Search

CN-122019153-A - Dynamic planning method for intelligent computing center resources based on artificial intelligent prediction

CN122019153ACN 122019153 ACN122019153 ACN 122019153ACN-122019153-A

Abstract

The application belongs to the field of computing resource planning, relates to an artificial intelligent prediction technology, and aims to solve the problem that the prior art cannot dynamically optimize a load prediction result according to a task execution state, in particular to an intelligent computing center resource dynamic planning method based on artificial intelligent prediction, which comprises the following steps of periodically predicting and analyzing a load of a computing center, namely generating a prediction period, dividing the prediction period into a plurality of prediction periods, and acquiring task queue historical data, resource utilization historical data and metadata at the beginning time of the prediction period; the application realizes the intelligent dynamic programming of the computing resources, and the system can rapidly respond to load change through periodical load prediction and resource demand interval division, thereby avoiding the problem of resource fragmentation caused by fixed-scale cluster configuration.

Inventors

  • QI YONG

Assignees

  • 中科信控(北京)科技有限公司

Dates

Publication Date
20260512
Application Date
20260121

Claims (8)

  1. 1. The intelligent computing center resource dynamic programming method based on artificial intelligent prediction is characterized by comprising the following steps of: The method comprises the steps of S1, periodically predicting and analyzing load of a computing center, namely generating a prediction period, dividing the prediction period into a plurality of prediction periods, and acquiring task queue historical data, resource utilization rate historical data and metadata at the beginning time of the prediction period; Step S2, task allocation is carried out on the computing nodes of the computing center, wherein a reserved object is randomly selected and forms a reserved processing cluster; Step S3, analyzing the task allocation rationality of a computing center, namely acquiring a uniformity coefficient of a reserved processing cluster at the end time of a prediction period, counting resource demand peaks of all the prediction periods at the end time of the prediction period, forming a resource demand range by the maximum value and the minimum value of the resource demand peaks, dividing the resource demand range into a plurality of resource demand intervals, and marking a configuration cluster of the resource demand interval; And S4, analyzing the task execution state of the computing center.
  2. 2. The intelligent computing center resource dynamic planning method based on artificial intelligence prediction according to claim 1, wherein in step S1, task queue history data is high-frequency time series data collected from a scheduler, the task queue data includes a time stamp, the number of submitted tasks, the number of tasks in a queue and the number of tasks in operation, resource utilization history data is collected from a monitoring system, the resource utilization history data includes a total CPU utilization, a memory utilization and a GPU utilization, and the metadata includes an average task duration, a requested CPU/GPU/memory.
  3. 3. The dynamic planning method for intelligent computing center resources based on artificial intelligence prediction according to claim 2, wherein in step S2, the specific process of task allocation for computing nodes of the computing center includes marking computing nodes of the computing center as allocation objects, randomly selecting the allocation objects as reservation objects, wherein the number of the reservation objects is equal to the allocation number, the reservation objects form a reservation processing cluster, and the computing tasks received in the prediction period are uniformly allocated to the reservation objects in the reservation processing cluster according to the calculated amount for processing.
  4. 4. The method for dynamic planning of resources in an intelligent computing center based on artificial intelligence prediction according to claim 3, wherein in step S3, the process of obtaining the uniformity coefficient includes counting the total time length of execution of the computing task by each reserved object and marking the total time length as the processing value of the reserved object, and performing variance calculation on the processing values of all reserved objects in the prediction period to obtain the uniformity coefficient.
  5. 5. The method for dynamic resource planning in an intelligent computing center based on artificial intelligence prediction according to claim 4, wherein in step S3, the specific marking process of the configuration cluster of the resource demand interval includes marking the reserved processing cluster of the prediction period with the peak value of the resource demand within the resource demand interval as the matching cluster of the resource demand interval, and marking the matching cluster with the smallest distribution coefficient as the configuration cluster of the resource demand interval.
  6. 6. The method for dynamically planning resources of an intelligent computing center based on artificial intelligence prediction according to claim 5, wherein in step S4, the specific process of analyzing the task execution state of the computing center includes summing up the processing values of all the reserved objects in the configuration cluster corresponding to the resource demand interval to obtain an execution coefficient of the resource demand interval at the end of the prediction period, determining whether the task execution state of the configuration cluster of the resource demand interval meets the requirement by the execution coefficient, and executing optimization analysis on the expansion interval when the task execution state does not meet the requirement.
  7. 7. The method for dynamically planning resources in an intelligent computing center based on artificial intelligence prediction according to claim 6, wherein the specific process of determining whether the task execution state of the configuration cluster in the resource demand interval meets the requirements includes comparing an execution coefficient with a preset execution threshold, determining that the task execution state of the configuration cluster in the resource demand interval meets the requirements if the execution coefficient is smaller than the execution threshold, and determining that the task execution state of the configuration cluster in the resource demand interval does not meet the requirements if the execution coefficient is greater than or equal to the execution threshold, and marking the corresponding resource demand interval as a pre-expansion interval.
  8. 8. The method for dynamic resource planning in intelligent computing center based on artificial intelligence prediction according to claim 7, wherein the specific process of performing optimization analysis on the pre-expansion interval includes marking the number of times that the allocation object is listed in the reserved processing cluster in the current prediction period as an in-listing value, obtaining an allocation update value FPg by a formula FPg = [ t1×fp ], where t1 is a proportionality coefficient, 1.15 is equal to or less than 1.25, FP is a numerical value of the allocation number, adding the allocation object with the largest in-listing value outside the reserved object in the configuration cluster corresponding to the pre-expansion interval into the configuration cluster until the number of the reserved objects in the configuration cluster reaches the allocation update value FPg.

Description

Dynamic planning method for intelligent computing center resources based on artificial intelligent prediction Technical Field The invention belongs to the field of computing resource planning, relates to an artificial intelligent prediction technology, and in particular relates to an intelligent computing center resource dynamic planning method based on artificial intelligent prediction. Background The core of the intelligent computing center resource dynamic planning method is to realize the fundamental transition from passive response to active prediction by using artificial intelligent technology. The method and the system conduct intelligent resource planning and scheduling in advance by accurately predicting future workload, so that efficient, green and automatic operation of computing resources is realized. The invention patent with publication number CN113703952B discloses a resource allocation method for queue resource scheduling based on a supercomputer, which optimizes the allocation of computing resources, improves the efficiency, can keep a vigorous resource queue for resource call in emergency, but can only allocate tasks according to a computing load prediction result and an independent computing node saturation state, and cannot carry out integral analysis on the task execution state of a computing cluster formed by the computing nodes, so that task blocking caused by conflict of display memory allocation in the GPU cluster and task accumulation caused by unbalanced computing power of the computing nodes in a heterogeneous cluster (CPU+GPU), and further, the prior art cannot dynamically optimize the load prediction result according to the task execution state, so that the completion and time rate of the computing tasks cannot be guaranteed. The application provides a solution to the technical problem. Disclosure of Invention The invention aims to provide an intelligent computing center resource dynamic planning method based on artificial intelligent prediction, which is used for solving the problem that the load prediction result cannot be dynamically optimized according to the task execution state in the prior art; The invention aims to provide an intelligent computing center resource dynamic planning method based on artificial intelligent prediction, which can dynamically optimize a load prediction result according to a task execution state. The aim of the invention can be achieved by the following technical scheme: The intelligent computing center resource dynamic programming method based on artificial intelligent prediction comprises the following steps: The method comprises the steps of S1, periodically predicting and analyzing load of a computing center, namely generating a prediction period, dividing the prediction period into a plurality of prediction periods, and acquiring task queue historical data, resource utilization rate historical data and metadata at the beginning time of the prediction period; Step S2, task allocation is carried out on the computing nodes of the computing center, wherein a reserved object is randomly selected and forms a reserved processing cluster; Step S3, analyzing the task allocation rationality of a computing center, namely acquiring a uniformity coefficient of a reserved processing cluster at the end time of a prediction period, counting resource demand peaks of all the prediction periods at the end time of the prediction period, forming a resource demand range by the maximum value and the minimum value of the resource demand peaks, dividing the resource demand range into a plurality of resource demand intervals, and marking a configuration cluster of the resource demand interval; And S4, analyzing the task execution state of the computing center. Further, in step S1, the task queue history data is high-frequency time-series data collected from the scheduler, the task queue data includes a time stamp, the number of submitted tasks, the number of tasks in queue, and the number of tasks in operation, the resource utilization history data is collected from the monitoring system, the resource utilization history data includes a total CPU utilization, a memory usage, and a GPU utilization, and the metadata includes an average task duration, and a requested CPU/GPU/memory. Further, in step S2, the specific process of task allocation to the computing nodes of the computing center includes marking the computing nodes of the computing center as allocation objects, randomly selecting the allocation objects as reserved objects, wherein the number of the reserved objects is equal to the allocation number, forming a reserved processing cluster by the reserved objects, and uniformly distributing the computing tasks received in the prediction period to the reserved objects in the reserved processing cluster for processing according to the calculated amount. Further, in step S3, the process of obtaining the uniformity coefficient includes counting the total time length