电波科学学报2025，Vol.40Issue(2)：246-251,275,7.DOI:10.12265/j.cjors.2024115
一种基于核数分配的任务智能调度方法
A task intelligent scheduling method based on the number of cores
刘柳 1凡益民 2刘田 2张琰3
作者信息
- 1. 西安电子科技大学通信工程学院,西安 710071||中国电子科技集团公司第十研究所,成都 610036
- 2. 中国电子科技集团公司第十研究所,成都 610036
- 3. 西安电子科技大学通信工程学院,西安 710071
- 折叠
摘要
Abstract
The new generation of ground systems for space tracking,telemetry,and command(TT&C)and communication is composed of a large-scale heterogeneous cloud computing resource pool.To allocate available computing power to different TT&C and communication tasks reasonably,an intelligent scheduling method based on core allocation is proposed.This algorithm allocates different CPUs/GPUs for task processing based on the number of cores required by each task.This paper establishes a Markov decision process(MDP)model for the task core allocation problem,constructs a Q-table based on the number of tasks to be executed and the number of available processors for scheduling,formulates a principle of maximizing reward values for load balancing across processor resources,searches for the task allocation path that optimizes scheduling performance,iteratively calculates rewards and updates Q-values,completes multi-cycle training,and continuously iterates and optimizes the output scheduling results based on the reward circuit.Experimental analysis results show that this algorithm can maximize the utilization rate of computing cores while ensuring balanced task deployment.Compared with existing algorithms,it enhances the rationality of multi-core allocation and optimizes the overall system performance.关键词
任务/利用率/核数/调度Key words
task/utilization/number of cores/scheduling分类
计算机与自动化引用本文复制引用
刘柳,凡益民,刘田,张琰..一种基于核数分配的任务智能调度方法[J].电波科学学报,2025,40(2):246-251,275,7.
