一种考虑局部性的作业执行时间预测算法

打开文本图片集

中图分类号：TP39 文献标志码：A 文章编号：1671-5489（2025）06-1685-09

Abstract： Aiming at the problem of underutilization of job locality and low prediction accuracy in highperformance computing systems，we proposed a job runtime prediction algorithm considering locality. The algorithm comprehensively utilized the global and local features of job log data，and improved prediction accuracy through a voting mechanism that combined with machine learning prediction and locality-based time-series prediction. Experimental results show that on actual scheduling log datasets such as Unliu Gaia and PIK IPLEX，the JRPL algorithm outperforms or is not inferior to the machine learning algorithms as the baseline in all three metrics： average absolute error， average prediction accuracy，and hit rate. This research result provides an improved prediction model for job scheduling in high-performance computing systems， which hepls to make more accurate execution time forecasting，improve system resource utilization，and reduce computational costs.

Keywords： high-performance computing system; job runtime prediction； machine learning; localized; prediction accuracy

高性能计算在工程模拟、人工智能、工业制造等领域应用广泛．在高性能计算系统中，常用的调度算法[1]有先来先服务（first come first service，FCFS）和短作业优先（short job first，SJF）等．Tang等[2]提出高效的调度策略需要对作业执行时间有一个大概估计，估计越精确，调度越高效[3]。（剩余12815字）