Speed of parallel processing for random task graphs

Marco Isopi, Charles Newman

Research output: Contribution to journalArticle

Abstract

The random graph model of parallel computation introduced by Gelenbe et al. depends on three parameters: n, the number of tasks (vertices); F, the common distribution of T1,…, Tn, the task processing times, and p = pn, the probability for a given i < j that task i must be completed before task j is started. The total processing time is Rn, the maximum sum of Ti's along directed paths of the graph. We study the large n behavior of Rn when npn grows sublinearly but superlogarithmically, the regime where the longest directed path contains about enpn tasks. For an exponential (mean one) F, we prove that Rn is about 4npn. The “discrepancy” between 4 and e is a large deviation effect. Related results are obtained when npn grows exactly logarithmically and when F is not exponential, but has a tail which decays (at least) exponentially fast. © 1994 John Wiley L Sons, Inc.

Original languageEnglish (US)
Pages (from-to)361-376
Number of pages16
JournalCommunications on Pure and Applied Mathematics
Volume47
Issue number3
DOIs
StatePublished - 1994

Fingerprint

Task Graph
Parallel Processing
Random Graphs
Path
Parallel Computation
Graph Model
Processing
Large Deviations
Discrepancy
Tail
Decay
Graph in graph theory

ASJC Scopus subject areas

  • Mathematics(all)
  • Applied Mathematics

Cite this

Speed of parallel processing for random task graphs. / Isopi, Marco; Newman, Charles.

In: Communications on Pure and Applied Mathematics, Vol. 47, No. 3, 1994, p. 361-376.

Research output: Contribution to journalArticle

@article{c678ff52b7694a668856d6f8816f9e8f,
title = "Speed of parallel processing for random task graphs",
abstract = "The random graph model of parallel computation introduced by Gelenbe et al. depends on three parameters: n, the number of tasks (vertices); F, the common distribution of T1,…, Tn, the task processing times, and p = pn, the probability for a given i < j that task i must be completed before task j is started. The total processing time is Rn, the maximum sum of Ti's along directed paths of the graph. We study the large n behavior of Rn when npn grows sublinearly but superlogarithmically, the regime where the longest directed path contains about enpn tasks. For an exponential (mean one) F, we prove that Rn is about 4npn. The “discrepancy” between 4 and e is a large deviation effect. Related results are obtained when npn grows exactly logarithmically and when F is not exponential, but has a tail which decays (at least) exponentially fast. {\circledC} 1994 John Wiley L Sons, Inc.",
author = "Marco Isopi and Charles Newman",
year = "1994",
doi = "10.1002/cpa.3160470307",
language = "English (US)",
volume = "47",
pages = "361--376",
journal = "Communications on Pure and Applied Mathematics",
issn = "0010-3640",
publisher = "Wiley-Liss Inc.",
number = "3",

}

TY - JOUR

T1 - Speed of parallel processing for random task graphs

AU - Isopi, Marco

AU - Newman, Charles

PY - 1994

Y1 - 1994

N2 - The random graph model of parallel computation introduced by Gelenbe et al. depends on three parameters: n, the number of tasks (vertices); F, the common distribution of T1,…, Tn, the task processing times, and p = pn, the probability for a given i < j that task i must be completed before task j is started. The total processing time is Rn, the maximum sum of Ti's along directed paths of the graph. We study the large n behavior of Rn when npn grows sublinearly but superlogarithmically, the regime where the longest directed path contains about enpn tasks. For an exponential (mean one) F, we prove that Rn is about 4npn. The “discrepancy” between 4 and e is a large deviation effect. Related results are obtained when npn grows exactly logarithmically and when F is not exponential, but has a tail which decays (at least) exponentially fast. © 1994 John Wiley L Sons, Inc.

AB - The random graph model of parallel computation introduced by Gelenbe et al. depends on three parameters: n, the number of tasks (vertices); F, the common distribution of T1,…, Tn, the task processing times, and p = pn, the probability for a given i < j that task i must be completed before task j is started. The total processing time is Rn, the maximum sum of Ti's along directed paths of the graph. We study the large n behavior of Rn when npn grows sublinearly but superlogarithmically, the regime where the longest directed path contains about enpn tasks. For an exponential (mean one) F, we prove that Rn is about 4npn. The “discrepancy” between 4 and e is a large deviation effect. Related results are obtained when npn grows exactly logarithmically and when F is not exponential, but has a tail which decays (at least) exponentially fast. © 1994 John Wiley L Sons, Inc.

UR - http://www.scopus.com/inward/record.url?scp=84990706843&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990706843&partnerID=8YFLogxK

U2 - 10.1002/cpa.3160470307

DO - 10.1002/cpa.3160470307

M3 - Article

VL - 47

SP - 361

EP - 376

JO - Communications on Pure and Applied Mathematics

JF - Communications on Pure and Applied Mathematics

SN - 0010-3640

IS - 3

ER -