AI-driven software development resource intelligent scheduling method and system in cloud-native environment

Abstract

The application discloses an AI-driven software development resource intelligent scheduling method and system in a cloud-native environment, and relates to the technical field of software development.The method comprises the following specific steps: S1: collecting resource data, task data and environment state data in the software development process in the cloud-native environment; S2: performing cleaning, normalization and feature extraction processing on the collected data, removing abnormal data and redundant data, converting data in different formats into a unified format, and extracting key features in the data to obtain a standardized data set; the application builds a resource demand prediction sub-model based on a time series prediction model and incorporating a feature weight reinforcement mechanism, deeply mines the complex correlation between task data and historical resource data, and enables accurate prediction of resource demand of each software development task, thereby avoiding the problem that traditional methods cannot accurately predict task resource demand.

Description

Technical Field

[0001] This invention relates to the field of software development technology, specifically to an AI-driven intelligent scheduling method and system for software development resources in a cloud-native environment. Background Technology

[0002] With the rapid development of cloud-native technologies, cloud-native environments, with their significant characteristics such as elastic scaling, containerized deployment, and microservice architecture, have become the core environment relied upon by many software development projects. In a cloud-native environment, the software development process involves various types of resources, including CPU, memory, storage, and network bandwidth. The reasonable allocation and efficient scheduling of these resources play a crucial role in ensuring the smooth execution of software development tasks and improving overall development efficiency. However, due to the diversity, dynamism, and priority differences of software development tasks, different types of tasks have different resource requirements, and resource requirements may also change dynamically as the development progresses during task execution.

[0003] In the existing field of resource scheduling technology, traditional resource scheduling methods mainly suffer from two problems. On the one hand, some resource scheduling methods use simple load balancing algorithms, such as round-robin and weighted round-robin. These algorithms only consider the current load of nodes but ignore the resource requirements of tasks and historical data. Because they do not fully combine the actual needs of tasks, the scheduling accuracy is low, and they cannot accurately allocate appropriate resources to different tasks. This can easily lead to some nodes being overloaded while others are idle, resulting in serious waste of resources. At the same time, it will also negatively affect the execution efficiency of software development tasks. On the other hand, although some methods introduce machine learning models for resource prediction, the structure of these models is relatively simple and fails to fully explore the complex relationship between task data and resource data. This results in limited accuracy in predicting resource demand, making it difficult to meet the accuracy requirements of resource scheduling in cloud-native environments. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an AI-driven intelligent scheduling method and system for software development resources in a cloud-native environment. This system can intelligently allocate and dynamically adjust various resources during software development by constructing an AI-driven resource scheduling model and combining it with the dynamic characteristics of the cloud-native environment. It collects resource and task data from the entire software development process, preprocesses it, and inputs it into a trained resource demand prediction sub-model to obtain resource demand prediction results for each task. Based on real-time resource status data in the cloud-native environment, an optimal resource scheduling scheme is generated through a resource optimization allocation sub-model. Finally, resource scheduling is executed according to this scheme, and resource usage and task progress are monitored in real time during the scheduling process. If any anomalies are detected, the scheduling strategy can be dynamically adjusted, significantly improving the utilization rate of software development resources in the cloud-native environment, effectively reducing resource waste, and greatly improving software development efficiency.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: On one hand, an AI-driven intelligent scheduling method for software development resources in a cloud-native environment, comprising the following specific steps:

[0006] S1: Collect resource data, task data, and environment status data during the software development process in a cloud-native environment;

[0007] S2.: Clean, normalize, and extract features from the collected data to remove abnormal and redundant data, convert data of different formats into a unified format, and extract key features from the data to obtain a standardized dataset.

[0008] S3: Input the task data and historical resource usage data from the obtained standardized dataset into the pre-trained resource demand prediction sub-model, and output the resource demand prediction results for each software development task.

[0009] S4: Real-time data collection of resource availability, resource utilization, node communication latency, and node fault status for each node is generated using monitoring components in the cloud-native environment.

[0010] S5: Input the obtained resource demand forecast results and the obtained real-time resource status table into the resource optimization allocation sub-model, and output the optimal resource scheduling scheme for each software development task.

[0011] S6: Based on the generated optimal resource scheduling scheme, the resource allocation and scheduling are completed through the scheduling execution component in the cloud-native environment. At the same time, the task execution progress, resource usage deviation and node status changes are monitored in real time. When the resource usage deviation exceeds the preset threshold or a node fails, the real-time monitoring data is fed back to the resource demand prediction sub-model and the resource optimization allocation sub-model to regenerate the adjusted resource scheduling scheme.

[0012] Furthermore, in step S1, the resource data includes CPU utilization, memory usage, storage capacity, and network bandwidth; the task data includes task type, task priority, estimated task execution time, and task dependencies; and the environment status data includes node load and service health status.

[0013] Furthermore, in step S3, the resource demand prediction sub-model is constructed based on the time-series prediction model and incorporates a feature weight enhancement mechanism. It calculates and outputs the resource demand prediction results for each software development task using a prediction formula, which is: ,in, At any moment , No. Resource requirement forecasts for each software development task. It is the number of time steps in the historical data used for prediction. It is a time-series correlation weight. It is the first At that moment, with the first Actual resource consumption data for historical tasks of the same type. It is the total number of key features of the task involved in the prediction. These are feature weight coefficients. It is the first At that moment, the The first task Standardized values ​​for each feature.

[0014] Furthermore, the aforementioned It is the time-series correlation weight, and its calculation formula is: ,in, It is a time-series sensitivity coefficient, used to adjust the correlation sensitivity between historical data and current predictions. It is the first Historical average resource consumption for this type of task It is the first The standard deviation of historical resource consumption for this type of task It is a summation index that iterates through all historical time steps within all time series windows to ensure that all historical data participate in the weight calculation; These are the feature weight coefficients, and their calculation formula is: ,in, It is the feature enhancement coefficient, used to control the overall strength of feature weights. It is an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise. For the first Key feature set of the class of tasks, Corr It is the first The first feature and the second Correlation coefficients of resource consumption for similar tasks It is a summation index that iterates through all task features and is used to normalize the feature correlation coefficient.

[0015] Furthermore, in step S3, the resource demand prediction results include the number of CPU cores, memory size, storage capacity, and network bandwidth required for each task.

[0016] Furthermore, in step S5, the resource optimization allocation sub-model takes maximizing resource utilization as its objective function and task priority, node load threshold, and resource supply-demand matching degree as constraints. It solves for and outputs the optimal resource scheduling scheme for each software development task. Its objective function is: ,in, It refers to the overall utilization rate of software development resources in a cloud-native environment. It is the total number of software development tasks. It represents the total number of nodes with allocable resources in a cloud-native environment. It is a task-node assignment indicator variable, if the first The task was assigned to the first 1 node If not allocated, , It is the first The task was assigned to the first The actual amount of resources used per node It is the first Total resources of each node.

[0017] Furthermore, the task priority constraint is: when task Priority Task Priority hour, The node load threshold constraint is: for all nodes : The resource supply and demand matching constraint is: for all tasks ,node : ,in, , These are any two different software development tasks. It is the first The load threshold of each node. It is the first Forecast of resource requirements for each task. It is the threshold for the mismatch between resource supply and demand.

[0018] Furthermore, the optimal resource scheduling scheme includes resource allocation nodes, resource allocation amount, and resource allocation time.

[0019] Furthermore, the specific content monitored in real time in step S6 includes:

[0020] Task execution progress monitoring: By periodically collecting task execution logs, the ratio of the completed workload to the total workload is calculated to obtain the task execution progress;

[0021] Resource usage deviation monitoring: Calculate the difference between actual resource usage and predicted resource usage. When the difference accounts for more than 5%-10% of the predicted resource usage, it is judged as an abnormal resource usage deviation.

[0022] Node status monitoring: Real-time detection of node CPU usage, memory usage, network connectivity, and service operation status. When the node CPU usage or memory usage exceeds 80%-90%, or when the network is interrupted or the service stops, it is determined that the node status is abnormal.

[0023] On the other hand, an AI-driven intelligent scheduling system for software development resources in a cloud-native environment, which includes:

[0024] Data acquisition module: used to collect resource data, task data, and environment status data during the software development process in a cloud-native environment;

[0025] Data preprocessing module: used to clean, normalize and extract features from the collected data to obtain a standardized dataset;

[0026] Resource demand forecasting module: It has a built-in resource demand forecasting sub-model, which is used to receive standardized datasets and output the resource demand forecasting results for each software development task.

[0027] Real-time resource awareness module: Collects resource status data of each node in real time through monitoring components in the cloud-native environment and generates a real-time resource status table;

[0028] The optimal scheduling scheme generation module has a built-in resource optimization allocation sub-model, which receives resource demand prediction results and real-time resource status tables. With the goal of maximizing resource utilization and combined with preset constraints, it generates and outputs the optimal resource scheduling scheme through intelligent optimization algorithms.

[0029] The scheduling execution and dynamic adjustment module is used to execute resource scheduling based on the optimal resource scheduling scheme, monitor the status of tasks and resources in real time, and dynamically adjust the scheduling scheme.

[0030] Compared with existing technologies, this AI-driven intelligent scheduling method and system for software development resources in a cloud-native environment has the following advantages:

[0031] I. This invention constructs a resource demand prediction sub-model based on a time-series prediction model and incorporates a feature weight enhancement mechanism. It deeply mines the complex relationship between task data and historical resource data, enabling it to accurately predict the resource requirements of each software development task. This avoids the problem that traditional methods are unable to accurately predict task resource requirements, ensuring that resource allocation is more in line with the actual needs of the task. It also reduces the phenomenon of some nodes being overloaded and some nodes being idle due to unreasonable resource allocation, effectively improving the rationality of resource allocation.

[0032] Second, this invention uses visualization tools to display task execution progress, resource usage, and node status in real time, and sets up an early warning mechanism. When abnormal task progress, abnormal resource usage, or node failure is detected, real-time monitoring data can be fed back to the resource demand prediction sub-model and the resource optimization allocation sub-model in a timely manner. The adjusted resource scheduling scheme is regenerated and the resources are reallocated, so that resource scheduling can adapt to the dynamic changes of the cloud-native environment in real time, ensuring the continuous and efficient progress of software development tasks, significantly improving software development efficiency, and avoiding task delays or failures due to resource issues.

[0033] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0035] Figure 1 A structural block diagram of an AI-driven intelligent scheduling system for software development resources in a cloud-native environment;

[0036] Figure 2 This is a flowchart of an AI-driven intelligent scheduling method for software development resources in a cloud-native environment. Detailed Implementation

[0037] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0038] This invention provides a method and system for intelligent scheduling of AI-driven software development resources in a cloud-native environment, such as... Figure 1As shown, it includes: a data acquisition module, a data preprocessing module, a resource demand prediction module, a real-time resource perception module, an optimal scheduling scheme generation module, and a scheduling execution and dynamic adjustment module. By constructing an AI-driven resource scheduling model and combining it with the dynamic characteristics of the cloud-native environment, it achieves intelligent allocation and dynamic adjustment of various resources during the software development process. It collects resource and task data from the entire software development process, preprocesses it, and inputs it into a trained resource demand prediction sub-model to obtain resource demand prediction results for each task. Based on the real-time resource status data of the cloud-native environment, it generates an optimal resource scheduling scheme through a resource optimization allocation sub-model. Finally, it executes resource scheduling according to this scheme, and monitors resource usage and task progress in real time during the scheduling process. Once an anomaly is detected, it can dynamically adjust the scheduling strategy, significantly improving the utilization rate of software development resources in the cloud-native environment, effectively reducing resource waste, and greatly improving software development efficiency.

[0039] Example 1

[0040] A small to medium-sized technology company launched an office collaboration system development project, which requires the development and launch of three core functional modules: document management, instant messaging, and scheduling. The project is deployed in a cloud-native environment and includes several Kubernetes worker nodes, with node configurations that meet the resource requirements of conventional software development.

[0041] like Figure 2 As shown, cloud-native environment monitoring tools are used to continuously collect resource usage information for each node, including the real-time occupancy and remaining status of computing, storage, and network resources. Task information is obtained from the project management platform and code management platform to clarify the task type, planned execution duration, and task priority for each development stage. Furthermore, the dependencies between tasks must be considered. Module-specific testing can only begin after the code has been compiled, while integration testing can only begin after all modules have passed independent testing. Pre-production deployment requires successful integration testing as a prerequisite. Simultaneously, the environmental monitoring interface is used to monitor the operational health status of each node, data transmission latency between nodes, and other environmental information to ensure comprehensive and real-time basic data, providing parameter support for subsequent formula calculations.

[0042] The collected data is processed and optimized, abnormal data caused by temporary malfunctions of monitoring equipment is removed, and missing node resource data is supplemented to ensure... , The core parameters are accurate; resource and environmental data of different types and scales are converted into a unified standard format to facilitate subsequent formula calculations; key features are extracted from task data, such as the functional module to which the task belongs, the basic conditions required for task execution, and the pre- and post-task related links, to form a standardized dataset, providing input data for the application of resource demand prediction formulas and resource optimization allocation formulas.

[0043] Based on the experience of resource consumption in similar historical office system development projects, and combined with the characteristics of each task in the current project, the predicted resource requirements for each task are calculated using the core formula for resource requirement forecasting. Its formula is ,in, That is, the corresponding task At the present moment Resource demand forecast , This data represents the actual resource consumption of similar historical tasks. The weighting of the impact of data from different historical time steps is calculated based on the deviation between historical data and the average consumption of similar tasks. For the current task Standardized values ​​for key features Used to enhance the impact of key features on resource requirements, such as task priority. corresponding The value is higher than other features; for the code compilation task of the document management module, the calculation result is the task's... For integration testing tasks, the resource consumption data of historical integration tests is combined with the characteristics of the number of modules currently participating in the test, and its value is calculated using a formula. This ensures that the resource requirement forecasts for each task are aligned with the actual execution scenario.

[0044] Monitoring components deployed on each node track dynamic changes in node resources in real time, including real-time resource utilization and remaining available resources (for updating). (Related calculations), and regularly update the resource status information table to clarify the status of each node, including whether nodes are experiencing operational abnormalities and network connectivity. The current value ensures timely monitoring of resource distribution and health across the entire cloud-native environment, providing a basis for subsequent resource optimization and allocation formulas. The parameters provide the latest values ​​to ensure the accuracy of the formula calculation results.

[0045] At the same time, the rationality of scheduling is ensured by combining the three types of constraint formulas:

[0046] For task priority constraint formulas: such as pre-production deployment tasks Integration testing tasks ,because It must meet the following requirements: ;

[0047] Node load threshold constraint formula: for all nodes Ensure that resource usage does not exceed the safe limit: ;

[0048] Resource supply and demand matching constraint formula: for all tasks ,node Ensure the allocation of resources and (basically the same) ,in, Assign indicator variables to task-nodes. The actual resource usage after the task is assigned to a node. The node load safety threshold, The allowable range for resource supply and demand deviations;

[0049] Based on the calculation results of the above formula, a resource allocation plan is formulated: the code compilation task is allocated to the current... There are a sufficient number of remaining and stable nodes; independent module test tasks are started on nodes with available resources after the corresponding modules have been compiled; integration test tasks coordinate the idle resources of multiple nodes to ensure... Pre-production deployment tasks are reserved for adequate, Nodes with ample capacity ensure a smooth launch process.

[0050] According to the established scheduling plan, resource allocation and task initiation are completed using cloud-native environment scheduling tools. Task execution progress and resource usage are tracked in real time. If a node experiences a sudden event during task execution, [further action is required]. The remaining amount decreases sharply, or the actual resource usage for a certain task is... Deviation exceeds Adjust resource allocation promptly: Update The value will migrate non-critical tasks on resource-constrained nodes to resource-sufficient nodes, or adjust... Supplement the increased resources for tasks with increased demand to ensure that all tasks can proceed normally and meet the requirements of the various constraint formulas, without being affected by resource issues.

[0051] Throughout the project development process, calculations using the resource utilization formula show that the overall resource utilization rate is significantly higher than that of traditional methods, with no obvious resource idleness or overload. Tasks at each stage are executed in an orderly manner according to plan, without any delays due to resource allocation issues. The pre-production deployment stage is completed smoothly, and the office collaboration system is launched on schedule, meeting the company's daily office needs. The project development efficiency is significantly higher than that of traditional manual resource allocation methods.

[0052] Example 2

[0053] To prepare for an upcoming promotional event, a large e-commerce company launched a platform system upgrade project, which involves functional optimization and performance improvement of four core business systems: product display, order processing, payment settlement, and logistics tracking. The project is deployed in a large-scale cloud-native environment, which includes multiple Kubernetes worker nodes. Some nodes are used to support the existing normal business operations, while others are dedicated to this upgrade project.

[0054] Through a distributed monitoring system in a cloud-native environment, resource data is collected simultaneously from dedicated nodes used for upgrade projects and shared nodes supporting existing businesses, as well as the baseline of resource usage and fluctuations in idle resources for various nodes. The task list for upgrade projects is obtained from the enterprise DevOps platform, clarifying the specific content, planned execution cycle, and dependencies between tasks for code compilation, testing, and deployment of each business system. For example, performance testing of the order processing system must be initiated after the system's code compilation and functional testing are completed, and gray-scale deployment in the production environment requires all business systems to pass performance testing. Furthermore, environmental data such as the service operation status and network transmission stability of each node are collected in real time to ensure data coverage of the entire project process.

[0055] The collected resource data is filtered to remove abnormal resource usage data caused by temporary fluctuations during peak business periods, while retaining valid data that reflects the normal operating status of nodes. The format standards of resource data, task data, and environmental data are unified to eliminate differences between data from different sources. Key information is extracted from task data, such as the business system to which the task belongs, the special requirements of task execution for hardware resources, and the time window limit for task execution, to build a standardized dataset and provide a foundation for subsequent resource analysis and scheduling.

[0056] Drawing on past experience with system upgrades prior to large-scale promotional events on e-commerce platforms, and considering the functional complexity and data processing volume of each business system upgrade in this instance, we analyzed the resource requirements for each task. For example, the product display system requires the addition of multiple promotional activity display modules, resulting in a significant increase in code volume; therefore, we anticipated that its compilation tasks would require more resources than a typical upgrade. The order processing system's performance stress test needs to simulate high-concurrency order scenarios, and we estimated that its resource requirements would be substantial. The payment settlement system's deployment tasks have high requirements for node stability, so we needed to reserve stable node resources for it to ensure that the predicted resource requirements for each task match the actual upgrade scenario.

[0057] Through real-time monitoring tools, we continuously track the resource dynamics of all nodes, including changes in resource usage of dedicated nodes and the idle resource status of shared nodes under the current business load fluctuations. For shared nodes, we focus on the resource idle window during business off-peak periods to find suitable opportunities for tasks that need to be executed on shared nodes. We regularly update the resource status table to ensure that we can keep track of the availability and health status of resources on each node in a timely manner, providing real-time reference for resource allocation.

[0058] With the goal of maximizing the use of dedicated node resources and making reasonable use of idle resources on shared nodes, priority is given to ensuring the resource requirements of production environment canary deployment and performance stress testing tasks. Strict control is maintained over the use of all node resources to ensure they do not exceed safe thresholds, and the resources allocated to tasks are largely consistent with the estimated demand. A scheduling plan is formulated as follows: code compilation tasks for each business system are prioritized to start on dedicated nodes; if dedicated node resources are scarce, they are executed during off-peak hours on shared nodes. Functional testing tasks are conducted using idle resources on dedicated nodes after the corresponding system compilation is completed. Performance stress testing tasks are assigned to dedicated nodes with higher configurations and stable performance to ensure accurate test results. Production environment canary deployment tasks are deployed in batches on nodes with low load and high stability to minimize the impact on existing business operations.

[0059] Task execution and resource allocation are carried out according to the scheduling plan, and task progress and node resource status are monitored in real time. During performance stress testing, if a dedicated node's resources approach the threshold due to excessive test load, some test tasks are promptly diverted to other idle dedicated nodes. When executing compilation tasks on shared nodes, if the load of existing business suddenly increases, the compilation tasks on the shared nodes are immediately suspended, and resources are prioritized to ensure the existing business. The compilation tasks are restarted after the business load decreases. Through dynamic adjustment, it is ensured that the upgrade project and existing business do not interfere with each other and can both run normally.

[0060] During the project upgrade, dedicated node resources were utilized efficiently, and idle shared node resources were used effectively without any waste or overload. The upgrade tasks for each core business system were completed within the planned timeframe, performance stress testing was thorough, and the gray-scale deployment in the production environment proceeded smoothly. The upgraded system operated stably during the promotional period, effectively handling high-concurrency business requests and meeting user shopping needs. The project upgrade goals were successfully achieved.

[0061] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligent scheduling of software development resources driven by AI in a cloud-native environment, characterized in that, The method includes the following specific steps: S1: Collect resource data, task data, and environment status data during the software development process in a cloud-native environment; S2.: Clean, normalize, and extract features from the collected data to remove abnormal and redundant data, convert data of different formats into a unified format, and extract key features from the data to obtain a standardized dataset. S3: Input the task data and historical resource usage data from the obtained standardized dataset into the pre-trained resource demand prediction sub-model, and output the resource demand prediction results for each software development task. S4: Real-time data collection of resource availability, resource utilization, node communication latency, and node fault status for each node is generated using monitoring components in the cloud-native environment. S5: Input the obtained resource demand forecast results and the obtained real-time resource status table into the resource optimization allocation sub-model, and output the optimal resource scheduling scheme for each software development task. S6: Based on the generated optimal resource scheduling scheme, the resource allocation and scheduling are completed through the scheduling execution component in the cloud-native environment. At the same time, the task execution progress, resource usage deviation and node status changes are monitored in real time. When the resource usage deviation exceeds the preset threshold or a node fails, the real-time monitoring data is fed back to the resource demand prediction sub-model and the resource optimization allocation sub-model to regenerate the adjusted resource scheduling scheme.

2. The AI-driven intelligent scheduling method and system for software development resources in a cloud-native environment according to claim 1, characterized in that, In step S1, the resource data includes CPU utilization, memory usage, storage capacity, and network bandwidth; the task data includes task type, task priority, estimated task execution time, and task dependencies; and the environment status data includes node load and service health status.

3. The AI-driven intelligent scheduling method and system for software development resources in a cloud-native environment according to claim 1, characterized in that, In step S3, the resource demand prediction sub-model is constructed based on the time-series prediction model and incorporates a feature weight enhancement mechanism. It calculates and outputs the resource demand prediction results for each software development task using a prediction formula, which is: ,in, At any moment , No. Resource requirement forecasts for each software development task. It is the number of time steps in the historical data used for prediction. It is a time-series correlation weight. It is the first At that moment, with the first Actual resource consumption data for historical tasks of the same type. It is the total number of key features of the task involved in the prediction. These are feature weight coefficients. It is the first At that moment, the The first task Standardized values ​​for each feature.

4. The AI-driven intelligent scheduling method and system for software development resources in a cloud-native environment according to claim 3, characterized in that, The It is the time-series correlation weight, and its calculation formula is: ,in, It is a time-series sensitivity coefficient, used to adjust the correlation sensitivity between historical data and current predictions. It is the first Historical average resource consumption for this type of task It is the first The standard deviation of historical resource consumption for this type of task It is a summation index that iterates through all historical time steps within all time series windows to ensure that all historical data participate in the weight calculation; These are the feature weight coefficients, and their calculation formula is: ,in, It is the feature enhancement coefficient, used to control the overall strength of feature weights. It is an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise. For the first Key feature set of the class of tasks, Corr It is the first The first feature and the second Correlation coefficients of resource consumption for similar tasks It is a summation index that iterates through all task features and is used to normalize the feature correlation coefficient.

5. The AI-driven intelligent scheduling method for software development resources in a cloud-native environment according to claim 1, characterized in that, In step S3, the resource demand prediction results include the number of CPU cores, memory size, storage capacity, and network bandwidth required for each task.

6. The AI-driven intelligent scheduling method for software development resources in a cloud-native environment according to claim 1, characterized in that, In step S5, the resource optimization allocation sub-model takes maximizing resource utilization as its objective function and task priority, node load threshold, and resource supply-demand matching degree as constraints. It solves for and outputs the optimal resource scheduling scheme for each software development task. Its objective function is: ,in, It refers to the overall utilization rate of software development resources in a cloud-native environment. It is the total number of software development tasks. It represents the total number of nodes with allocable resources in a cloud-native environment. It is a task-node assignment indicator variable, if the first The task was assigned to the first 1 node If not allocated, , It is the first The task was assigned to the first The actual amount of resources used per node It is the first Total resources of each node.

7. The AI-driven intelligent scheduling method for software development resources in a cloud-native environment according to claim 6, characterized in that, The task priority constraint is: when the task... Priority Task Priority hour, The node load threshold constraint is: for all nodes : The resource supply and demand matching constraint is: for all tasks ,node : ,in, , These are any two different software development tasks. It is the first The load threshold of each node. It is the first Forecast of resource requirements for each task. It is the threshold for the mismatch between resource supply and demand.

8. The AI-driven intelligent scheduling method for software development resources in a cloud-native environment according to claim 1, characterized in that, The optimal resource scheduling scheme includes resource allocation nodes, resource allocation amount, and resource allocation time.

9. The AI-driven intelligent scheduling method for software development resources in a cloud-native environment according to claim 1, characterized in that, The specific content of real-time monitoring in step S6 includes: Task execution progress monitoring: By periodically collecting task execution logs, the ratio of the completed workload to the total workload is calculated to obtain the task execution progress; Resource usage deviation monitoring: Calculate the difference between actual resource usage and predicted resource usage. When the difference accounts for more than 5%-10% of the predicted resource usage, it is judged as an abnormal resource usage deviation. Node status monitoring: Real-time detection of node CPU usage, memory usage, network connectivity, and service operation status. When the node CPU usage or memory usage exceeds 80%-90%, or when the network is interrupted or the service stops, it is determined that the node status is abnormal.

10. An AI-driven intelligent scheduling system for software development resources in a cloud-native environment, the system being applicable to the AI-driven intelligent scheduling method for software development resources in a cloud-native environment as described in any one of claims 1-9, characterized in that, The system includes: Data acquisition module: used to collect resource data, task data, and environment status data during the software development process in a cloud-native environment; Data preprocessing module: used to clean, normalize and extract features from the collected data to obtain a standardized dataset; Resource demand forecasting module: It has a built-in resource demand forecasting sub-model, which is used to receive standardized datasets and output the resource demand forecasting results for each software development task. Real-time resource awareness module: Collects resource status data of each node in real time through monitoring components in the cloud-native environment and generates a real-time resource status table; The optimal scheduling scheme generation module has a built-in resource optimization allocation sub-model, which receives resource demand prediction results and real-time resource status tables. With the goal of maximizing resource utilization and combined with preset constraints, it generates and outputs the optimal resource scheduling scheme through intelligent optimization algorithms. The scheduling execution and dynamic adjustment module is used to execute resource scheduling based on the optimal resource scheduling scheme, monitor the status of tasks and resources in real time, and dynamically adjust the scheduling scheme.

Images