Scholarlink

Abstract <p>Fragmentation of supercomputer resources is usually considered a negative factor that reduces the utilization of HPC systems. Most research focuses on minimizing the downtime of supercomputer nodes and cores by reducing fragmentation. This article presents a method for increasing computing speed based on the artificial fragmentation. Artificial fragmentation adds supercomputer nodes to certain jobs in order to reduce competition for resources among parallel program threads. The results of simulation modeling and experiments show that artificial fragmentation not only increases the speed of individual jobs, but also improves the performance of the supercomputer system when processing a jobs flow.</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

Job Scheduling Based on the Supercomputer Resources Fragmentation Management

  • 期刊论文

  • 发表日期：

    2025年11月1日

查看全文
  • A. V. Baranov,
  • K. A. Bragin

摘要

Abstract

Fragmentation of supercomputer resources is usually considered a negative factor that reduces the utilization of HPC systems. Most research focuses on minimizing the downtime of supercomputer nodes and cores by reducing fragmentation. This article presents a method for increasing computing speed based on the artificial fragmentation. Artificial fragmentation adds supercomputer nodes to certain jobs in order to reduce competition for resources among parallel program threads. The results of simulation modeling and experiments show that artificial fragmentation not only increases the speed of individual jobs, but also improves the performance of the supercomputer system when processing a jobs flow.