CN-121979803-A - Multi-level intelligent distribution method and system for full-performance detection test tasks

CN121979803ACN 121979803 ACN121979803 ACN 121979803ACN-121979803-A

Abstract

A multi-level intelligent distribution method and system for full-performance detection test tasks. The method comprises the steps of obtaining a full-performance detection test task, analyzing the full-performance detection test task to obtain basic information, performing task feature mapping on the basic information to obtain task feature vectors, performing self-adaptive clustering on the task according to the task feature vectors to generate task clusters, detecting all available resource information, obtaining resource capacity vectors based on the resource information, layering to obtain resource pools of different types, calculating the adaptation degree of the task clusters and the resource pools, distributing the task clusters to an optimal resource pool based on the adaptation degree, performing multi-level task distribution on the tasks in the task clusters based on task priorities after distributing all the task clusters to the resource pools, and optimizing multi-level task distribution results through reinforcement learning to achieve multi-level intelligent distribution of the full-performance detection test task. The scheme of the invention improves the task scheduling efficiency.

Inventors

ZHAO SHUANGSHUANG
LI JUN
CAO XIAODONG
CHEN WENGUANG
Wang siyun
GONG DAN
BAO JIN
YI YONGXIAN
CHEN YUHAN
XIA GUOFANG

Assignees

国网江苏省电力有限公司营销服务中心
国网江苏省电力有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. The multi-level intelligent distribution method for the full-performance detection test task is characterized by comprising the following steps of: Step S1, acquiring a full-performance detection test task, analyzing the full-performance detection test task to obtain basic information, performing task feature mapping on the basic information to obtain a task feature vector, and performing self-adaptive clustering on the task according to the task feature vector to generate a task cluster; Step S2, detecting all available resource information, obtaining a resource capacity vector based on the resource information, layering to obtain resource pools of different types, calculating the adaptation degree of task clusters and the resource pools, and distributing the task clusters to an optimal resource pool based on the adaptation degree; And step S3, after all the task clusters are distributed to the resource pool, carrying out multi-level task distribution on the tasks in the task clusters based on task priorities, and optimizing the multi-level task distribution result through reinforcement learning to realize multi-level intelligent distribution of the full-performance detection test tasks.
2. The multi-level intelligent allocation method of full-performance test tasks according to claim 1, wherein the analyzing the full-performance test tasks to obtain basic information further comprises: And analyzing the acquired task by using a structured task analysis algorithm, an unstructured task data analysis algorithm, task feature extraction based on rule matching and a task dependency analysis algorithm to obtain task basic information, wherein the basic information comprises task complexity, priority, execution time, dependency relationship, environmental constraint, resource affinity and historical task feedback.
3. The multi-level intelligent allocation method of full-performance detection test tasks according to claim 2, wherein the self-adaptive clustering is performed on the tasks according to the task feature vectors to generate task clusters, and further comprising: Normalizing the task feature vector to obtain normalized task feature vector representation, and introducing a nonlinear topological similarity measurement calculation formula: Wherein, the Is a task And tasks The nonlinear topological similarity is used for measuring the proximity degree of two tasks in the task feature space; Is the dimension of the task feature vector; is a task feature index; Is the first Weighting coefficients of the individual task features; Is a task In the first place Standardized features on dimensional features; Is a task In the first place Standardized features on dimensional features; Is the first An exponential scaling parameter for each task feature; Is the scaling factor for the overall similarity calculation.
4. The multi-level intelligent allocation method of full-performance detection test tasks according to claim 3, wherein the self-adaptive clustering is performed on the tasks according to the task feature vectors to generate task clusters, and further comprising: similarity, dependency weight and computational complexity correlation among comprehensive tasks are defined, the comprehensive task similarity is defined, and an initial clustering center is determined by constructing a task association graph.
5. The multi-level intelligent allocation method of full-performance test tasks according to claim 4, wherein the resource pools are partitioned according to computing power, storage power or sensors.
6. The multi-level intelligent allocation method of full-performance detection test tasks according to claim 5, wherein the calculation formula of the adaptation degree of the calculation task cluster and the resource pool is as follows: Wherein, the Is a task cluster And resource pool Degree of adaptation between; is a first weighting coefficient for adjusting the weighted similarity A weight in the fitness calculation; Is the first Weights of the individual features; is a second weighting coefficient for adjusting the difference term A weight in the fitness calculation; Is the number of tasks in the task cluster; Is a task cluster Middle (f) The task is at the first Task features in the individual dimensions; is the first of the resource pool A plurality of features; Is a norm calculation; is a constant.
7. The multi-level intelligent distribution method of full-performance test tasks according to claim 6, wherein optimizing the multi-level task distribution results by reinforcement learning, further comprises: The method comprises the steps of initializing a Q-table, training a reinforcement learning model according to historical task execution data, predicting execution effects of different scheduling strategies, calculating a Q value in real time in a task scheduling process, selecting an optimal action to adjust the task scheduling strategy, updating the Q value after task execution is completed, optimizing a next round of task scheduling decision, and learning the optimal task scheduling strategy after multiple rounds of training.
8. A multi-level intelligent distribution system for full performance test tasks, comprising: The task cluster generation module is used for acquiring a full-performance detection test task, analyzing the full-performance detection test task to obtain basic information, performing task feature mapping on the basic information to obtain a task feature vector, and performing self-adaptive clustering on the task according to the task feature vector to generate a task cluster; the task cluster allocation module is used for detecting all available resource information, obtaining a resource capacity vector based on the resource information, layering to obtain different types of resource pools, calculating the adaptation degree of the task clusters and the resource pools, and allocating the task clusters to the optimal resource pools based on the adaptation degree; and the multi-level distribution module is used for carrying out multi-level task distribution on the tasks in the task clusters based on the task priorities after distributing all the task clusters to the resource pool, optimizing the multi-level task distribution results through reinforcement learning, and realizing multi-level intelligent distribution of the full-performance detection test tasks.
9. A terminal comprises a processor and a storage medium, and is characterized in that: The storage medium is used for storing instructions; The processor is configured to operate in accordance with the instructions to perform the steps of the multi-level intelligent distribution method of full performance test tasks according to any one of claims 1-7.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the program when executed by a processor performs the steps of the multi-level intelligent distribution method of full performance test tasks according to any of claims 1-7.

Description

Multi-level intelligent distribution method and system for full-performance detection test tasks Technical Field The invention belongs to the field of task scheduling, and particularly relates to a multi-level intelligent allocation method and system for full-performance detection test tasks. Background With the continuous development of information technology and computing power, the demand of computing resources is rapidly increasing, and particularly in the fields of cloud computing, big data, artificial intelligence and the like, task scheduling and resource management have become one of key technologies for improving system performance. In conventional computing resource management, task scheduling algorithms mostly adopt static or preset rules to allocate tasks, and complex relationships between tasks and resources are ignored. The static allocation mode always looks at the catch of the forepart when facing diversified and dynamically-changed task demands, so that the resource allocation is uneven, the resource utilization rate is low and the system response time is long. In the prior art, scheduling algorithms focus on static management of resources without fully taking into account the diversity of tasks and the dynamically changing demands. For example, simple round robin scheduling algorithms, priority scheduling algorithms, and time slice based scheduling strategies, while capable of functioning in certain specific environments, are not effective in complex scenarios such as resource load non-uniformity, task-to-task relevance ambiguity, etc. In addition, conventional methods fail to take into account historical execution data of tasks and dynamic changes in resources, resulting in the accuracy and efficiency of system scheduling being compromised when complex computing tasks are processed. In addition, the prior art also has the technical problems that the distribution efficiency is reduced due to insufficient analysis of task information, and the resource utilization rate is lower due to poor matching effect of tasks and resources. Disclosure of Invention In order to solve the defects in the prior art, the invention provides a multi-level intelligent allocation method and system for full-performance detection test tasks, which are used for solving the technical problems of reduced allocation efficiency and lower task and resource utilization rate. In order to solve the technical problems, the invention adopts the following technical scheme. The invention firstly discloses a multi-level intelligent distribution method of full-performance detection test tasks, which comprises the following steps: Step S1, acquiring a full-performance detection test task, analyzing the full-performance detection test task to obtain basic information, performing task feature mapping on the basic information to obtain a task feature vector, and performing self-adaptive clustering on the task according to the task feature vector to generate a task cluster; Step S2, detecting all available resource information, obtaining a resource capacity vector based on the resource information, layering to obtain resource pools of different types, calculating the adaptation degree of task clusters and the resource pools, and distributing the task clusters to an optimal resource pool based on the adaptation degree; And step S3, after all the task clusters are distributed to the resource pool, carrying out multi-level task distribution on the tasks in the task clusters based on task priorities, and optimizing the multi-level task distribution result through reinforcement learning to realize multi-level intelligent distribution of the full-performance detection test tasks. The invention further comprises the following preferable schemes: The analyzing the full-performance detection test task to obtain basic information further comprises: And analyzing the acquired task by using a structured task analysis algorithm, an unstructured task data analysis algorithm, task feature extraction based on rule matching and a task dependency analysis algorithm to obtain task basic information, wherein the basic information comprises task complexity, priority, execution time, dependency relationship, environmental constraint, resource affinity and historical task feedback. The self-adaptive clustering is carried out on the tasks according to the task feature vectors to generate task clusters, and the method further comprises the following steps: Normalizing the task feature vector to obtain normalized task feature vector representation, and introducing a nonlinear topological similarity measurement calculation formula: Wherein, the Is a taskAnd tasksThe nonlinear topological similarity is used for measuring the proximity degree of two tasks in the task feature space; Is the dimension of the task feature vector; is a task feature index; Is the first Weighting coefficients of the individual task features; Is a task In the first placeStandardized f