An artificial intelligence-based intelligent computing center resource management system

Abstract

The application relates to the technical field of resource management, in particular to an intelligent calculation center resource management system based on artificial intelligence, which comprises a data processing module, a task scheduling module and a control module; the data processing module comprises a collection unit used for collecting resource information of the intelligent calculation center to output resource data and a preprocessing unit connected with the collection unit and used for preprocessing the resource data to output resource correlation data; the task scheduling module is connected with the data processing module and comprises a resource state monitoring unit used for monitoring the resource state of the intelligent calculation center to output task state data and a task scheduling unit connected with the resource state monitoring unit and used for scheduling resource tasks to resource nodes according to the task state data; the control module is connected with the data processing module and the task scheduling module respectively and is used for determining the random scheduling proportion of the resource tasks according to the overload misjudgment rate of the resource nodes. The application improves the management stability of the resources of the intelligent calculation center.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resource management, in particular to an intelligent calculation center resource management system based on artificial intelligence. BACKGROUND

[0002] In the prior art, the intelligent calculation center resource management system based on artificial intelligence is a system that uses artificial intelligence technology to efficiently manage and optimally allocate various resources of an intelligent calculation center. It can monitor resource status in real time, predict resource demand, intelligently schedule and allocate resources to improve the overall performance, resource utilization rate and service quality of the intelligent calculation center, reduce operating costs, and monitor the running state of tasks in real time during task execution, including task progress, resource usage, whether errors occur, etc. If a task is abnormal, the system will promptly issue an alarm and handle it according to the preset strategy, such as automatic retry, re-allocation of resources or termination of the task, etc.

[0003] Chinese Patent Publication No. CN117472587B discloses an AI intelligent calculation center resource scheduling system, which includes a task identification module, a resource demand prediction module, a dynamic resource allocation module, a cloud-edge collaborative scheduling module, a multi-objective optimization decision module, a scheduling strategy adaptive module, a performance analysis and monitoring module, and a scheduling result feedback module. In the present application, support vector machines and clustering analysis algorithms improve task identification and classification accuracy, time series and recurrent neural networks combined with resource demand prediction reduce waste, dynamic resource allocation applies heuristic algorithms and linear programming to improve efficiency, cloud-edge collaborative scheduling uses MapReduce technology to speed up data processing, genetic algorithms balance efficiency, energy and cost in multi-objective optimization decision making, and adaptive scheduling strategies apply machine learning.

[0004] As can be seen, the prior art has the following problems: As a result of long-term use, the software resources such as operating system, software framework, library file, etc. may become obsolete or incompatible with new tasks, and as tasks are continuously scheduled and executed, system resources may become fragmented, which will reduce resource utilization, make it difficult to find continuous available resources when allocating tasks, affect the efficiency and effectiveness of resource scheduling, and result in insufficient management stability of the resources of the intelligent calculation center. SUMMARY

[0005] Therefore, the present application provides an intelligent calculation center resource management system based on artificial intelligence to overcome the problem that in the prior art, as a result of long-term use, the software resources such as operating system, software framework, library file, etc. may become obsolete or incompatible with new tasks, and as tasks are continuously scheduled and executed, system resources may become fragmented, which will reduce resource utilization, make it difficult to find continuous available resources when allocating tasks, affect the efficiency and effectiveness of resource scheduling, and result in insufficient management stability of the resources of the intelligent calculation center.

[0006] To achieve the above object, the application provides an intelligent calculation center resource management system based on artificial intelligence, comprising:

[0007] A data processing module, comprising a collection unit configured to collect resource information of the intelligent calculation center to output resource data, a preprocessing unit connected to the collection unit and configured to preprocess the resource data to output resource correlation data, and a model training unit connected to the preprocessing unit and configured to train a mapping relationship according to the resource correlation data to output an artificial intelligence model;

[0008] A task scheduling module connected to the data processing module, comprising a resource state monitoring unit configured to monitor a resource state of the intelligent calculation center to output task state data, and a task scheduling unit connected to the resource state monitoring unit and configured to schedule resource tasks to resource nodes according to the task state data;

[0009] A control module connected to the data processing module and the task scheduling module, configured to determine a random scheduling proportion of the resource tasks according to an overload misjudgment rate of the resource nodes, or to determine a scheduling interval of the resource tasks according to a lag time length of the task state data, and to determine an update frequency of the task state data according to a conflict frequency of the resource task scheduling.

[0010] Further, the control module is configured to determine whether the management stability of the intelligent calculation center resources meets the requirements according to the overload misjudgment rate of the resource nodes, and if the overload misjudgment rate of the resource nodes is greater than a preset first misjudgment rate, it is determined that the management stability of the intelligent calculation center resources does not meet the requirements.

[0011] Further, the control module is configured to preliminarily determine that the scheduling accuracy of the intelligent calculation center resources does not meet the requirements when the overload misjudgment rate of the resource nodes is greater than the preset first misjudgment rate and less than or equal to a preset second misjudgment rate, and to determine whether the scheduling accuracy of the intelligent calculation center resources meets the requirements according to the lag time length of the task state data.

[0012] Further, the control module is configured to increase the random scheduling proportion of the resource tasks when the overload misjudgment rate of the resource nodes is greater than the preset second misjudgment rate.

[0013] Wherein, the increase range of the random scheduling proportion of the resource tasks is determined by the difference between the overload misjudgment rate of the resource nodes and the preset second misjudgment rate.

[0014] Further, the control module is configured to determine whether the scheduling accuracy of the intelligent calculation center resources meets the requirements according to the lag time length of the task state data, and if the lag time length of the task state data is greater than a preset first lag time length, it is determined that the scheduling accuracy of the intelligent calculation center resources does not meet the requirements.

[0015] Further, the control module is configured to decrease the scheduling interval of the resource task when the lag time of the task state data is greater than the preset first lag time and less than or equal to a preset second lag time.

[0016] Further, the control module is configured to preliminarily determine that the scheduling adaptability of the resource of the intelligent computing center does not meet the requirement when the lag time of the task state data is greater than the preset second lag time, and determine whether the scheduling adaptability of the resource of the intelligent computing center meets the requirement according to the conflict frequency of the resource task scheduling.

[0017] Further, the decrease range of the scheduling interval of the resource task is determined by the difference between the lag time of the task state data and the preset first lag time.

[0018] Further, the control module is configured to determine whether the scheduling adaptability of the resource of the intelligent computing center meets the requirement according to the conflict frequency of the resource task scheduling, and determine that the scheduling adaptability of the resource of the intelligent computing center does not meet the requirement and increase the update frequency of the task state data when the conflict frequency of the resource task scheduling is greater than a preset conflict frequency.

[0019] Further, the increase range of the update frequency of the task state data is determined by the difference between the conflict frequency of the resource task scheduling and the preset conflict frequency.

[0020] Compared with the prior art, the beneficial effects of the present application are that the system adjusts the random scheduling proportion of resource tasks according to the overload misjudgment rate of the resource nodes, increases the random scheduling proportion of resource tasks, reduces the dependence on outdated load data, randomly allocates tasks to other nodes that may not be overloaded, reduces the impact of misjudgment, adjusts the scheduling interval of resource tasks according to the lag length of task state data, reduces the scheduling interval of resource tasks, more frequently adjusts resources based on the latest collected state data, rather than relying on outdated historical prediction results, avoids the continuous deterioration of the overload problem caused by long-interval scheduling, adjusts the update frequency of task state data according to the conflict frequency of resource task scheduling, increases the update frequency of task state data, frequently acquires the latest state of the task, enables the model to more accurately track the rapid fluctuations of the load in the time dimension, timely feedbacks the task state change, helps the model dynamically adjust the prediction window, and avoids the pattern matching deviation caused by outdated state information.

[0021] Further, the system adjusts the random scheduling proportion of resource tasks by setting the preset first misjudgment rate and the preset second misjudgment rate, increases the random scheduling proportion of resource tasks, reduces the dependence on outdated load data, randomly allocates tasks to other nodes that may not be overloaded, reduces the impact of misjudgment, and improves the management stability of the intelligent computing center resources.

[0022] Further, the system adjusts the scheduling interval of resource tasks by setting the preset first lag length and the preset second lag length, reduces the scheduling interval of resource tasks, more frequently adjusts resources based on the latest collected state data, rather than relying on outdated historical prediction results, avoids the continuous deterioration of the overload problem caused by long-interval scheduling, and further improves the management stability of the intelligent computing center resources.

[0023] Further, the system of the present application adjusts the update frequency of the task state data by a preset update frequency. Since the resource occupation of different types of tasks is superimposed in time and space, forming a complex composite load mode, the load prediction model is difficult to capture the nonlinear and multimodal composite load characteristics, resulting in a significant increase in pattern matching error during model training. By increasing the update frequency of the task state data, the latest state of the task can be obtained at a high frequency, so that the model can more accurately track the rapid fluctuations of the load in the time dimension and timely feedback the changes in the task state, which can help the model dynamically adjust the prediction window, avoid the pattern matching deviation caused by outdated state information, and further improve the management stability of the intelligent computing center resources. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 is a whole structure block diagram of the intelligent computing center resource management system based on artificial intelligence of the embodiment of the present application;

[0025] Figure 2 is a logic flow chart of the process of determining the random scheduling ratio of the resource task of the intelligent computing center resource management system based on artificial intelligence of the embodiment of the present application;

[0026] Figure 3 is a logic flow chart of the process of determining the scheduling interval of the resource task of the intelligent computing center resource management system based on artificial intelligence of the embodiment of the present application;

[0027] Figure 4 is a logic flow chart of the process of determining the update frequency of the task state data of the intelligent computing center resource management system based on artificial intelligence of the embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the purpose and advantages of the present application more clear and explicit, the present application will be further described below in combination with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0029] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application, and are not intended to limit the protection scope of the present application.

[0030] Please refer to Figure 1 , Figure 2 , Figure 3 and Figure 4As shown, they are the overall structure block diagram of the artificial intelligence-based intelligent computing center resource management system, the logic flow chart of the process of determining the random scheduling ratio of resource tasks, the logic flow chart of the process of determining the scheduling interval of resource tasks, and the logic flow chart of the process of determining the update frequency of task state data, respectively. The artificial intelligence-based intelligent computing center resource management system comprises:

[0031] The data processing module comprises a collection unit for collecting resource information of the intelligent computing center to output resource data, a preprocessing unit connected with the collection unit for preprocessing the resource data to output resource correlation data, and a model training unit connected with the preprocessing unit for training a mapping relationship according to the resource correlation data to output an artificial intelligence model.

[0032] The task scheduling module is connected with the data processing module and comprises a resource state monitoring unit for monitoring the resource state of the intelligent computing center to output task state data, and a task scheduling unit connected with the resource state monitoring unit for scheduling resource tasks to resource nodes according to the task state data.

[0033] The control module is connected with the data processing module and the task scheduling module, respectively, for determining the random scheduling ratio of resource tasks according to the overload misjudgment rate of the resource nodes, or for determining the scheduling interval of resource tasks according to the lag length of the task state data, and for determining the update frequency of the task state data according to the conflict frequency of resource task scheduling.

[0034] Specifically, the artificial intelligence-based intelligent computing center resource management system further comprises a storage module connected with the data processing module and the task scheduling module, respectively, for storing the resource information, the resource data, the resource correlation data, the artificial intelligence model, the resource state, the task state data, the resource tasks, and the resource nodes.

[0035] Specifically, the resource information includes the type of storage medium, the storage location of resource data, and the number of running resource tasks.

[0036] Specifically, the resource data includes CPU cache size, available memory capacity, and display bandwidth.

[0037] Specifically, preprocessing includes removing noise, processing missing values, and normalization processing.

[0038] Specifically, the resource correlation data includes CPU cache size after removing noise, display bandwidth after processing missing values, and available memory capacity after normalization processing.

[0039] Specifically, the artificial intelligence model includes a linear regression model, a reinforcement learning model, and a decision tree.

[0040] Specifically, the resource state includes a CPU state, a system memory capacity state, and a network transmission delay duration.

[0041] Specifically, the task state data includes a task name, an execution progress, and an execution node.

[0042] Specifically, the resource task includes a resource state, a task demand, and real-time monitoring.

[0043] Specifically, the resource node includes a computing node, a storage node, and a network node.

[0044] In implementation, the system adjusts the random scheduling proportion of the resource task according to the overload misjudgment rate of the resource node. Since the node load data collection period is too long, it cannot capture the second-level burst load, which leads to misjudgment of the node state by the scheduling system, and the new task is allocated to the actual overloaded node. By increasing the random scheduling proportion of the resource task, the dependence on outdated load data can be reduced, the task can be randomly allocated to other nodes that may not be overloaded, the impact of misjudgment can be reduced, the scheduling interval of the resource task is adjusted according to the lag length of the task state data. Since the collection frequency of the task state is lower than the load change frequency, the historical state information input by the model is invalid, which leads to deviation of the prediction from the actual situation, thereby increasing the load prediction error. By reducing the scheduling interval of the resource task, resource adjustment can be performed more frequently based on the latest collected state data, rather than relying on outdated historical prediction results, thereby avoiding continuous deterioration of the overload problem caused by long-interval scheduling. The update frequency of the task state data is adjusted according to the conflict frequency of the resource task scheduling. Since the resource occupation of different types of tasks is superimposed in time and space, a complex composite load mode is formed, and the load prediction model is difficult to capture this nonlinear and multi-modal composite load characteristic, which significantly increases the pattern matching error during model training. By increasing the update frequency of the task state data, the latest state of the task can be obtained at a high frequency, so that the model can more accurately track the rapid fluctuations of the load in the time dimension and timely feedback the task state change, which can help the model dynamically adjust the prediction window and avoid pattern matching deviation caused by outdated state information.

[0045] Specifically, the control module is used to determine whether the management stability of the intelligent computing center resource meets the requirements according to the overload misjudgment rate of the resource node. If the overload misjudgment rate of the resource node is greater than a preset first misjudgment rate, it is determined that the management stability of the intelligent computing center resource does not meet the requirements.

[0046] Specifically, the control module is configured to preliminarily determine that the scheduling accuracy of the intelligent computing center resource does not meet the requirement when the overload misjudgment rate of the resource node is greater than the preset first misjudgment rate and less than or equal to a preset second misjudgment rate, and determine whether the scheduling accuracy of the intelligent computing center resource meets the requirement according to the lag length of the task state data.

[0047] It can be understood that the three intervals divided by the preset first misjudgment rate and the preset second misjudgment rate correspond to three situations respectively.

[0048] The first interval is that the overload misjudgment rate of the resource node is less than or equal to the preset first misjudgment rate, and the corresponding situation is that it is determined that the management stability of the intelligent computing center resource meets the requirement.

[0049] The second interval is that the overload misjudgment rate of the resource node is greater than the preset first misjudgment rate and less than or equal to the preset second misjudgment rate, and the corresponding situation is that the historical state information input by the model is invalid due to the fact that the collection frequency of the task state is lower than the load change frequency, which leads to the deviation of the prediction from the actual situation, thereby increasing the load prediction error.

[0050] The third interval is that the overload misjudgment rate of the resource node is greater than the preset second misjudgment rate, and the corresponding situation is that the node load data collection period is too long to capture the second-level burst load, which leads to the misjudgment of the node state by the scheduling system and the allocation of new tasks to the actually overloaded node.

[0051] In implementation, the preset first misjudgment rate is generally selected in the range of [1%, 3%], and the preset second misjudgment rate is generally selected in the range of [4%, 6%].

[0052] Preferably, the preferred embodiment of the preset first misjudgment rate is 2%, and the preferred embodiment of the preset second misjudgment rate is 5%.

[0053] Specifically, the overload misjudgment rate of the resource node is the ratio of the number of data nodes misjudged as overloaded to the total number of resource data nodes.

[0054] Specifically, overload refers to the working load borne by the resource node exceeding the limit state that it can bear.

[0055] Specifically, the control module is configured to increase the random scheduling proportion of resource tasks when the overload misjudgment rate of the resource node is greater than the preset second misjudgment rate.

[0056] The increase range of the random scheduling proportion of resource tasks is determined by the difference between the overload misjudgment rate of the resource node and the preset second misjudgment rate.

[0057] Specifically, the random scheduling proportion of resource tasks refers to the ratio of the number of times of randomly selecting resource nodes for task scheduling to the total number of scheduling times.

[0058] Specifically, when the difference between the overload misjudgment rate of the resource node and the preset second misjudgment rate is within 1%, the random scheduling ratio of the resource task is increased to 1.1 times of the original; when the difference between the overload misjudgment rate of the resource node and the preset second misjudgment rate exceeds 1%, the random scheduling ratio of the resource task is increased by 2% for each 0.5% exceeding 1.1 times of the original, for example, when the difference between the overload misjudgment rate of the resource node and the preset second misjudgment rate is 2%, the current random scheduling ratio of the resource task is 30%, and the increased random scheduling ratio of the resource task is 30x1.1+2x2=37%.

[0059] In implementation, the system of the present application adjusts the random scheduling ratio of the resource task by setting the preset first misjudgment rate and the preset second misjudgment rate. Since the node load data collection period is too long, it cannot capture the second-level burst load, resulting in misjudgment of the scheduling system node state and allocation of new tasks to the actual overloaded node. By increasing the random scheduling ratio of the resource task, the dependence on outdated load data can be reduced, the task can be randomly allocated to other possible non-overloaded nodes, the impact of misjudgment can be reduced, and the management stability of the intelligent computing center resource is improved.

[0060] Specifically, the control module is used to determine whether the scheduling accuracy of the intelligent computing center resource meets the requirements according to the lag length of the task state data. If the lag length of the task state data is greater than the preset first lag length, it is determined that the scheduling accuracy of the intelligent computing center resource does not meet the requirements.

[0061] Specifically, the control module is used to decrease the scheduling interval of the resource task when the lag length of the task state data is greater than the preset first lag length and less than or equal to the preset second lag length.

[0062] Specifically, the control module is used to preliminarily determine that the scheduling adaptability of the intelligent computing center resource does not meet the requirements when the lag length of the task state data is greater than the preset second lag length, and to determine whether the scheduling adaptability of the intelligent computing center resource meets the requirements according to the conflict frequency of the resource task scheduling.

[0063] Specifically, the conflict frequency of the resource task scheduling refers to the frequency of conflict of multiple resource tasks in the task scheduling process.

[0064] Specifically, the conflict refers to the situation that the tasks cannot be normally scheduled due to insufficient resources, resource failure, and imperfect scheduling algorithm in the resource task scheduling process.

[0065] Specifically, the lag length of the task state data refers to a time interval between a time when an actual state of the task actually changes and a target time of the task state data in the system.

[0066] Specifically, the scheduling interval of the resource task is a time interval between two adjacent scheduling times of the resource task.

[0067] It can be understood that the three intervals divided by the preset first lag length and the preset second lag length correspond to three situations respectively.

[0068] The first interval is that the lag length of the task state data is less than or equal to the preset first lag length, and the corresponding situation is that the scheduling accuracy of the intelligent computing center resource is determined to meet the requirements.

[0069] The second interval is that the lag length of the task state data is greater than the preset first lag length and less than or equal to the preset second packet loss rate, and the corresponding situation is that the historical state information input by the model is invalid due to the fact that the collection frequency of the task state is lower than the load change frequency, which leads to deviation of the prediction from the actual situation, thereby increasing the load prediction error.

[0070] The third interval is that the lag length of the task state data is greater than the preset second lag length, and the corresponding situation is that the resource occupation of different types of tasks is superimposed in time and space to form a complex composite load mode, and the load prediction model is difficult to capture the nonlinear and multi-modal composite load characteristics, thereby significantly increasing the pattern matching error during model training.

[0071] In implementation, the preset first lag length is generally selected in the range of [90 ms, 95 ms], and the preset second lag length is generally selected in the range of [96 ms, 100 ms].

[0072] Preferably, the preferred embodiment of the preset first lag length is 93 ms, and the preferred embodiment of the preset second lag length is 98 ms.

[0073] Specifically, the reduction range of the scheduling interval of the resource task is determined by the difference between the lag length of the task state data and the preset first lag length.

[0074] Specifically, when the difference between the lag length of the task state data and the preset first lag length is within 2 ms, the scheduling interval of the resource task is reduced to 0.9 times of the original scheduling interval, and when the difference between the lag length of the task state data and the preset first lag length exceeds 2 ms, the scheduling interval of the resource task is reduced by 1 s for each 1 ms in addition to the reduction to 0.9 times of the original scheduling interval, for example, the difference between the lag length of the task state data and the preset first lag length is 4 ms, the current scheduling interval of the resource task is 20 s, and the reduced scheduling interval of the resource task is 20*0.9-2*1=16 s.

[0075] In implementation, the system described in the application adjusts the scheduling interval of resource tasks by setting a preset first hysteresis duration and a preset second hysteresis duration. Since the collection frequency of task states is lower than the frequency of load changes, the historical state information input by the model is invalid, which leads to deviation of prediction from reality, thereby increasing the load prediction error. By reducing the scheduling interval of resource tasks, resource adjustment can be performed more frequently based on the latest collected state data rather than relying on outdated historical prediction results, thereby avoiding continuous deterioration of the overload problem caused by long-interval scheduling and further improving the management stability of the resource of the intelligent computing center.

[0076] Specifically, the control module is configured to determine whether the scheduling adaptability of the resource of the intelligent computing center meets the requirements according to the conflict frequency of resource task scheduling. If the conflict frequency of resource task scheduling is greater than a preset conflict frequency, it is determined that the scheduling adaptability of the resource of the intelligent computing center does not meet the requirements, and the update frequency of task state data is increased.

[0077] It can be understood that the two intervals divided by the preset conflict frequency correspond to two situations respectively.

[0078] The first interval is that the conflict frequency of resource task scheduling is less than or equal to the preset conflict frequency, and the corresponding situation is that it is determined that the scheduling adaptability of the resource of the intelligent computing center meets the requirements.

[0079] The second interval is that the conflict frequency of resource task scheduling is greater than the preset conflict frequency, and the corresponding situation is that the resource occupation of different types of tasks is superimposed in time and space to form a complex composite load mode, and the load prediction model is difficult to capture the nonlinear and multimodal composite load characteristics, thereby significantly increasing the pattern matching error during model training.

[0080] In implementation, the preset conflict frequency is generally selected in the range of [6 times / minute, 10 times / minute].

[0081] Preferably, the preferred embodiment of the preset conflict frequency is 8 times / minute.

[0082] Specifically, the increase amplitude of the update frequency of the task state data is determined by the difference between the conflict frequency of resource task scheduling and the preset conflict frequency.

[0083] Specifically, when the difference between the conflict frequency of resource task scheduling and the preset conflict frequency is within 2 times / min, the update frequency of the task state data is increased to 1.2 times of the original; when the difference between the conflict frequency of resource task scheduling and the preset conflict frequency exceeds 2 times / min, on the basis of increasing to 1.2 times of the original, the update frequency of the task state data is increased by 5 Hz for every 1 time / min, for example, when the difference between the conflict frequency of resource task scheduling and the preset conflict frequency is 3 times / min, the update frequency of the current task state data is 100 Hz, and the increased update frequency of the task state data is 100*1.2+1*5=125 Hz.

[0084] In implementation, the system described in the application adjusts the update frequency of the task state data by the preset update frequency. Since the resource occupation of different types of tasks is superimposed in time and space, forming a complex composite load mode, the load prediction model is difficult to capture the nonlinear and multimodal composite load characteristics, resulting in a significant increase in pattern matching error during model training. By increasing the update frequency of the task state data, the latest state of the task can be obtained at a high frequency, so that the model can more accurately track the rapid fluctuations of the load in the time dimension and timely feedback the task state changes, which can help the model dynamically adjust the prediction window, avoid the pattern matching deviation caused by outdated state information, and further improve the management stability of the intelligent computing center resources.

[0085] So far, the technical solutions of the application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the application, and the technical solutions after the changes or replacements will fall within the protection scope of the application.

Claims

1. An artificial intelligence-based intelligent computing center resource management system, characterized by, The method comprises the following steps: a data processing module comprising a collection unit configured to collect resource information of the intelligent computing center to output resource data, a preprocessing unit connected to the collection unit and configured to preprocess the resource data to output resource correlation data, and a model training unit connected to the preprocessing unit and configured to train a mapping relationship based on the resource correlation data to output an artificial intelligence model; a task scheduling module connected to the data processing module, comprising a resource state monitoring unit configured to monitor a resource state of the intelligent computing center to output task state data, and a task scheduling unit connected to the resource state monitoring unit and configured to schedule resource tasks to resource nodes based on the task state data; a control module connected to the data processing module and the task scheduling module, configured to determine a random scheduling proportion of resource tasks based on an overload misjudgment rate of the resource nodes, to determine a scheduling interval of the resource tasks based on a lag time of the task state data, and to determine an update frequency of the task state data based on a conflict frequency of the resource task scheduling; the control module is configured to determine whether the management stability of the intelligent computing center resources meets the requirements based on the overload misjudgment rate of the resource nodes, and if the overload misjudgment rate of the resource nodes is greater than a preset first misjudgment rate, it is determined that the management stability of the intelligent computing center resources does not meet the requirements; the control module is configured to preliminarily determine that the scheduling accuracy of the intelligent computing center resources does not meet the requirements when the overload misjudgment rate of the resource nodes is greater than the preset first misjudgment rate and less than or equal to a preset second misjudgment rate, and to determine whether the scheduling accuracy of the intelligent computing center resources meets the requirements based on the lag time of the task state data; the control module is configured to increase the random scheduling proportion of the resource tasks when the overload misjudgment rate of the resource nodes is greater than the preset second misjudgment rate; wherein the increase range of the random scheduling proportion of the resource tasks is determined by the difference between the overload misjudgment rate of the resource nodes and the preset second misjudgment rate; the control module is configured to determine whether the scheduling accuracy of the intelligent computing center resources meets the requirements based on the lag time of the task state data, and if the lag time of the task state data is greater than a preset first lag time, it is determined that the scheduling accuracy of the intelligent computing center resources does not meet the requirements; the control module is configured to decrease the scheduling interval of the resource tasks when the lag time of the task state data is greater than the preset first lag time and less than or equal to a preset second lag time; the control module is configured to preliminarily determine that the scheduling adaptability of the intelligent computing center resources does not meet the requirements when the lag time of the task state data is greater than the preset second lag time, and to determine whether the scheduling adaptability of the intelligent computing center resources meets the requirements based on the conflict frequency of the resource task scheduling. 2.The AI-based intelligent computing center resource management system of claim 1, wherein, The decrease range of the scheduling interval of the resource tasks is determined by the difference between the lag time of the task state data and the preset first lag time. 3.The AI-based cognitive center resource management system of claim 2, wherein, The control module is used to determine whether the scheduling adaptability of the resource of the intelligent calculation center meets the requirement according to the conflict frequency of the resource task scheduling, and if the conflict frequency of the resource task scheduling is greater than the preset conflict frequency, it is determined that the scheduling adaptability of the resource of the intelligent calculation center does not meet the requirement, and the update frequency of the task state data is increased. 4.The AI-based intelligent computing center resource management system of claim 3, wherein, The increase range of the update frequency of the task state data is determined by the difference between the conflict frequency of the resource task scheduling and the preset conflict frequency.

Images