Publication detail

Performance-Cost Optimization of Moldable Scientific Workflows

JAROŠ, M. JAROŠ, J.

Original Title

Performance-Cost Optimization of Moldable Scientific Workflows

Type

conference paper

Language

English

Original Abstract

Moldable scientific workflows represent a special class of scientific workflows where the tasks are written as distributed programs being able to exploit various amounts of computer resources. However, current cluster job schedulers require the user to specify the amount of resources per task manually. This often leads to suboptimal execution time and related cost of the whole workflow execution since many users have only limited experience and knowledge of the parallel efficiency and scaling. This paper proposes several mechanisms to automatically optimize the execution parameters of moldable workflows using genetic algorithms. The paper introduces a local optimization of workflow tasks, a global optimization of the workflow on systems with on-demand resource allocation, and a global optimization for systems with static resource allocation. Several objectives including the workflow makespan, computational cost and the percentage of idling nodes are investigated together with a trade-off parameter putting stress on one objective or another. The paper also discusses the structure and quality of several evolved workflow schedules and the possible reduction in makespan or cost. Finally, the computational requirements of evolutionary process together with the recommended genetic algorithm settings are investigated. The most complex workflows may be evolved in less than two minutes using the global optimization while in only 14s using the local optimization.

Keywords

task graph scheduling, workflow, genetic algorithm, moldable tasks, makespan estimation

Authors

JAROŠ, M.; JAROŠ, J.

Released

30. 9. 2021

Publisher

Springer International Publishing

Location

Portland, Oregon USA

ISBN

978-3-030-88223-5

Book

Job Scheduling Strategies for Parallel Processing

Edition

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ISBN

0302-9743

Periodical

Lecture Notes in Computer Science

Number

12985

State

Federal Republic of Germany

Pages from

149

Pages to

167

Pages count

19

URL

BibTex

@inproceedings{BUT175770,
  author="Marta {Jaroš} and Jiří {Jaroš}",
  title="Performance-Cost Optimization of Moldable Scientific Workflows",
  booktitle="Job Scheduling Strategies for Parallel Processing",
  year="2021",
  series="Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
  journal="Lecture Notes in Computer Science",
  number="12985",
  pages="149--167",
  publisher="Springer International Publishing",
  address="Portland, Oregon USA",
  doi="10.1007/978-3-030-88224-2\{_}8",
  isbn="978-3-030-88223-5",
  issn="0302-9743",
  url="https://link.springer.com/book/10.1007%2F978-3-030-88224-2"
}