Intelligent computing power resource scheduling and optimizing system

Abstract

The invention provides an intelligent computing power resource scheduling and optimizing system, which relates to the field of electric digital data processing, and comprises a monitoring acquisition module, a scheduling distribution module, an optimization prediction module and a task management module, and is characterized in that the monitoring acquisition module is used for acquiring and monitoring operation data of computing power resources; the scheduling distribution module dynamically schedules computing power resources on the basis of task requirements, the optimization prediction module predicts resource requirements on the basis of historical data and optimizes resource distribution, and the task management module is used for managing task information of a user; according to the system, the resource is dynamically adjusted in advance by predicting the resource demand change of the task, so that the operation efficiency of the resource node can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of electronic digital data processing, and in particular to an intelligent computing resource scheduling and optimization system. Background Art

[0002] Traditional computing resource scheduling methods are mostly based on preset strategies, which are difficult to adapt to dynamically changing computing loads, resulting in low resource utilization, long task waiting times, and even computing node overloads. These bottlenecks not only affect the overall performance and response speed of the system, but also restrict the scalability and flexibility of computing resources. Therefore, an intelligent scheduling system is needed that can perceive task requirements and resource load status in real time, and make predictions and optimize resource allocation based on historical data.

[0003] The foregoing discussion of the background art is intended only to facilitate an understanding of the present invention. This discussion does not acknowledge or admit that any of the material referred to is part of the common general knowledge.

[0004] Now many resource optimization systems have been developed. After a lot of searching and reference, it is found that the existing resource optimization systems are such as the system disclosed in the publication number CN116471333B. These system methods generally include: determining the labels corresponding to the nodes based on the corresponding layers in the topological structure; determining the identifiers corresponding to the nodes based on the corresponding blocks in the topological structure; for the nodes corresponding to the electronic node identifiers, determining the electronic nodes to be scheduled based on the classical operations research nonlinear optimization global scheduling algorithm; for the nodes corresponding to the quantum node identifiers, determining the quantum nodes to be scheduled based on the quantum machine learning intelligent global scheduling algorithm; when the target task is obtained, determining the nodes to be scheduled from the electronic nodes to be scheduled and the quantum nodes to be scheduled, and determining the scheduling path. However, the system cannot predict the changes in the resource requirements of the task, resulting in the lag in the scheduling of resources, and the overall operation efficiency of the system needs to be improved. Summary of the invention

[0005] The purpose of the present invention is to propose an intelligent computing resource scheduling and optimization system to address the existing deficiencies.

[0006] The present invention adopts the following technical solution: An intelligent computing resource scheduling and optimization system, including a monitoring and acquisition module, a scheduling and allocation module, an optimization and prediction module, and a task management module; The monitoring and collection module is used to collect and monitor the operation data of computing resources, the scheduling and allocation module dynamically schedules computing resources based on task requirements, the optimization and prediction module predicts resource requirements based on historical data and optimizes resource allocation, and the task management module is used to manage user task information; The monitoring and acquisition module includes a hardware monitoring unit, a resource acquisition unit and a transmission control unit. The hardware monitoring unit is used to monitor the hardware status information of the computing node, the resource acquisition unit is used to collect the usage of resources by the task, and the transmission control unit is used to control the external transmission of the monitoring and acquisition data; The scheduling and allocation module includes a resource allocation unit, a task scheduling unit and a strategy selection unit, wherein the resource allocation unit is used to calculate the amount of resources allocated to each task, the task scheduling unit schedules the task to the corresponding node based on the allocated amount of resources, and the strategy selection unit is used to select the scheduling strategy applied by the task scheduling unit; The optimization prediction module includes a historical data storage unit, a resource prediction unit and a dynamic optimization unit. The historical data storage unit is used to store historical data of resource utilization of tasks. The resource prediction unit is used to predict changes in the demand for resources of tasks. The dynamic optimization unit optimizes the task resources based on the prediction information. The task management module includes a user information management unit, a task information management unit and a task submission control unit, wherein the user information management unit is used to store and manage user information, the task information management unit is used to store and manage task information, and the task submission control unit is used to control the process of submitting tasks to the scheduling allocation module; Further, the resource prediction unit includes a comprehensive feature extraction processor, a personalized feature matching processor and a prediction output processor, wherein the comprehensive feature extraction processor is used to extract the comprehensive features of resource utilization of the same type of tasks, the personalized feature matching processor matches the case data similar to the target task based on the personalized features, and the prediction output processor predicts the subsequent resource demand of the target task based on the comprehensive features and the personalized features; Further, the dynamic optimization unit includes a summary analysis processor, a dynamic optimization processor and an adjustment output processor, wherein the summary analysis processor is used to summarize and analyze the prediction information of all tasks, the dynamic adjustment processor is used to optimize and analyze the resource allocation of the target task, and the adjustment output processor is used to output the adjustment information to the corresponding node; Further, the comprehensive feature extraction processor divides the resource utilization rate change function in the case into m segments on average, and respectively calculates the mean ratio of the latter segment to the former segment, and obtains a sequence consisting of m-1 ratios. The comprehensive feature extraction processor calculates the standard deviation of the ratios at each rank of the same type of task case, and selects the rank with a standard deviation less than a threshold, and the selected rank constitutes the feature rank, and the feature rank and the corresponding ratio at the feature rank together constitute the comprehensive feature of the task of this type; The personality feature matching processor divides the resource utilization rate change function into n segments on average based on the current progress of the target task, and the value of n is calculated according to the following formula: ; in, is the current progress of the target task; The personality feature matching processor calculates the mean ratio of the latter section to the former section respectively, and obtains a sequence consisting of n-1 ratios. The personality feature matching processor calculates the personality difference index Sr between the target task and the case task according to the following formula: ; Among them, Ch is the feature sequence set corresponding to the task type, a i represents the i-th ratio of the target task, b i is the i-th ratio representing the case task; The personality feature matching processor uses the case task with the smallest personality difference index as a matching prediction benchmark case; Furthermore, the prediction output processor calculates the reference deviation value rd according to the following formula: ; Among them, c i represents the i-th ratio in the comprehensive feature, is the number of sequences less than n in the feature sequence set; The prediction output processor calculates the resource utilization mean au of the target task in the next period according to the following formula: ; Where v is the maximum resource utilization of the target task, and h(n) is the nth segment mean of the resource utilization change function in the prediction benchmark case.

[0007] The beneficial effects achieved by the present invention are: This system obtains comprehensive feature data by analyzing historical case data, and matches the most similar case tasks as benchmark cases based on individual features, predicts the average resource demand for the target task in the next time period, and adjusts computing resources in advance based on the predicted data, which can effectively improve the utilization rate of computing resources in the node and improve the operating efficiency of the node.

[0008] To further understand the features and technical contents of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are only for reference and description and are not intended to limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a schematic diagram of the overall structural framework of the present invention; Figure 2 This is a schematic diagram of the monitoring and acquisition module of the present invention; Figure 3 This is a schematic diagram of the scheduling and allocation module of the present invention; Figure 4 This is a schematic diagram of the optimization prediction module of the present invention; Figure 5 This is a schematic diagram of the task management module of the present invention; Figure 6 It is a comparison chart of the experimental test results of the present invention. DETAILED DESCRIPTION

[0010] The following is an explanation of the embodiments of the present invention through specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted according to actual sizes. It is stated in advance. The following embodiments will further explain the relevant technical contents of the present invention in detail, but the disclosed contents are not intended to limit the scope of protection of the present invention.

[0011] Embodiment 1: This embodiment provides an intelligent computing resource scheduling and optimization system, combining Figure 1 , including monitoring and acquisition module, scheduling and allocation module, optimization and prediction module and task management module.

[0012] The monitoring and acquisition module is used to collect and monitor the operating data of computing resources, the scheduling and allocation module dynamically schedules computing resources based on task requirements, the optimization and prediction module predicts resource requirements based on historical data and optimizes resource allocation, and the task management module is used to manage user task information.

[0013] The monitoring and acquisition module includes a hardware monitoring unit, a resource acquisition unit and a transmission control unit. The hardware monitoring unit is used to monitor the hardware status information of the computing node, the resource acquisition unit is used to collect the resource usage of the task, and the transmission control unit is used to control the external transmission of the monitoring and acquisition data.

[0014] The scheduling allocation module includes a resource allocation unit, a task scheduling unit and a strategy selection unit. The resource allocation unit is used to calculate the amount of resources allocated to each task. The task scheduling unit schedules the task to the corresponding node based on the allocated resource amount. The strategy selection unit is used to select the scheduling strategy applied by the task scheduling unit.

[0015] The optimization prediction module includes a historical data storage unit, a resource prediction unit and a dynamic optimization unit. The historical data storage unit is used to store historical data of task resource utilization, the resource prediction unit is used to predict changes in task demand for resources, and the dynamic optimization unit optimizes task resources based on the prediction information.

[0016] The task management module includes a user information management unit, a task information management unit and a task submission control unit. The user information management unit is used to store and manage user information, the task information management unit is used to store and manage task information, and the task submission control unit is used to control the process of submitting tasks to the scheduling allocation module.

[0017] The resource prediction unit includes a comprehensive feature extraction processor, an individual feature matching processor and a prediction output processor. The comprehensive feature extraction processor is used to extract the comprehensive features of resource utilization of tasks of the same type. The individual feature matching processor matches case data similar to the target task based on the individual features. The prediction output processor predicts the subsequent resource requirements of the target task based on the comprehensive features and the individual features.

[0018] The dynamic optimization unit includes a summary analysis processor, a dynamic optimization processor and an adjustment output processor. The summary analysis processor is used to summarize and analyze the prediction information of all tasks, the dynamic adjustment processor is used to optimize the resource allocation of the target task, and the adjustment output processor is used to output the adjustment information to the corresponding node.

[0019] The comprehensive feature extraction processor divides the resource utilization change function in the case into m segments on average, and calculates the mean ratio of the latter segment to the former segment respectively, and obtains a sequence consisting of m-1 ratios. The comprehensive feature extraction processor calculates the standard deviation of the ratios at each rank of the same type of task cases, and screens out the rank with a standard deviation less than a threshold. The screened rank constitutes the feature rank, and the feature rank and the corresponding ratio at the feature rank together constitute the comprehensive feature of this type of task.

[0020] The personality feature matching processor divides the resource utilization rate change function into n segments on average based on the current progress of the target task, and the value of n is calculated according to the following formula: .

[0021] in, The current progress of the target task.

[0022] The personality feature matching processor calculates the mean ratio of the latter section to the former section respectively, and obtains a sequence consisting of n-1 ratios. The personality feature matching processor calculates the personality difference index Sr between the target task and the case task according to the following formula: .

[0023] Among them, Ch is the feature sequence set corresponding to the task type, a i represents the i-th ratio of the target task, b i is the i-th ratio representing the case task.

[0024] The personality feature matching processor uses the case task with the smallest personality difference index as a matching prediction benchmark case.

[0025] The prediction output processor calculates the reference deviation value rd according to the following formula: .

[0026] Among them, c i represents the i-th ratio in the comprehensive feature, is the number of sequences less than n in the feature sequence set.

[0027] The prediction output processor calculates the resource utilization mean au of the target task in the next period according to the following formula: .

[0028] Where v is the maximum resource utilization of the target task, and h(n) is the nth segment mean of the resource utilization change function in the prediction benchmark case.

[0029] Embodiment 2: This embodiment includes all the contents of Embodiment 1, and provides an intelligent computing resource scheduling and optimization system, including a monitoring and acquisition module, a scheduling and allocation module, an optimization and prediction module, and a task management module.

[0030] The monitoring and acquisition module is used to collect and monitor the operating data of computing resources, the scheduling and allocation module dynamically schedules computing resources based on task requirements, the optimization and prediction module predicts resource requirements based on historical data and optimizes resource allocation, and the task management module is used to manage user task information.

[0031] Combination Figure 2 The monitoring and acquisition module includes a hardware monitoring unit, a resource acquisition unit and a transmission control unit. The hardware monitoring unit is used to monitor the hardware status information of the computing node, the resource acquisition unit is used to collect the resource usage of the task, and the transmission control unit is used to control the external transmission of the monitoring and acquisition data.

[0032] Combination Figure 3The scheduling allocation module includes a resource allocation unit, a task scheduling unit and a strategy selection unit. The resource allocation unit is used to calculate the amount of resources allocated to each task. The task scheduling unit schedules the task to the corresponding node based on the allocated resource amount. The strategy selection unit is used to select the scheduling strategy applied by the task scheduling unit.

[0033] Combination Figure 4 The optimization prediction module includes a historical data storage unit, a resource prediction unit and a dynamic optimization unit. The historical data storage unit is used to store historical data of task resource utilization, the resource prediction unit is used to predict changes in task demand for resources, and the dynamic optimization unit optimizes task resources based on the prediction information.

[0034] Combination Figure 5 The task management module includes a user information management unit, a task information management unit and a task submission control unit. The user information management unit is used to store and manage user information, the task information management unit is used to store and manage task information, and the task submission control unit is used to control the process of submitting tasks to the scheduling allocation module.

[0035] The hardware monitoring unit includes a temperature monitoring processor, a power monitoring processor and a load monitoring processor. The temperature monitoring processor is used to monitor the temperature status of the resource hardware, the power monitoring processor is used to monitor the operating power of the resource hardware, and the load monitoring processor is used to monitor the load information of the resource hardware.

[0036] The resource acquisition unit includes a GPU usage acquisition processor, a memory usage acquisition processor and an information cache processor. The GPU usage acquisition processor is used to collect the usage information of the GPU by the task, the memory usage acquisition processor is used to collect the usage information of the memory by the task, and the information cache processor is used to organize the collected information into a function form and store it.

[0037] The transmission control unit includes an application receiving processor, a data selection processor and a data transmission processor. The application receiving processor is used to receive an access application for data. The data selection processor selects corresponding data content based on the access application. The data transmission processor is used to transmit the selected data to the outside.

[0038] The resource allocation unit includes a demand analysis processor, an allocation adjustment processor and a list output processor. The demand analysis processor is used to parse the task information to obtain the basic resource demand. The allocation adjustment processor adjusts the resource quantity in combination with the monitoring information. The list output processor is used to send the resource configuration list of the task to the task scheduling unit.

[0039] The task scheduling unit includes a task queue processor, a node selection processor and a scheduling execution processor. The task queue processor is used to manage the scheduling order of tasks. The node selection processor selects a scheduling node for the task based on a resource configuration list. The scheduling execution processor is used to schedule task data to the corresponding resource node.

[0040] The policy selection unit includes a policy scheme register, a policy adaptation processor and a change output processor. The policy scheme register is used to store scheduling policy information. The policy adaptation processor adapts the corresponding scheduling policy based on the operating environment. The change output processor is used to output the changed scheduling policy when the policy changes.

[0041] The historical data storage unit includes a task classification processor, a case data register and a data retrieval processor. The task classification processor is used to classify tasks. The case data register stores case data based on classification results. The data retrieval processor is used to retrieve case data.

[0042] The resource prediction unit includes a comprehensive feature extraction processor, an individual feature matching processor and a prediction output processor. The comprehensive feature extraction processor is used to extract the comprehensive features of resource utilization of tasks of the same type. The individual feature matching processor matches case data similar to the target task based on the individual features. The prediction output processor predicts the subsequent resource requirements of the target task based on the comprehensive features and the individual features.

[0043] The dynamic optimization unit includes a summary analysis processor, a dynamic optimization processor and an adjustment output processor. The summary analysis processor is used to summarize and analyze the prediction information of all tasks, the dynamic adjustment processor is used to optimize the resource allocation of the target task, and the adjustment output processor is used to output the adjustment information to the corresponding node.

[0044] The user information management unit includes a basic information register, a task statistics processor and an interactive editing processor. The basic information register is used to store the user's basic information, the task statistics processor is used to count the number of calls to various tasks by each user, and the interactive editing processor is used for the user to edit the basic information and call information.

[0045] The task information management unit includes a task framework processor, an information verification processor and a task generation processor. The task framework processor is used to provide various types of task data frameworks, the information verification processor is used to verify the integrity of task information based on the corresponding framework, and the task generation processor is used to generate executable tasks and perform encoding processing.

[0046] The task submission control unit includes a task receiving processor, a coding registration processor and a task output processor. The task receiving processor is used to receive the encoded executable task, the coding registration processor is used to register the coding information of the executable task, and the task output processor is used to send the executable task to the scheduling allocation module.

[0047] The comprehensive feature extraction processor divides the resource utilization change function in the case into m segments on average, and calculates the mean ratio of the latter segment to the former segment respectively, and obtains a sequence consisting of m-1 ratios. The comprehensive feature extraction processor calculates the standard deviation of the ratios at each rank of the same type of task cases, and screens out the rank with a standard deviation less than a threshold. The screened rank constitutes the feature rank, and the feature rank and the corresponding ratio at the feature rank together constitute the comprehensive feature of this type of task.

[0048] The personality feature matching processor divides the resource utilization rate change function into n segments on average based on the current progress of the target task, and the value of n is calculated according to the following formula: .

[0049] in, The current progress of the target task.

[0050] The personality feature matching processor calculates the mean ratio of the latter section to the former section respectively, and obtains a sequence consisting of n-1 ratios. The personality feature matching processor calculates the personality difference index Sr between the target task and the case task according to the following formula: .

[0051] Among them, Ch is the feature sequence set corresponding to the task type, a i represents the i-th ratio of the target task, b i is the i-th ratio representing the case task.

[0052] The personality feature matching processor uses the case task with the smallest personality difference index as a matching prediction benchmark case.

[0053] The prediction output processor calculates the reference deviation value rd according to the following formula: .

[0054] Among them, c i represents the i-th ratio in the comprehensive feature, is the number of sequences less than n in the feature sequence set.

[0055] The prediction output processor calculates the resource utilization mean au of the target task in the next period according to the following formula: .

[0056] Where v is the maximum resource utilization of the target task, and h(n) is the nth segment mean of the resource utilization change function in the prediction benchmark case.

[0057] Each value in the resource utilization change function represents the ratio of the resource utilization at the corresponding time point to the maximum resource utilization.

[0058] The summary analysis processor classifies and summarizes the predicted data according to the node where the task is located, and calculates the incremental value between the total predicted resource utilization and the current resource allocation of each node. When the incremental value is less than the idle resource amount of the node, it is determined as normal optimization; when the incremental value is greater than the idle resource amount of the node, it is determined as reduction optimization.

[0059] During normal optimization, the dynamic adjustment processor directly calculates the change value between the predicted resource utilization of the target task and the current resource allocation as the adjustment output value. When the dynamic adjustment processor is performing reduction optimization, the actual adjustment value is calculated according to the following formula: .

[0060] Among them, an is the current resource allocation of the target task, and uV is the idle resource amount of the node.

[0061] The i appearing in the above text is an ordinal number used to indicate a sequence number.

[0062] Some code information of this system is as follows: class HardwareMonitor: def get_status(self): # Get computing node hardware status information (such as temperature, power, load, etc.) pass class ResourceCollector: def collect_usage(self): # Collect the usage of computing resources (such as CPU, memory, bandwidth, etc.) by the task pass class TransmissionController: def control_transmission(self, hardware_status, resource_usage): # Control the transmission of monitoring and collection data pass class ResourceAllocator: def allocate_resources(self, task, strategy): # Allocate resources based on task requirements and strategies pass class TaskScheduler: def schedule_task(self, task, resources): # Assign tasks to appropriate nodes pass class StrategySelector: def select_strategy(self, task): # Choose an appropriate scheduling strategy (such as priority, load balancing, etc.) pass class HistoricalDataStorage: def retrieve_data(self): # Get resource utilization data from the historical database pass class ResourcePredictor: def predict_resource_needs(self, history): # Predict resource requirements based on historical data pass class DynamicOptimizer: def optimize(self, predictions): # Dynamically optimize resource allocation strategy pass class TaskManagementModule: def __init__(self): self.user_manager = UserInfoManager() self.task_manager = TaskInfoManager() self.task_submission_controller = TaskSubmissionController() def manage_tasks(self, user, task): self.user_manager.store_user_info(user) self.task_manager.store_task_info(task) self.task_submission_controller.submit_task(task) class UserInfoManager: def store_user_info(self, user): #Store user information pass class TaskInfoManager: def store_task_info(self, task): #Store task information pass class TaskSubmissionController: def submit_task(self, task): # Control task submission to the scheduling allocation module Pass.

[0063] Now we use 10 task samples to test and get the completion time comparison chart of the experimental group of this system and the control group of the ordinary system, as shown in the figure below: Figure 6 shown.

[0064] The contents disclosed above are only preferred feasible embodiments of the present invention, and do not limit the protection scope of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated as technology develops.

Claims

1. An intelligent computing resource scheduling and optimization system, characterized in that: It includes monitoring and acquisition module, scheduling and allocation module, optimization and prediction module and task management module; The monitoring and collection module is used to collect and monitor the operation data of computing resources, the scheduling and allocation module dynamically schedules computing resources based on task requirements, the optimization and prediction module predicts resource requirements based on historical data and optimizes resource allocation, and the task management module is used to manage user task information; The monitoring and acquisition module includes a hardware monitoring unit, a resource acquisition unit and a transmission control unit. The hardware monitoring unit is used to monitor the hardware status information of the computing node, the resource acquisition unit is used to collect the usage of resources by the task, and the transmission control unit is used to control the external transmission of the monitoring and acquisition data; The scheduling and allocation module includes a resource allocation unit, a task scheduling unit and a strategy selection unit, wherein the resource allocation unit is used to calculate the amount of resources allocated to each task, the task scheduling unit schedules the task to the corresponding node based on the allocated amount of resources, and the strategy selection unit is used to select the scheduling strategy applied by the task scheduling unit; The optimization prediction module includes a historical data storage unit, a resource prediction unit and a dynamic optimization unit. The historical data storage unit is used to store historical data of resource utilization of tasks. The resource prediction unit is used to predict changes in the demand for resources of tasks. The dynamic optimization unit optimizes the task resources based on the prediction information. The task management module includes a user information management unit, a task information management unit and a task submission control unit. The user information management unit is used to store and manage user information, the task information management unit is used to store and manage task information, and the task submission control unit is used to control the process of submitting tasks to the scheduling allocation module.

2. The intelligent computing resource scheduling and optimization system according to claim 1, characterized in that: The resource prediction unit includes a comprehensive feature extraction processor, an individual feature matching processor and a prediction output processor. The comprehensive feature extraction processor is used to extract the comprehensive features of resource utilization of tasks of the same type. The individual feature matching processor matches case data similar to the target task based on the individual features. The prediction output processor predicts the subsequent resource requirements of the target task based on the comprehensive features and the individual features.

3. The intelligent computing resource scheduling and optimization system according to claim 2, characterized in that: The dynamic optimization unit includes a summary analysis processor, a dynamic optimization processor and an adjustment output processor. The summary analysis processor is used to summarize and analyze the prediction information of all tasks, the dynamic adjustment processor is used to optimize the resource allocation of the target task, and the adjustment output processor is used to output the adjustment information to the corresponding node.

4. The intelligent computing resource scheduling and optimization system according to claim 3, characterized in that: The comprehensive feature extraction processor divides the resource utilization rate change function in the case into m segments on average, and respectively calculates the mean ratio of the latter segment to the former segment to obtain a sequence consisting of m-1 ratios. The comprehensive feature extraction processor calculates the standard deviation of the ratios at each rank of the same type of task case, and selects the rank with a standard deviation less than a threshold value. The selected rank constitutes the feature rank, and the feature rank and the corresponding ratio at the feature rank together constitute the comprehensive feature of the task of this type. The personality feature matching processor divides the resource utilization rate change function into n segments on average based on the current progress of the target task, and the value of n is calculated according to the following formula: ； in, is the current progress of the target task; The personality feature matching processor calculates the mean ratio of the latter section to the former section respectively, and obtains a sequence consisting of n-1 ratios. The personality feature matching processor calculates the personality difference index Sr between the target task and the case task according to the following formula: ； Among them, Ch is the feature sequence set corresponding to the task type, a i represents the i-th ratio of the target task, b i is the i-th ratio representing the case task; The personality feature matching processor uses the case task with the smallest personality difference index as a matching prediction benchmark case.

5. The intelligent computing resource scheduling and optimization system according to claim 4, characterized in that: The prediction output processor calculates the reference deviation value rd according to the following formula: ； Among them, c i represents the i-th ratio in the comprehensive feature, is the number of sequences less than n in the feature sequence set; The prediction output processor calculates the resource utilization mean au of the target task in the next period according to the following formula: ； Where v is the maximum resource utilization of the target task, and h(n) is the nth segment mean of the resource utilization change function in the prediction benchmark case.

Images