Packing schemes for gang scheduling

Dror G. Feitelson

doi:10.1007/bfb0022289

Packing schemes for gang scheduling

Dror G. Feitelson^*

^*Corresponding author for this work

The Rachel and Selim Benin School of Engineering and Computer Science

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

160 Scopus citations

Abstract

Jobs that do not require all processors in the system can be packed together for gang scheduling. We examine accounting traces from several parallel computers to show that indeed many jobs have small sizes and can be packed together. We then formulate a number of such packing algorithms, and evaluate their effectiveness using simulations based on our workload study. The results are that two algorithms are the best: either perform the mapping based on a buddy system of processors, or use migration to re-map the jobs more tightly whenever a job arrives or terminates. Other approaches, such as mapping to the least loaded PEs, proved to be counterproductive. The buddy system approach depends on the capability to gang-schedule jobs in multiple slots, if there is space. The migration algorithm is more robust, but is expected to suffer greatly due to the overhead of the migration itself. In either case fragmentation is not an issue, and utilization may top 90% with sufficiently high loads.

Original language	English
Title of host publication	Job Scheduling Strategies for Parallel Processing - IPPS 1996 Workshop, Proceedings
Editors	Dror G. Feitelson, Larry Rudolph
Publisher	Springer Verlag
Pages	89-110
Number of pages	22
ISBN (Print)	3540618643, 9783540618645
DOIs	https://doi.org/10.1007/bfb0022289
State	Published - 1996
Event	2nd Workshop on Job Scheduling Strategies for Parallel Processing, IPPS 1996 - Honolulu, United States Duration: 16 Apr 1996 → 16 Apr 1996

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	1162
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	2nd Workshop on Job Scheduling Strategies for Parallel Processing, IPPS 1996
Country/Territory	United States
City	Honolulu
Period	16/04/96 → 16/04/96

Bibliographical note

Publisher Copyright:
© Springer-Verlag Berlin Heidelberg 1996.

Access to Document

10.1007/bfb0022289

Cite this

Feitelson, D. G. (1996). Packing schemes for gang scheduling. In D. G. Feitelson, & L. Rudolph (Eds.), Job Scheduling Strategies for Parallel Processing - IPPS 1996 Workshop, Proceedings (pp. 89-110). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1162). Springer Verlag. https://doi.org/10.1007/bfb0022289

@inproceedings{d160d069a654453ba3ad5e14aaee428c,

title = "Packing schemes for gang scheduling",

abstract = "Jobs that do not require all processors in the system can be packed together for gang scheduling. We examine accounting traces from several parallel computers to show that indeed many jobs have small sizes and can be packed together. We then formulate a number of such packing algorithms, and evaluate their effectiveness using simulations based on our workload study. The results are that two algorithms are the best: either perform the mapping based on a buddy system of processors, or use migration to re-map the jobs more tightly whenever a job arrives or terminates. Other approaches, such as mapping to the least loaded PEs, proved to be counterproductive. The buddy system approach depends on the capability to gang-schedule jobs in multiple slots, if there is space. The migration algorithm is more robust, but is expected to suffer greatly due to the overhead of the migration itself. In either case fragmentation is not an issue, and utilization may top 90% with sufficiently high loads.",

author = "Feitelson, {Dror G.}",

note = "Publisher Copyright: {\textcopyright} Springer-Verlag Berlin Heidelberg 1996.; 2nd Workshop on Job Scheduling Strategies for Parallel Processing, IPPS 1996 ; Conference date: 16-04-1996 Through 16-04-1996",

year = "1996",

doi = "10.1007/bfb0022289",

language = "אנגלית",

isbn = "3540618643",

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

publisher = "Springer Verlag",

pages = "89--110",

editor = "Feitelson, {Dror G.} and Larry Rudolph",

booktitle = "Job Scheduling Strategies for Parallel Processing - IPPS 1996 Workshop, Proceedings",

address = "גרמניה",

}

Feitelson, DG 1996, Packing schemes for gang scheduling. in DG Feitelson & L Rudolph (eds), Job Scheduling Strategies for Parallel Processing - IPPS 1996 Workshop, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1162, Springer Verlag, pp. 89-110, 2nd Workshop on Job Scheduling Strategies for Parallel Processing, IPPS 1996, Honolulu, United States, 16/04/96. https://doi.org/10.1007/bfb0022289

Packing schemes for gang scheduling. / Feitelson, Dror G.
Job Scheduling Strategies for Parallel Processing - IPPS 1996 Workshop, Proceedings. ed. / Dror G. Feitelson; Larry Rudolph. Springer Verlag, 1996. p. 89-110 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1162).

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

TY - GEN

T1 - Packing schemes for gang scheduling

AU - Feitelson, Dror G.

N1 - Publisher Copyright: © Springer-Verlag Berlin Heidelberg 1996.

PY - 1996

Y1 - 1996

N2 - Jobs that do not require all processors in the system can be packed together for gang scheduling. We examine accounting traces from several parallel computers to show that indeed many jobs have small sizes and can be packed together. We then formulate a number of such packing algorithms, and evaluate their effectiveness using simulations based on our workload study. The results are that two algorithms are the best: either perform the mapping based on a buddy system of processors, or use migration to re-map the jobs more tightly whenever a job arrives or terminates. Other approaches, such as mapping to the least loaded PEs, proved to be counterproductive. The buddy system approach depends on the capability to gang-schedule jobs in multiple slots, if there is space. The migration algorithm is more robust, but is expected to suffer greatly due to the overhead of the migration itself. In either case fragmentation is not an issue, and utilization may top 90% with sufficiently high loads.

AB - Jobs that do not require all processors in the system can be packed together for gang scheduling. We examine accounting traces from several parallel computers to show that indeed many jobs have small sizes and can be packed together. We then formulate a number of such packing algorithms, and evaluate their effectiveness using simulations based on our workload study. The results are that two algorithms are the best: either perform the mapping based on a buddy system of processors, or use migration to re-map the jobs more tightly whenever a job arrives or terminates. Other approaches, such as mapping to the least loaded PEs, proved to be counterproductive. The buddy system approach depends on the capability to gang-schedule jobs in multiple slots, if there is space. The migration algorithm is more robust, but is expected to suffer greatly due to the overhead of the migration itself. In either case fragmentation is not an issue, and utilization may top 90% with sufficiently high loads.

UR - http://www.scopus.com/inward/record.url?scp=84955614945&partnerID=8YFLogxK

U2 - 10.1007/bfb0022289

DO - 10.1007/bfb0022289

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:84955614945

SN - 3540618643

SN - 9783540618645

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

SP - 89

EP - 110

BT - Job Scheduling Strategies for Parallel Processing - IPPS 1996 Workshop, Proceedings

A2 - Feitelson, Dror G.

A2 - Rudolph, Larry

PB - Springer Verlag

T2 - 2nd Workshop on Job Scheduling Strategies for Parallel Processing, IPPS 1996

Y2 - 16 April 1996 through 16 April 1996

ER -

Packing schemes for gang scheduling

Abstract

Publication series

Conference

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this