Computing power pool scheduling management system based on AI intelligent technology

Abstract

This invention discloses a computing power pooling scheduling and management system based on AI intelligent technology, relating to the field of computing power pooling scheduling and management. It solves the technical problem in existing technologies where it is impossible to determine the rules for the coordinated use of various types of resources within a computing power resource pool. The pooling layer manages various computing resources in the resource layer in a pooled manner. If the real-time resource configuration at the resource layer meets the computing power requirements of the actual usage scenario, the current resource configuration is pooled to construct a computing power resource pool. After constructing the computing power resource pool and the rules for coordinated use of resources, resources are orchestrated according to the usage scenario of the tasks to be executed in real time. Non-currently executed tasks are re-matched, alleviating the computing power execution pressure on the computing power pool, achieving optimal resource utilization, improving the execution efficiency of task computation, reducing computing time, and effectively solving the resource scheduling problem for sudden tasks.

Description

Technical Field

[0001] This invention relates to the field of computing power pooling scheduling and management, specifically to a computing power pooling scheduling and management system based on AI intelligent technology. Background Technology

[0002] AI intelligent technology, also known as artificial intelligence technology; computing power pooling scheduling and management system is an innovative technical architecture that aims to integrate and optimize various computing resources to improve resource utilization, reduce costs, and meet diverse business needs; with the development of computer hardware technology, various heterogeneous acceleration chips and devices are constantly emerging to meet the computing needs of various upper-layer applications.

[0003] However, in the existing technology, the resource configuration of the resource layer cannot be identified and updated during the computing power pooling stage, and the configuration screening cannot be performed, which reduces the efficiency of controlling the invalid occupation of the resource layer. In addition, resources cannot be recycled and reused, which reduces the operating efficiency of the computing power pool. Furthermore, it is impossible to determine the rules for the coordinated use of various types of resources within the computing power resource pool, and the inability to efficiently occupy resources reduces the feasibility of optimizing the scheduling of the computing power pool.

[0004] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to solve the problems mentioned above by proposing a computing power pooling scheduling and management system based on AI intelligent technology.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] The computing power pooling scheduling and management system based on AI intelligent technology has a system architecture consisting of a resource layer, a pooling layer, and a scheduling layer.

[0008] The resource layer consists of various computing resources. Resources within the resource layer are identified in real time, and the resource layer configuration is updated based on the identification results. After the construction is completed, the resource status is monitored during the continuous update phase of the resource layer. The health status of the resources is inferred based on real-time monitoring analysis. After status monitoring, resources are reclaimed based on the real-time computing task execution progress. At the same time, the resource layer configuration is optimized in real time based on changes in the computing task usage scenario.

[0009] The pooling layer pools and manages various computing resources in the resource layer. If the real-time resource configuration in the resource layer meets the computing power requirements of the actual use scenario, the current resource configuration is pooled to build a computing power resource pool. After building the computing power resource pool and the resource coordination and use criteria, resources are orchestrated according to the use scenario of the real-time tasks to be executed.

[0010] The scheduling layer is used to reasonably match tasks with computing resources based on task requirements and the real-time status of the computing resource pool; when the computing resource pool and the currently executing task are matched, the processing efficiency of the current computing resource pool for the executing task is evaluated by synchronous comparison; after confirming that there is no delay in synchronous comparison, the progress of the executing task matched by the computing resource pool is monitored.

[0011] In a preferred embodiment of the present invention, existing resources within the resource layer are obtained and real-time resource configuration is set, resources added in real time are marked as to be added to the configuration, and new optimization data and new risk data are obtained.

[0012] If the newly added optimization data does not exceed the parameter excess threshold, or if the newly added risk data exceeds the conflict value threshold, it will not be added to the current real-time resource configuration; if the newly added optimization data exceeds the parameter excess threshold, but the newly added risk data does not exceed the conflict value threshold, it will be added to the current real-time resource configuration.

[0013] The newly added optimized data and newly added risk data are the excess amount of computing power data processing parameters of the same type of resources corresponding to the real-time resource configuration and the configuration to be added, and the conflict value of the running parameters of the different types of cooperative resources corresponding to the real-time resource configuration and the configuration to be added.

[0014] In a preferred embodiment of the present invention, the fluctuation trend of the running parameters of any resource before and after executing a task in real-time resource configuration is collected within a set parameter range, and a parameter curve is constructed based on the values ​​of the running parameters before and after the task execution; if the parameter curve shows an upward trend at the time of task execution, and the downward slope of the running parameter curve continues to decrease after the task is completed, then hardware maintenance and performance testing are performed on the current resource; if the parameter curve does not show an upward trend at the time of task execution, or the downward slope of the running parameter curve does not continue to decrease after the task is completed, then task matching is performed on the current resource.

[0015] In a preferred embodiment of the present invention, when real-time resource configuration at the resource layer is used to coordinate task execution, a preset execution period for the task is obtained, and the output data volume corresponding to the data stream of the real-time task execution is compared. If the output data volume corresponding to the data stream decreases to or is lower than the minimum data stream threshold, the current moment is marked as the computing power shutdown moment.

[0016] Based on the comparison of the amount of data processed by the task at the time of computing power shutdown, if the amount of data processed is within the preset data range and the data processing flow of the task is completed, the current computing power shutdown time is marked as the resource reclamation time, and the resources matched with the current task at the resource layer are reclaimed and rematched with the task.

[0017] If the amount of data being processed is not within the preset range, or if the data processing flow of the task is completed, then a data processing performance test will be performed on the resources used in the current resource layer.

[0018] In a preferred embodiment of the present invention, the computing power requirement of the current execution task is obtained based on the execution task received by the resource layer, and the processing parameters are compared during the task execution stage. If the fluctuation range of the processing parameters exceeds the span threshold when facing different computing power requirements, and the fluctuation frequency of the processing parameters continues to increase with the increase of the execution task, it is inferred that the configuration computing power of the resource layer usage scenario needs to be upgraded. Based on the current real-time resource configuration as the type standard, the performance is upgraded on the basis of the original set performance parameters according to the type standard.

[0019] If the processing parameters continue to increase when facing different computing power requirements, and the duration of each fluctuation of the processing parameters continues to increase, it is inferred that the configuration type of the resource layer usage scenario needs to be changed. Taking the current real-time resource configuration as the starting point of the configuration standard, the resource type is adjusted and the overall computing power performance of the configuration is improved.

[0020] In a preferred embodiment of the present invention, if the real-time resource configuration at the resource layer meets the computing power requirements of the actual use scenario, the current resource configuration is pooled to construct a computing power resource pool; and the criteria for the coordinated use of various types of resources in the computing power resource pool are determined according to the task execution requirements of each use scenario.

[0021] In a preferred embodiment of the present invention, after constructing a computing power resource pool and resource coordination rules, resources are orchestrated according to the usage scenarios of real-time tasks to be executed. The usage rules corresponding to the execution procedures of real-time tasks to be executed are statistically analyzed and arranged according to the execution procedure order. After the arrangement is completed, the usage rules of the current coordination are statistically analyzed, and resources with overlapping usage rules are used uniformly, that is, resources are not used according to the execution procedure order, and the execution procedures required by the current usage scenario are executed synchronously according to the usage rules.

[0022] The completed data is stored and transmitted during the orchestration of the corresponding execution process to complete the current execution process. If there is no need to execute the same type of resource after the execution process is completed during the orchestration process, the currently configured resource is released and a new non-current execution task is matched.

[0023] In a preferred embodiment of the present invention, the scheduling process of the tasks executed by the real-time computing power resource pool is synchronously compared to obtain the frequency of increase in the delay between the completion time of the data processing using overlapping resources and the execution time of the corresponding execution process of the processed data. If the frequency of increase in delay exceeds the frequency threshold, it is inferred that the same processing using overlapping resources has a delay, and the computing power resource pool is scheduled to increase the number of resources adapted to the criteria for the current execution process or increase the processing time; if the frequency of increase in delay does not exceed the frequency threshold, it is inferred that the same processing using overlapping resources has no delay.

[0024] In a preferred embodiment of the present invention, the progress of the execution tasks matched by the computing power resource pool is monitored. The progress completion time of each execution task is analyzed to infer whether there is a delay. If there is a progress interruption in the task execution stage, no data stream is generated during the current task execution period. The resources corresponding to the execution process of the current execution period are tested for operation.

[0025] In a preferred embodiment of the present invention, if resources are running but cannot process the currently executing process, the arrangement of the currently executing process is adjusted to reduce the data processing of the current type of resources, and the resources that can complete the matching process in advance are used as the scheduling resources for the current executing process.

[0026] If there are idle resources, the order and progress of instruction sending for the current execution process will be controlled to ensure that the start time of the preset execution period of each execution progress is consistent with the time when the corresponding matching resource receives the execution instruction.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. In this invention, resources within the resource layer are identified in real time, and the resource layer configuration is updated based on the identification results. During the update phase, the resource performance is screened and optimized to avoid resource conflicts or resource incompatibility, which would cause a decrease in the resource data processing performance of the resource layer and occupy the storage space of the resource layer. The resource identification and update further improves the computing power of the resource layer.

[0029] Real-time monitoring and analysis are used to infer the health status of resources and avoid resource anomalies during data processing. These anomalies can reduce real-time computing power, causing data processing tasks to be suspended or delayed, and reducing the data processing performance of the computing power pool. Real-time resource status assessment is used to ensure the efficient execution of the current computing power.

[0030] Resource reclamation is performed based on the real-time computing task execution progress to avoid resource gaps in the resource layer during the current computing task, which would prevent the reasonable utilization of resources, reduce the scheduling of the resource layer, and even affect the efficiency of the computing pool in planning and processing tasks.

[0031] Based on actual needs, the configuration of the resource layer is updated while keeping the resource layer configuration type unchanged, so as to improve the computing power foundation of the resource layer and reasonably solve the computing power task requirements of the usage scenario, without the phenomenon of resource waste or resource performance excess.

[0032] 2. In this invention, the criteria for the coordinated use of various types of resources in the computing power resource pool are determined according to the task execution requirements of each usage scenario. For example, the criteria for the use of data containers and hardware acceleration devices added for data acquisition and storage are determined to complete task execution and improve execution efficiency without involving the execution standby of unnecessary resources, thereby improving the efficiency of reasonable coordination of resources in the entire computing power resource pool.

[0033] Once the execution process is completed during the orchestration process, if there is no need to execute the same type of resource, the currently configured resource will be released and re-matched with non-current execution tasks. This alleviates the computing power execution pressure on the computing power pool, achieves optimal resource utilization, improves the execution efficiency of task computation, reduces computing time, and effectively solves the resource scheduling problem for sudden tasks.

[0034] 3. In this invention, the computing power resource pool is matched with the currently executed task. The efficiency of the current computing power resource pool in processing the executed task is evaluated by synchronous comparison, which improves the efficiency of the executed task and facilitates timely resource scheduling when the resource allocation cannot meet the computing power demand, so as to facilitate the more efficient completion of the executed task. Attached Figure Description

[0035] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0036] Figure 1 This is a system architecture block diagram of the present invention;

[0037] Figure 2 This is a flowchart of Embodiment 1 of the present invention;

[0038] Figure 3 This is a flowchart of Embodiment 2 of the present invention;

[0039] Figure 4 This is a flowchart of Embodiment 3 of the present invention. Detailed Implementation

[0040] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0042] Please see Figure 1 As shown, the AI-based computing power pooling scheduling and management system has an architecture consisting of a resource layer, a pooling layer, and a scheduling layer. The resource layer comprises various computing resources, including physical servers, virtual machines, containers, GPU clusters, and FPGA hardware acceleration devices. The pooling layer abstracts and pools these computing resources to form a unified computing power resource pool. The scheduling layer allocates tasks to the most suitable computing resources based on user-submitted task requirements and the real-time status of the resource pool.

[0043] Example 1

[0044] Please see Figure 2 As shown, the operation steps of the resource layer are as follows:

[0045] S101. Resource identification and update within the resource layer: Real-time identification of resources within the resource layer and updating of resource layer configuration based on the identification results. During the update phase, screening and optimization are performed based on real-time resource performance to avoid resource conflicts or resource incompatibility, which could lead to a decrease in resource data processing performance and occupy of resource layer storage space. Resource identification and update further improves the computing power of the resource layer.

[0046] The system acquires existing resources within the resource layer and sets real-time resource configurations. Newly added resources are marked as configurations to be added. The system also acquires the excess processing parameters of the computing power data for the same type of resources corresponding to the real-time resource configuration and the configurations to be added. Processing parameters represent the speed of task data processing or the processing response time, and the excess indicates a positive proportionality between the processing parameters (i.e., larger parameters result in better performance). The excess processing parameters of the computing power data for the same type of resources corresponding to the real-time resource configuration and the configurations to be added are marked as newly added optimization data. The system also acquires conflicting values ​​of operating parameters for non-compatible resources corresponding to the real-time resource configuration and the configurations to be added. Conflicting operating parameters indicate that the real-time resource configuration controls hardware temperature, while the configurations to be added cause the hardware temperature to rise; the resulting temperature rise is considered conflicting data. Finally, the conflicting values ​​of operating parameters for non-compatible resources corresponding to the real-time resource configuration and the configurations to be added are marked as newly added risk data.

[0047] Furthermore, the newly added optimized data and newly added risk data are compared with the parameter excess threshold and conflict value threshold, respectively:

[0048] If the newly added optimized data does not exceed the parameter excess threshold, or the newly added risk data exceeds the conflict value threshold, it is inferred that the identification and detection of the data to be added to the configuration is abnormal and will not be added to the current real-time resource configuration.

[0049] If the newly added optimized data exceeds the parameter excess threshold, and the newly added risk data does not exceed the conflict value threshold, it is inferred that the identification and detection of the configuration to be added is normal, and it is added to the current real-time resource configuration, that is, the resource configuration is updated.

[0050] It should be explained that if the real-time task data increases instantaneously, and the newly added optimized data is a geometric multiple of the corresponding parameter overload threshold, but the newly added risk data is not a geometric multiple of the conflict value threshold, then it will be used as a temporary configuration.

[0051] S102. Resource status monitoring: During the continuous update phase of the resource layer, resource status is monitored. Based on real-time monitoring analysis, the health status of the resources is inferred to avoid resource layer anomalies during data processing, which could reduce real-time computing power and cause data processing tasks to be suspended or delayed, thus reducing the data processing performance of the computing power pool. Based on real-time resource status assessment, the efficiency of the execution of the current computing power is ensured.

[0052] The system monitors the status of real-time resource configurations, collecting data on the fluctuation trends of running parameters of any resource before and after task execution, within a set parameter range. Parameter curves are constructed based on the values ​​of these running parameters before and after task execution, where running parameters include CPU utilization, memory usage, network bandwidth, etc. If the parameter curves show an upward trend during task execution, and the slope of the curves continues to decrease after task completion, it is inferred that the corresponding resource status is at risk, and hardware maintenance and performance testing are performed on the current resource.

[0053] If the parameter curve does not show an upward trend during task execution, or if the downward slope of the running parameter curve does not continue to decrease after the task is completed, it is inferred that there is no risk in the resource status corresponding to the real-time resource configuration, and task matching is performed on the current resource.

[0054] S103. Resource allocation and reclamation: Resource reclamation is carried out according to the real-time computing task execution progress to avoid the resource allocation within the resource layer from generating computing power gaps during the current computing power task, which would prevent the reasonable utilization of resources, reduce the scheduling of the resource layer, and thus affect the efficiency of the computing power pool in planning and processing tasks.

[0055] When coordinating real-time resource configuration with task execution, the preset execution period of the task is obtained. The output data volume corresponding to the real-time task execution data stream is compared. If the output data volume drops to or falls below the minimum data stream threshold, the current moment is marked as the computing power shutdown moment. The processing data volume of the task is compared based on the computing power shutdown moment. If the processing data volume is within the preset range and the task's data processing flow is completed, the current computing power shutdown moment is marked as the resource reclamation moment. Resources matched to the currently executing task in the resource layer are reclaimed and re-matched with the task. If the processing data volume is not within the preset range, or the task's data processing flow is completed, a data processing performance test is performed on the resources used in the current resource layer to determine whether to increase the cached data volume during task execution. If so, resource repair is performed after the currently matched task is completed.

[0056] S104. Resource configuration planning: Real-time performance optimization of resource layer configuration based on changes in computing power task usage scenarios. On the basis of unchanged resource layer configuration type, configuration is updated according to actual needs to improve the computing power foundation of resource layer and can reasonably solve the computing power task needs of usage scenarios without resource waste or resource performance excess.

[0057] Based on the execution tasks received by the resource layer, the computing power requirement of the current execution task is obtained. During the task execution phase, the processing parameters are compared. If the fluctuation range of the processing parameters exceeds the range threshold when facing different computing power requirements, and the frequency of the fluctuation of the processing parameters continues to increase with the increase of execution tasks, it is inferred that the configuration computing power of the resource layer usage scenario needs to be upgraded. Based on the current real-time resource configuration as the type standard, the performance is upgraded on the basis of the original performance parameters. If the processing parameters continue to increase when facing different computing power requirements, and the single fluctuation time of the processing parameters continues to increase, it is inferred that the configuration type of the resource layer usage scenario needs to be changed. Based on the current real-time resource configuration as the configuration standard, the resource type is adjusted and the overall computing power performance is improved.

[0058] Example 2

[0059] Please see Figure 3 As shown, the operation steps of the pooling layer are as follows:

[0060] S201. Pool the resources in the resource layer. If the real-time resource configuration in the resource layer meets the computing power requirements of the actual use scenario, then pool the current resource configuration to build a computing power resource pool. Based on the task execution requirements of each use scenario, determine the rules for the coordinated use of various types of resources in the computing power resource pool, such as the rules for the use of data containers and hardware acceleration devices added for data acquisition and storage. Complete the task execution and improve the execution efficiency without involving the execution standby of unnecessary resources, thereby improving the efficiency of reasonable coordination of resources in the entire computing power resource pool.

[0061] S202. Based on the computing power pool requirements, after constructing the computing power resource pool and resource coordination rules, resource orchestration is performed according to the usage scenarios of real-time tasks to be executed. The usage rules corresponding to the execution procedures of real-time tasks to be executed are statistically analyzed and orchestrated according to the execution procedure order. After orchestration is completed, the current coordination usage rules are statistically analyzed, and resources with overlapping usage rules are used uniformly, that is, resources are not used according to the execution procedure order. According to the usage rules, the execution procedures required by the current usage scenario are executed synchronously, and the completed execution data is stored. Data transmission is performed when the corresponding execution procedure is orchestrated to complete the current execution procedure. During the orchestration process, if there is no execution requirement for the same type of resource after the execution procedure is completed, the currently configured resources are released and re-matched with non-current execution tasks, which alleviates the computing power execution pressure of the computing power pool, achieves optimal resource utilization, improves the execution efficiency of task computing, reduces computing time, and effectively solves the resource scheduling of sudden tasks.

[0062] Example 3

[0063] Please see Figure 4 As shown, the operation steps of the scheduling layer are as follows:

[0064] S301. Synchronous comparison of task requirements and resource status: The computing power resource pool is matched with the currently executed task. The synchronous comparison is used to infer the processing efficiency of the current computing power resource pool for the executed task, which improves the efficiency of the executed task and facilitates timely resource scheduling when the resource allocation cannot meet the computing power requirements, so as to facilitate the more efficient completion of the executed task.

[0065] The scheduling process of tasks executed by the real-time computing resource pool is synchronously compared to obtain the frequency of increase in the delay between the completion time of data processing using overlapping resources and the execution time of the corresponding execution process. If the frequency of increase in delay exceeds the increase frequency threshold, it is inferred that the same processing using overlapping resources has a delay, and the computing resource pool is rescheduled to increase the number of resources adapted to the current execution process or increase the processing time. If the frequency of increase in delay does not exceed the increase frequency threshold, it is inferred that the same processing using overlapping resources has no delay.

[0066] S302. Execute task detection and real-time scheduling decision-making. After confirming that there is no delay in synchronization comparison, monitor the progress of the execution tasks matched by the computing power resource pool. Analyze the completion time of each execution task to infer whether there is a delay in progress. If there is a progress interruption in the task execution phase, no data stream will be generated during the current task execution period. Perform operation detection on the resources corresponding to the execution process of the current execution period. If the resources are running but cannot process the current execution process, no data stream will be generated. Adjust the orchestration of the current execution process to reduce the data processing of the current type of resources. Use resources of the same type that can process data and complete the matching process in advance as the scheduling resources for the current execution process. If the resources have an operation gap, i.e., are not running, no data stream will be generated. Control the instruction sending order and progress of the orchestration of the current execution process to ensure that the start time of the preset execution period of each execution progress is consistent with the time when the corresponding matching resource receives the execution instruction. It should be explained that the consistency of time is the instruction sending deadline. If there is a delay, the current resource operation will be interrupted, the progress of the execution task will be paused, and an early warning will be issued.

[0067] In this invention, the resource layer consists of various computing resources. Resources within the resource layer are identified in real time, and the resource layer configuration is updated based on the identification results. After construction, resource status is monitored during the continuous update phase of the resource layer. The health of the resources is inferred based on real-time monitoring analysis. After status monitoring, resources are reclaimed based on the real-time computing task execution progress. Simultaneously, the resource layer configuration is optimized in real-time based on changes in the computing task usage scenario. The pooling layer manages the various computing resources in the resource layer in a pooled manner. If the real-time resource configuration of the resource layer meets the computing power requirements of the actual usage scenario, the current resource configuration is pooled to construct a computing power resource pool. After constructing the computing power resource pool and resource matching criteria, resources are orchestrated according to the usage scenario of the real-time tasks to be executed. The scheduling layer is used to reasonably match tasks and computing resources based on task requirements and the real-time status of the computing power resource pool.

[0068] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A computing power pooling scheduling and management system based on AI intelligent technology, characterized in that, The system architecture consists of a resource layer, a pooling layer, and a scheduling layer. The resource layer consists of various computing resources. Resources within the resource layer are identified in real time, and resource layer configurations are updated based on the identification results. After construction, resource status is monitored during the continuous update phase of the resource layer. The health of resource status is inferred based on real-time monitoring analysis. After status monitoring, resources are reclaimed based on the real-time computing task execution progress. Simultaneously, the resource layer configuration is optimized in real-time based on changes in computing task usage scenarios. Existing resources within the resource layer are acquired and real-time resource configurations are set. Newly added resources are marked as pending configuration additions, and new optimization data and new risk data are obtained. If the newly added optimized data does not exceed the parameter excess threshold, or if the newly added risky data exceeds the conflict value threshold, it will not be added to the current real-time resource configuration. If the newly added optimized data exceeds the parameter excess threshold, and the newly added risky data does not exceed the conflict value threshold, then add it to the current real-time resource configuration; Resource layer configuration update means that the identification and detection of the configuration to be added is normal, and it is added to the current real-time resource configuration; The newly added optimized data and newly added risk data are the excess amount of computing power data processing parameters of the same type of resources corresponding to the real-time resource configuration and the configuration to be added, and the conflict value of the running parameters of the non-same type of cooperative resources corresponding to the real-time resource configuration and the configuration to be added. The pooling layer pools and manages various computing resources in the resource layer. If the real-time resource configuration in the resource layer meets the computing power requirements of the actual use scenario, the current resource configuration is pooled to build a computing power resource pool. After building the computing power resource pool and the resource coordination and use criteria, resources are orchestrated according to the use scenario of the real-time tasks to be executed. The scheduling layer is used to reasonably match tasks with computing resources based on task requirements and the real-time status of the computing resource pool; when the computing resource pool and the currently executing task are matched, the processing efficiency of the current computing resource pool for the executing task is evaluated by synchronous comparison; after confirming that there is no delay in synchronous comparison, the progress of the executing task matched by the computing resource pool is monitored.

2. The AI-based intelligent computing power pooling scheduling and management system according to claim 1, characterized in that, The system collects the fluctuation trend of running parameters within a set parameter range before and after executing a task using any resource in real-time resource configuration, and constructs parameter curves based on the values ​​of running parameters before and after task execution. If the parameter curves show an upward trend during task execution, and the downward slope of the running parameter curves continues to decrease after the task is completed, then hardware maintenance and performance testing will be performed on the current resources; if the parameter curves do not show an upward trend during task execution, or the downward slope of the running parameter curves does not continue to decrease after the task is completed, then task matching will be performed on the current resources.

3. The AI-based intelligent computing power pooling scheduling and management system according to claim 2, characterized in that, When coordinating task execution with real-time resource configuration at the resource layer, the preset execution period of the task is obtained, and the output data volume corresponding to the data stream of the real-time task execution is compared. If the output data volume corresponding to the data stream decreases to or is lower than the minimum data stream threshold, the current moment is marked as the computing power shutdown moment. Based on the comparison of the amount of data processed by the task at the time of computing power shutdown, if the amount of data processed is within the preset data range and the data processing flow of the task is completed, the current computing power shutdown time is marked as the resource reclamation time, and the resources matched with the current task at the resource layer are reclaimed and rematched with the task. If the amount of data being processed is not within the preset range, or if the data processing flow of the task is completed, then a data processing performance test will be performed on the resources used in the current resource layer.

4. The computing power pooling scheduling and management system based on AI intelligent technology according to claim 3, characterized in that, Based on the execution tasks received by the resource layer, the computing power requirement of the current execution task is obtained, and the processing parameters are compared during the task execution phase. If the fluctuation range of the processing parameters exceeds the span threshold when facing different computing power requirements, and the frequency of the fluctuation of the processing parameters continues to increase with the increase of the execution tasks, it is inferred that the configuration computing power of the resource layer usage scenario needs to be upgraded. Based on the current real-time resource configuration as the type standard, the performance is upgraded on the basis of the original performance parameters. If the processing parameters continue to increase when facing different computing power requirements, and the duration of each fluctuation of the processing parameters continues to increase, it is inferred that the configuration type of the resource layer usage scenario needs to be changed. Taking the current real-time resource configuration as the starting point of the configuration standard, the resource type is adjusted and the overall computing power performance of the configuration is improved.

5. The AI-based intelligent computing power pooling scheduling and management system according to claim 4, characterized in that, If the real-time resource configuration at the resource layer meets the computing power requirements of the actual use scenario, then the current resource configuration will be pooled to build a computing power resource pool; and the criteria for the coordinated use of various types of resources in the computing power resource pool will be determined according to the task execution requirements of each use scenario.

6. The AI-based intelligent computing power pooling scheduling and management system according to claim 5, characterized in that, After constructing the computing power resource pool and resource coordination rules, resources are orchestrated according to the usage scenarios of real-time tasks to be executed. The usage rules corresponding to the execution procedures of real-time tasks to be executed are statistically analyzed and arranged according to the execution procedure order. After the orchestration is completed, the usage rules of the current coordination are statistically analyzed, and resources with overlapping usage rules are used uniformly. That is, resources are not used according to the execution procedure order, and the execution procedures required by the current usage scenario are executed synchronously according to the usage rules. The completed data is stored and transmitted during the orchestration of the corresponding execution process to complete the current execution process. If there is no need to execute the same type of resource after the execution process is completed during the orchestration process, the currently configured resource is released and a new non-current execution task is matched.

7. The AI-based intelligent computing power pooling scheduling and management system according to claim 6, characterized in that, The scheduling process of tasks executed by the real-time computing resource pool is synchronously compared to obtain the frequency of increase in the delay between the completion time of data processing using overlapping resources and the execution time of the corresponding execution process. If the frequency of increase in delay exceeds the increase frequency threshold, it is inferred that the same processing using overlapping resources has a delay, and the computing resource pool is rescheduled to increase the number of resources adapted to the current execution process or increase the processing time. If the frequency of increase in delay does not exceed the increase frequency threshold, it is inferred that the same processing using overlapping resources has no delay.

8. The AI-based intelligent computing power pooling scheduling and management system according to claim 7, characterized in that, The progress of the execution tasks matched by the computing power resource pool is monitored. The completion time of each execution task is analyzed to infer whether there is a delay in the progress. If there is a progress interruption in the task execution stage, no data stream is generated during the current task execution period. The resources corresponding to the execution process of the current execution period are monitored.

9. The AI-based intelligent computing power pooling scheduling and management system according to claim 8, characterized in that, If resources are operational but unable to process the currently executing procedure, the orchestration of the currently executing procedure will be adjusted to reduce data processing of the current type of resources. Resources that can process data of the same type as the current type of resources and can complete the matching procedure in advance will be used as the scheduling resources for the currently executing procedure. If there are idle resources, the order and progress of instruction sending for the current execution process will be controlled to ensure that the start time of the preset execution period of each execution progress is consistent with the time when the corresponding matching resource receives the execution instruction.

Images