On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems

Henri Casanova; Yick Ching Wong; Loïc Pottier; Rafael Ferreira da Silva

doi:10.1007/978-3-031-22698-4_1

On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems

Henri Casanova
, Yick Ching Wong
, Loïc Pottier
, Rafael Ferreira da Silva

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Runtime systems that automate the execution of applications on distributed cyberinfrastructures need to make scheduling decisions. Researchers have proposed many scheduling algorithms, but most of them are designed based on analytical models and assumptions that may not hold in practice. The literature is thus rife with algorithms that have been evaluated only within the scope of their underlying assumptions but whose practical effectiveness is unclear. It is thus difficult for developers to decide which algorithm to implement in their runtime systems. To obviate the above difficulty, we propose an approach by which the runtime system executes, throughout application execution, simulations of this very execution. Each simulation is for a different algorithm in a scheduling algorithm portfolio, and the best algorithm is selected based on simulation results. The main objective of this work is to evaluate the feasibility and potential merit of this portfolio scheduling approach, even in the presence of simulation inaccuracy, when compared to the traditional one-algorithm approach. We perform this evaluation via a case study in the context of scientific workflows. Our main finding is that portfolio scheduling can outperform the best one-algorithm approach even in the presence of relatively large simulation inaccuracies.

Original language	English
Title of host publication	Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers
Editors	Dalibor Klusácek, Corbalán Julita, Gonzalo P. Rodrigo
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	3-24
Number of pages	22
ISBN (Print)	9783031226977
DOIs	https://doi.org/10.1007/978-3-031-22698-4_1
State	Published - 2023
Event	25th International Workshop on Job Scheduling Strategies for Parallel Processing, JSSPP 2022, held in conjunction with the 36th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2022 - Virtual, Online Duration: Jun 3 2022 → Jun 3 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13592 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	25th International Workshop on Job Scheduling Strategies for Parallel Processing, JSSPP 2022, held in conjunction with the 36th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2022
City	Virtual, Online
Period	06/3/22 → 06/3/22

Funding

This manuscript has been authored in part by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The publisher acknowledges the US government license to provide public access under the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan). Acknowledgments. This work is funded by NSF contracts #2106059 and #2106147: “Collaborative Research: OAC Core: Simulation-driven runtime resource management for distributed workflow applications”; and partially funded by NSF contracts #2103489 and #2103508. This research used resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. Finally, we thank the NSF Chameleon Cloud for providing time grants to access their resources.

Keywords

On-line simulation
Portfolio scheduling
Workflows

Access to Document

10.1007/978-3-031-22698-4_1

Cite this

Casanova, H., Wong, Y. C., Pottier, L., & da Silva, R. F. (2023). On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems. In D. Klusácek, C. Julita, & G. P. Rodrigo (Eds.), Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers (pp. 3-24). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13592 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-22698-4_1

Casanova, Henri ; Wong, Yick Ching ; Pottier, Loïc et al. / On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems. Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers. editor / Dalibor Klusácek ; Corbalán Julita ; Gonzalo P. Rodrigo. Springer Science and Business Media Deutschland GmbH, 2023. pp. 3-24 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{e868bb5067544a769651ef3ae7f8203b,

title = "On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems",

abstract = "Runtime systems that automate the execution of applications on distributed cyberinfrastructures need to make scheduling decisions. Researchers have proposed many scheduling algorithms, but most of them are designed based on analytical models and assumptions that may not hold in practice. The literature is thus rife with algorithms that have been evaluated only within the scope of their underlying assumptions but whose practical effectiveness is unclear. It is thus difficult for developers to decide which algorithm to implement in their runtime systems. To obviate the above difficulty, we propose an approach by which the runtime system executes, throughout application execution, simulations of this very execution. Each simulation is for a different algorithm in a scheduling algorithm portfolio, and the best algorithm is selected based on simulation results. The main objective of this work is to evaluate the feasibility and potential merit of this portfolio scheduling approach, even in the presence of simulation inaccuracy, when compared to the traditional one-algorithm approach. We perform this evaluation via a case study in the context of scientific workflows. Our main finding is that portfolio scheduling can outperform the best one-algorithm approach even in the presence of relatively large simulation inaccuracies.",

keywords = "On-line simulation, Portfolio scheduling, Workflows",

author = "Henri Casanova and Wong, \{Yick Ching\} and Lo{\"i}c Pottier and \{da Silva\}, \{Rafael Ferreira\}",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 25th International Workshop on Job Scheduling Strategies for Parallel Processing, JSSPP 2022, held in conjunction with the 36th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2022 ; Conference date: 03-06-2022 Through 03-06-2022",

year = "2023",

doi = "10.1007/978-3-031-22698-4\_1",

language = "English",

isbn = "9783031226977",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "3--24",

editor = "Dalibor Klus{\'a}cek and Corbal{\'a}n Julita and Rodrigo, \{Gonzalo P.\}",

booktitle = "Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers",

}

Casanova, H, Wong, YC, Pottier, L & da Silva, RF 2023, On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems. in D Klusácek, C Julita & GP Rodrigo (eds), Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13592 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 3-24, 25th International Workshop on Job Scheduling Strategies for Parallel Processing, JSSPP 2022, held in conjunction with the 36th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2022, Virtual, Online, 06/3/22. https://doi.org/10.1007/978-3-031-22698-4_1

On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems. / Casanova, Henri; Wong, Yick Ching; Pottier, Loïc et al.
Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers. ed. / Dalibor Klusácek; Corbalán Julita; Gonzalo P. Rodrigo. Springer Science and Business Media Deutschland GmbH, 2023. p. 3-24 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13592 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems

AU - Casanova, Henri

AU - Wong, Yick Ching

AU - Pottier, Loïc

AU - da Silva, Rafael Ferreira

PY - 2023

Y1 - 2023

N2 - Runtime systems that automate the execution of applications on distributed cyberinfrastructures need to make scheduling decisions. Researchers have proposed many scheduling algorithms, but most of them are designed based on analytical models and assumptions that may not hold in practice. The literature is thus rife with algorithms that have been evaluated only within the scope of their underlying assumptions but whose practical effectiveness is unclear. It is thus difficult for developers to decide which algorithm to implement in their runtime systems. To obviate the above difficulty, we propose an approach by which the runtime system executes, throughout application execution, simulations of this very execution. Each simulation is for a different algorithm in a scheduling algorithm portfolio, and the best algorithm is selected based on simulation results. The main objective of this work is to evaluate the feasibility and potential merit of this portfolio scheduling approach, even in the presence of simulation inaccuracy, when compared to the traditional one-algorithm approach. We perform this evaluation via a case study in the context of scientific workflows. Our main finding is that portfolio scheduling can outperform the best one-algorithm approach even in the presence of relatively large simulation inaccuracies.

AB - Runtime systems that automate the execution of applications on distributed cyberinfrastructures need to make scheduling decisions. Researchers have proposed many scheduling algorithms, but most of them are designed based on analytical models and assumptions that may not hold in practice. The literature is thus rife with algorithms that have been evaluated only within the scope of their underlying assumptions but whose practical effectiveness is unclear. It is thus difficult for developers to decide which algorithm to implement in their runtime systems. To obviate the above difficulty, we propose an approach by which the runtime system executes, throughout application execution, simulations of this very execution. Each simulation is for a different algorithm in a scheduling algorithm portfolio, and the best algorithm is selected based on simulation results. The main objective of this work is to evaluate the feasibility and potential merit of this portfolio scheduling approach, even in the presence of simulation inaccuracy, when compared to the traditional one-algorithm approach. We perform this evaluation via a case study in the context of scientific workflows. Our main finding is that portfolio scheduling can outperform the best one-algorithm approach even in the presence of relatively large simulation inaccuracies.

KW - On-line simulation

KW - Portfolio scheduling

KW - Workflows

UR - https://www.scopus.com/pages/publications/85148685481

U2 - 10.1007/978-3-031-22698-4_1

DO - 10.1007/978-3-031-22698-4_1

M3 - Conference contribution

AN - SCOPUS:85148685481

SN - 9783031226977

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 3

EP - 24

BT - Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers

A2 - Klusácek, Dalibor

A2 - Julita, Corbalán

A2 - Rodrigo, Gonzalo P.

PB - Springer Science and Business Media Deutschland GmbH

T2 - 25th International Workshop on Job Scheduling Strategies for Parallel Processing, JSSPP 2022, held in conjunction with the 36th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2022

Y2 - 3 June 2022 through 3 June 2022

ER -

Casanova H, Wong YC, Pottier L, da Silva RF. On the Feasibility of Simulation-Driven Portfolio Scheduling for Cyberinfrastructure Runtime Systems. In Klusácek D, Julita C, Rodrigo GP, editors, Job Scheduling Strategies for Parallel Processing - 25th International Workshop, JSSPP 2022, Revised Selected Papers. Springer Science and Business Media Deutschland GmbH. 2023. p. 3-24. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-22698-4_1