Scholarlink

Abstract <p>This paper proposes a survey of ODA (operational data analytics) approaches for HPC centers. The goal of such approaches is to provide tools that constantly analyze monitoring data collected on a supercomputer and, based on this information, help to optimize the functioning of an HPC system. Since the collected and processed data is contained in huge volumes, the use of smart analytics such as machine learning techniques is necessary for tools development. In this paper, we extend the original classification of ODA approaches by dividing existing solutions based on the type of machine learning used in them. We also provide a review of several actual HPC problems in this area and describe the most prominent papers related to each problem.</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

Review of ML-based Solutions for HPC Systems Analysis

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • D. I. Lichmanov,
  • V. V. Voevodin,
  • V. A. Matveev,
  • A. S. Antonov

摘要

Abstract

This paper proposes a survey of ODA (operational data analytics) approaches for HPC centers. The goal of such approaches is to provide tools that constantly analyze monitoring data collected on a supercomputer and, based on this information, help to optimize the functioning of an HPC system. Since the collected and processed data is contained in huge volumes, the use of smart analytics such as machine learning techniques is necessary for tools development. In this paper, we extend the original classification of ODA approaches by dividing existing solutions based on the type of machine learning used in them. We also provide a review of several actual HPC problems in this area and describe the most prominent papers related to each problem.