate, then retrieve the past task execution records of that path from the historical scheduling database and supplement them to the current task execution record; The total number of tasks and the number of completed tasks are recalculated after the supplementation. Based on the comparison between the number of completed tasks and the total number of tasks, the success rate of the supplemented scheduling execution is determined. The path identifier of each path, the scheduling execution success rate, and the number of task execution records used for calculation are compiled into a path success rate statistics table, and the path success rate statistics table is stored in the path optimization module of the task scheduling decision tree.

8. The task scheduling optimization method based on decision tree algorithm according to claim 1, characterized in that, The process involves inputting the task attribute data of the task to be scheduled into the optimized task scheduling decision tree, and executing the scheduling judgment rules layer by layer through the child nodes of each growth dimension to generate a multi-dimensional adaptive scheduling scheme for the task to be scheduled, including: Extract the task attribute data of the task to be scheduled. The task attribute data includes the task identifier, the type and quantity of resources required for task execution, the task execution time requirement, the identifiers of the tasks that depend on the task before, the identifiers of the tasks that are associated with the task after, the temperature tolerance range of the task, and the network requirement level of the task. Input the task pre-dependent task identifier and task post-related task identifier from the task attribute data into the task dependency growth dimension of the optimized task scheduling decision tree; By executing the prerequisite dependency matching rule through the first-level child node of the task dependency growth dimension, it is determined whether the prerequisite task corresponding to the task prerequisite dependency task identifier meets the completion status requirements, and the first-level child node of the match is determined. Based on the first-level child node of the matching, the subsequent association matching rule is executed through the corresponding second-level child node to determine whether the subsequent task corresponding to the task subsequent association task identifier meets the start condition requirements, and to determine the matching second-level child node. Input the resource type and quantity required for task execution and the task execution duration requirement from the task attribute data into the resource constraint growth dimension of the optimized task scheduling decision tree; By executing the resource quantity matching rule through the first-level child node of the resource constraint growth dimension, it is determined whether the resource quantity required by the task conforms to the corresponding available resource quantity range, and the matching first-level child node is determined. Based on the first-level child node of the matching, the resource duration matching rule is executed through the corresponding second-level child node to determine whether the task execution duration requirement meets the corresponding resource usage duration range, and to determine the matching second-level child node. Input the task temperature tolerance range and task network requirement level from the task attribute data into the environmental disturbance growth dimension of the optimized task scheduling decision tree; By executing temperature matching rules through the first-level child node of the environmental interference growth dimension, it is determined whether the task temperature tolerance range conforms to the corresponding temperature change range, and the first-level child node for matching is determined. Based on the first-level child node of the matching, the network matching rules are executed through the corresponding second-level child node to determine whether the task network requirement level meets the corresponding network stability level, and to determine the matching second-level child node. Integrate the output results of the sub-nodes matched by each growth dimension to generate task execution sequence planning, resource allocation scheme, and environmental interference response scheme; The task execution sequence planning, the resource allocation scheme, and the environmental interference response scheme are combined to form a multi-dimensional adaptive scheduling scheme for the task to be scheduled. Set a scheme version identifier for the multi-dimensional adaptation scheduling scheme, and record the scheme generation time and the corresponding optimized task scheduling decision tree version.

9. A task scheduling optimization system based on decision tree algorithm, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the task scheduling optimization method based on the decision tree algorithm according to any one of claims 1 to 8 by executing the machine executable instructions.

10. A computer program product, characterized in that, The computer program product includes machine-executable instructions stored in a computer-readable storage medium. The processor of the task scheduling optimization system based on the decision tree algorithm reads the machine-executable instructions from the computer-readable storage medium and executes the machine-executable instructions, causing the task scheduling optimization system based on the decision tree algorithm to perform the task scheduling optimization method based on the decision tree algorithm as described in any one of claims 1 to 8.

Images