Scholarlink

Abstract <p>Authors consider an applicability of machine learning methods to predict task execution time on nodes of high-performance computing devices in order to guarantee an efficient usage of computational resources and therefore optimal planning. The purpose is to find methods to minimize difference between estimated and registered value of the execution time. Authors use system logs with data on computational resource consumptions. Results are presented with estimated dispersion and confidence intervals.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Methods for Predicting Task Execution Time in a Heterogeneous Network of Computers

  • 期刊论文

  • 发表日期：

    2026年7月20日

查看全文
  • V. O. Piskovski,
  • D. R. Lysenko

摘要

Abstract

Authors consider an applicability of machine learning methods to predict task execution time on nodes of high-performance computing devices in order to guarantee an efficient usage of computational resources and therefore optimal planning. The purpose is to find methods to minimize difference between estimated and registered value of the execution time. Authors use system logs with data on computational resource consumptions. Results are presented with estimated dispersion and confidence intervals.