An Analytical Bound for Choosing Trivial Strategies in Co-scheduling

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

1 Scopus citations


Efficient usage of shared high-performance computing (HPC) resources raises the problem of HPC applications co-scheduling, i.e. the problem of execution of multiple applications simultaneously on the same shared computing nodes. Each application may have different requirements for shared resources (e.g. network bandwidth or memory bus bandwidth). When these resources are used concurrently, their resource throughputs may decrease, which leads to performance degradation. In this paper we define application behavior model in co-scheduling environment and formalize a scheduling problem. Within the model we evaluate trivial strategies and compare them with an optimal strategy. The comparison provides a simple analytical criteria for choosing between a naive strategy of running all applications in parallel or any sophisticated strategies that account for applications performance degradation.

Original languageEnglish
Title of host publicationComputational Science and Its Applications – ICCSA 2021
Subtitle of host publication21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part VIII
EditorsOsvaldo Gervasi, et al.
PublisherSpringer Nature
ISBN (Print)9783030870096
StatePublished - 2021
Event21st International Conference on Computational Science and Its Applications, ICCSA 2021 - Virtual, Online
Duration: 13 Sep 202116 Sep 2021

Publication series

NameLecture Notes in Computer Science
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference21st International Conference on Computational Science and Its Applications, ICCSA 2021
CityVirtual, Online

Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)


  • Co-scheduling
  • High performance computing
  • Linear programming
  • Scheduling theory


Dive into the research topics of 'An Analytical Bound for Choosing Trivial Strategies in Co-scheduling'. Together they form a unique fingerprint.

Cite this