Load Partitioning and Trade-Off Study for Large Matrix-Vector Computations in Multicast Bus Networks with Communication Delays

Debasish Ghose, Hyong Joong Kim

Research output: Contribution to journalArticle

33 Citations (Scopus)

Abstract

In this paper we consider the problem of computing a large matrix-vector product in a network-based distributed computing environment comprising computers equipped with communication co-processors that may be used for communication off-loading. Communication delays, which are significant in such systems, are specifically taken into account. The important contribution of this study is to show that the optimal load partitioning, and the subsequent performance of the network, depends critically on many network parameters and load characteristics. In particular, it is shown that the size of the load plays an important role in determining the performance of the network. We consider only row-wise striping of the matrix in order to better allocate the computational burden among the processors. We derive closed-form solutions to the optimal load partitioning problem and show the existence of optimal load sharing conditions. An important and practically relevant trade-off study, from the architecture point of view, between the number of processors and the bus bandwidth is presented. Several practical load distribution strategies are considered and complete analyses for each of them is presented

Original languageEnglish
Pages (from-to)32-59
Number of pages28
JournalJournal of Parallel and Distributed Computing
Volume55
Issue number1
DOIs
Publication statusPublished - 1998 Nov 25
Externally publishedYes

Fingerprint

Communication Delay
Multicast
Partitioning
Trade-offs
Communication
Distributed computer systems
Load Sharing
Load Distribution
Cross product
Matrix Product
Distributed Computing
Closed-form Solution
Bandwidth
Computing

ASJC Scopus subject areas

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

Cite this

@article{dd6bd563b5ae497b966d66ac5b665536,
title = "Load Partitioning and Trade-Off Study for Large Matrix-Vector Computations in Multicast Bus Networks with Communication Delays",
abstract = "In this paper we consider the problem of computing a large matrix-vector product in a network-based distributed computing environment comprising computers equipped with communication co-processors that may be used for communication off-loading. Communication delays, which are significant in such systems, are specifically taken into account. The important contribution of this study is to show that the optimal load partitioning, and the subsequent performance of the network, depends critically on many network parameters and load characteristics. In particular, it is shown that the size of the load plays an important role in determining the performance of the network. We consider only row-wise striping of the matrix in order to better allocate the computational burden among the processors. We derive closed-form solutions to the optimal load partitioning problem and show the existence of optimal load sharing conditions. An important and practically relevant trade-off study, from the architecture point of view, between the number of processors and the bus bandwidth is presented. Several practical load distribution strategies are considered and complete analyses for each of them is presented",
author = "Debasish Ghose and Kim, {Hyong Joong}",
year = "1998",
month = "11",
day = "25",
doi = "10.1006/jpdc.1998.1494",
language = "English",
volume = "55",
pages = "32--59",
journal = "Journal of Parallel and Distributed Computing",
issn = "0743-7315",
publisher = "Academic Press Inc.",
number = "1",

}

TY - JOUR

T1 - Load Partitioning and Trade-Off Study for Large Matrix-Vector Computations in Multicast Bus Networks with Communication Delays

AU - Ghose, Debasish

AU - Kim, Hyong Joong

PY - 1998/11/25

Y1 - 1998/11/25

N2 - In this paper we consider the problem of computing a large matrix-vector product in a network-based distributed computing environment comprising computers equipped with communication co-processors that may be used for communication off-loading. Communication delays, which are significant in such systems, are specifically taken into account. The important contribution of this study is to show that the optimal load partitioning, and the subsequent performance of the network, depends critically on many network parameters and load characteristics. In particular, it is shown that the size of the load plays an important role in determining the performance of the network. We consider only row-wise striping of the matrix in order to better allocate the computational burden among the processors. We derive closed-form solutions to the optimal load partitioning problem and show the existence of optimal load sharing conditions. An important and practically relevant trade-off study, from the architecture point of view, between the number of processors and the bus bandwidth is presented. Several practical load distribution strategies are considered and complete analyses for each of them is presented

AB - In this paper we consider the problem of computing a large matrix-vector product in a network-based distributed computing environment comprising computers equipped with communication co-processors that may be used for communication off-loading. Communication delays, which are significant in such systems, are specifically taken into account. The important contribution of this study is to show that the optimal load partitioning, and the subsequent performance of the network, depends critically on many network parameters and load characteristics. In particular, it is shown that the size of the load plays an important role in determining the performance of the network. We consider only row-wise striping of the matrix in order to better allocate the computational burden among the processors. We derive closed-form solutions to the optimal load partitioning problem and show the existence of optimal load sharing conditions. An important and practically relevant trade-off study, from the architecture point of view, between the number of processors and the bus bandwidth is presented. Several practical load distribution strategies are considered and complete analyses for each of them is presented

UR - http://www.scopus.com/inward/record.url?scp=0007399046&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0007399046&partnerID=8YFLogxK

U2 - 10.1006/jpdc.1998.1494

DO - 10.1006/jpdc.1998.1494

M3 - Article

AN - SCOPUS:0007399046

VL - 55

SP - 32

EP - 59

JO - Journal of Parallel and Distributed Computing

JF - Journal of Parallel and Distributed Computing

SN - 0743-7315

IS - 1

ER -