Empirical study of dynamic scheduling on rings of processors

Miranda E. Barrows, Dawn E. Gregory, Lixin Gao, Arnold L. Rosenberg, Paul R Cohen

Research output: Contribution to journalArticle

Abstract

We empirically analyze and compare two low-overhead, deterministic policies for scheduling dynamically evolving tree-structured computations on rings of identical processing elements (PEs). Our computations have each task either halt or spawn two independent children and then halt; they abstract, for instance, computations generated by multigrid methods. Our simpler policy, KOSO, has each PE keep one child of a spawning task and pass the other to its clockwise neighbor in the ring; our more sophisticated policy, KOSO*, operates similarly, but allows child-passing only from a more heavily loaded PE to a more lightly loaded one. Both policies execute waiting tasks in increasing order of their depths in the evolving task-tree. Our study focuses on two conjectures: (a) Both policies yield good parallel speedup on large classes of the computations we study. (b) Policy KOSO* outperforms policy KOSO in many important situations. We verify these conjectures via a suite of experiments, supplemented by supporting mathematical analyses. We view our methodology of experimental design and analysis as a major component of our study's contribution, which will prove useful in other such studies.

Original languageEnglish (US)
Pages (from-to)1063-1079
Number of pages17
JournalParallel Computing
Volume25
Issue number9
DOIs
StatePublished - Sep 1999
Externally publishedYes

Fingerprint

Dynamic Scheduling
Empirical Study
Scheduling
Ring
Processing
Design of experiments
Clockwise
Multigrid Method
Experimental Analysis
Policy
Experimental design
Speedup
Verify
Experiments
Methodology
Experiment
Children

ASJC Scopus subject areas

  • Computer Science Applications
  • Hardware and Architecture
  • Control and Systems Engineering

Cite this

Empirical study of dynamic scheduling on rings of processors. / Barrows, Miranda E.; Gregory, Dawn E.; Gao, Lixin; Rosenberg, Arnold L.; Cohen, Paul R.

In: Parallel Computing, Vol. 25, No. 9, 09.1999, p. 1063-1079.

Research output: Contribution to journalArticle

Barrows, ME, Gregory, DE, Gao, L, Rosenberg, AL & Cohen, PR 1999, 'Empirical study of dynamic scheduling on rings of processors', Parallel Computing, vol. 25, no. 9, pp. 1063-1079. https://doi.org/10.1016/S0167-8191(99)00039-3
Barrows, Miranda E. ; Gregory, Dawn E. ; Gao, Lixin ; Rosenberg, Arnold L. ; Cohen, Paul R. / Empirical study of dynamic scheduling on rings of processors. In: Parallel Computing. 1999 ; Vol. 25, No. 9. pp. 1063-1079.
@article{4e465c03595642dc9beed8160b088a3c,
title = "Empirical study of dynamic scheduling on rings of processors",
abstract = "We empirically analyze and compare two low-overhead, deterministic policies for scheduling dynamically evolving tree-structured computations on rings of identical processing elements (PEs). Our computations have each task either halt or spawn two independent children and then halt; they abstract, for instance, computations generated by multigrid methods. Our simpler policy, KOSO, has each PE keep one child of a spawning task and pass the other to its clockwise neighbor in the ring; our more sophisticated policy, KOSO*, operates similarly, but allows child-passing only from a more heavily loaded PE to a more lightly loaded one. Both policies execute waiting tasks in increasing order of their depths in the evolving task-tree. Our study focuses on two conjectures: (a) Both policies yield good parallel speedup on large classes of the computations we study. (b) Policy KOSO* outperforms policy KOSO in many important situations. We verify these conjectures via a suite of experiments, supplemented by supporting mathematical analyses. We view our methodology of experimental design and analysis as a major component of our study's contribution, which will prove useful in other such studies.",
author = "Barrows, {Miranda E.} and Gregory, {Dawn E.} and Lixin Gao and Rosenberg, {Arnold L.} and Cohen, {Paul R}",
year = "1999",
month = "9",
doi = "10.1016/S0167-8191(99)00039-3",
language = "English (US)",
volume = "25",
pages = "1063--1079",
journal = "Parallel Computing",
issn = "0167-8191",
publisher = "Elsevier",
number = "9",

}

TY - JOUR

T1 - Empirical study of dynamic scheduling on rings of processors

AU - Barrows, Miranda E.

AU - Gregory, Dawn E.

AU - Gao, Lixin

AU - Rosenberg, Arnold L.

AU - Cohen, Paul R

PY - 1999/9

Y1 - 1999/9

N2 - We empirically analyze and compare two low-overhead, deterministic policies for scheduling dynamically evolving tree-structured computations on rings of identical processing elements (PEs). Our computations have each task either halt or spawn two independent children and then halt; they abstract, for instance, computations generated by multigrid methods. Our simpler policy, KOSO, has each PE keep one child of a spawning task and pass the other to its clockwise neighbor in the ring; our more sophisticated policy, KOSO*, operates similarly, but allows child-passing only from a more heavily loaded PE to a more lightly loaded one. Both policies execute waiting tasks in increasing order of their depths in the evolving task-tree. Our study focuses on two conjectures: (a) Both policies yield good parallel speedup on large classes of the computations we study. (b) Policy KOSO* outperforms policy KOSO in many important situations. We verify these conjectures via a suite of experiments, supplemented by supporting mathematical analyses. We view our methodology of experimental design and analysis as a major component of our study's contribution, which will prove useful in other such studies.

AB - We empirically analyze and compare two low-overhead, deterministic policies for scheduling dynamically evolving tree-structured computations on rings of identical processing elements (PEs). Our computations have each task either halt or spawn two independent children and then halt; they abstract, for instance, computations generated by multigrid methods. Our simpler policy, KOSO, has each PE keep one child of a spawning task and pass the other to its clockwise neighbor in the ring; our more sophisticated policy, KOSO*, operates similarly, but allows child-passing only from a more heavily loaded PE to a more lightly loaded one. Both policies execute waiting tasks in increasing order of their depths in the evolving task-tree. Our study focuses on two conjectures: (a) Both policies yield good parallel speedup on large classes of the computations we study. (b) Policy KOSO* outperforms policy KOSO in many important situations. We verify these conjectures via a suite of experiments, supplemented by supporting mathematical analyses. We view our methodology of experimental design and analysis as a major component of our study's contribution, which will prove useful in other such studies.

UR - http://www.scopus.com/inward/record.url?scp=0032645114&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032645114&partnerID=8YFLogxK

U2 - 10.1016/S0167-8191(99)00039-3

DO - 10.1016/S0167-8191(99)00039-3

M3 - Article

AN - SCOPUS:0032645114

VL - 25

SP - 1063

EP - 1079

JO - Parallel Computing

JF - Parallel Computing

SN - 0167-8191

IS - 9

ER -